Skip to main content

Innovative inspection techniques for laser powder deposition quality control

Final Report Summary - INTRAPID (Innovative inspection techniques for laser powder deposition quality control)


Executive Summary:

The INTRAPID project aims to develop three non-destructive testing (NDT) techniques for inspection of parts and components manufactured by Additive Manufacturing Processes, in particular Laser Metal Deposition (LMD). LMD is a technology that has been maturing over the last 20 years and has found application in repair, coatings, hybrid build and 3D near net shape manufacture of small intricate parts that can be used in aero and automobile engines to improve efficiency. Currently this technology cannot be used to supply some of these parts because absence of flaws that might be left in the component during manufacture cannot be guaranteed.

Three NDT techniques (laser ultrasonics, eddy currents and laser thermography) were chosen in INTRAPID as it was expected that each would have limitations or would find a niche in the variety of shapes and materials that would eventually be used for the components. They were chosen because they each operate with a different physical principle, which enhanced the chances of overall success, and they each have the capability to test very small areas of a component, which was essential for this application.

Work has been carried out in developing the reference and test specimens and in each of the non-destructive testing techniques. The first design of the reference proved unexpectedly difficult to manufacture so a series of new designs were added in which laser drilling, micro-electron-discharge machining and diffusion bonding techniques were needed.

Each of the NDT methods has been the subject of considerable development, through modelling in the first instance, and then with practical and experimental data. Each piece of equipment was then required to be adapted to the robotic deployment. The robot used was the KUKA device at TWI Sheffield.

Extensive laboratory trials were carried out with each system, using the reference samples, and procedures were developed. The final devices were tested on the deposited samples, and performance data for the systems obtained in the form of a graph of flaw size against flaw depth below the surface. Overall, the systems were capable of detecting flaws down to 0.1mm in size at the surface up to 0.4mm in size at up to 1.4mm below the surface.

For the final demonstration, all the software and the connections were made to a single PC, and an integrated software package generated to acquire and analyse the indications.

The website (www.intrapid.eu) is available and a video produced and distributed by Youtube.

The project achieved its main objectives and methods of flaw manufacture and the data obtained on flaw detection can be used to choose a method of inspection for the additive manufacturing process for metals and is unique to INTRAPID. The impact of this on industries wishing to use additive manufacture to make critical components can incorporate flaw sizes in calculations to ensure that components are fit for purpose.

The increase in use of additive manufacture will reduce the amount of scrap generated from production processes, and in this way will reduce pollution and thereby allow environmental improvement. Economically the rapid adoption of this new technology will cause growth in manufacturing and thereby increase employment and growth prospects in an expanding industry.

Project Context and Objectives:

Traditionally manufacture of small complex industrial components has been by casting, followed perhaps by machining. This is a relatively slow process, and it will always involve wasting a great deal of material, which in some cases may be high cost alloys. Equally the use of coated materials to increase the performance of individual components has long been recognised, but cannot always be achieved with recent technology.

In the automotive industry there is always a drive for reduced costs and efficiency in manufacturing, while there is a similar driver for lighter weight and more efficient engines to give a competitive advantage in fuel consumption. This results in innovative manufacturing techniques of which LMD is a strong candidate to enable specialised material properties. One example is the deposition of a hard wearing layer (such as bronze) on to a softer lightweight substrate, such as aluminium, for valve seats within a lightweight engine. With 50 million engines manufactured per year worldwide, in around 300 engine production lines, the potential for inclusion of LMD machines within production lines for engines is significant. In these lines some 20,000 prototyping machines using LMD have been installed [1].

In aerospace, aero engine manufacturing has changed its business focus in recent years from selling engines, to the leasing and servicing of engines. This total market is now worth around €40 billion [2]. Research in this area focuses on the development of cost effective repair techniques for ageing engines, and efforts have been put into the creation of new engine components that are designed for repair [3]. One example is the Rolls-Royce Trent 1000 engine family of which the majority of its components are designed for repair. LMD is one of the promising techniques that have been developed for repair of aero engine components, such as compressor blade, turbine blade, blisk and seal segment.

The INTRAPID project aims to develop three non-destructive testing (NDT) techniques for inspection of parts and components manufactured by Additive Manufacturing Processes, in particular Laser Metal Deposition (LMD). LMD is a technology that has been maturing over the last 20 years and has found application in repair, coatings, hybrid build and 3D near net shape manufacture of small intricate parts that can be used in aero and automobile engines to improve efficiency. Currently this technology cannot be used to supply some of these parts because absence of flaws that might be left in the component during manufacture cannot be guaranteed. Quality has been assessed by sample destructive testing, which gives slow feedback and is expensive, and also by computed tomography radiography, but this has to be deployed on a completed object, so the cost of finding a fault is increased. Some form of non-contact NDT would be the desired solution so that parts can be made right first time and with reduced cost; however conventional NDT is generally used for much simpler and larger structures and cannot be adapted easily for these components. The original concept was to have NDT techniques attached to a laser metal deposition head and to follow the deposition with an inspection as it was deposited.

Three NDT techniques (laser ultrasonics, eddy currents and laser thermography) were chosen in INTRAPID as it was expected that each would have limitations or would find a niche in the variety of shapes and materials that would eventually be used for the components. They were chosen because they each operate with a different physical principle, which enhanced the chances of overall success, and they each have the capability to test very small areas of a component, which was essential for this application.

The specific project objective is to develop the three inspection methods to a stage where prototype systems have been integrated into a production process and to complete a demonstration of this. If the deposited or melted layers are 0.5mm thick, for example, the target sensitivities for the inspections are: (1) detection of 0.25mm diameter pores 0.25mm below the surface or less, (2) lack of fusion 0.5mm below the surface 1mm long and (3) to inspect curves in the layer of 5mm radius or greater.

The main objectives of this project were to develop the three techniques, show that they work and deploy them in a realistic environment. The project was divided into eight work packages. Each of these is described below, and the objectives given.

Work Package 1 included as an initial review, preparation of a specification document, which was to include the reference and test samples, and all the inspection equipment. It also included production of a set of machined reference samples to be used in development of the equipment. The purpose of the work package is to set up the project for the equipment development.

In Work Package 2 the objective was to manufacture a set of laser weld deposited test samples with flaws induced in a controlled way. The reason for generating these samples was to use them to establish, (in Work Package 7) performance data for each of the three methods.

In Work Package 3, the development of laser ultrasonics was considered and the objective was to build a system and determine its operating parameters. The steps to doing this included an initial feasibility stage including modelling of the process and transmission of the ultrasonic waves in small structures and some practical feasibility trials with reference samples with a view to establishing the main features of the prototype instrument. It was expected that a laser system would be needed for detection but other air coupled techniques were also to be tried and included if feasible.

The second stage was to purchase the components and build a system which could be ultimately mounted on a deposition robot. It was planned to carry out some signal analysis and processing with a view to being able to automate flaw detection. Finally once complete an operating procedure would be produced.

The eddy current system would be developed in Work Package 4. The first stage of this was to investigate the sensors that could be used. Magnetic and cross-wound probes would be tried. This task was needed because the small size of the samples meant that indications would be expected from features such as edges, which may make the inspection difficult. This would be followed by construction of any adaptive circuits needed to join the sensors to a standard instrument. The output of this would be processed suitably to enhance signals. Automatic flaw recognition was again the aim. An operating procedure would be developed for the system.

Work Package 5 was the development of the laser thermography system. This started with initial feasibility trials with available equipment, followed by a system design stage. The prototype was then to be constructed. It was anticipated that he output of the system would be an image or series of images from a thermal camera, and it was necessary to design a method of analysis of the images to recognise features related to flaws. It was expected that this would be done by comparing the image from good areas to that from a defective area. Again an operating procedure would be produced.

In order to operate the above systems with a deposition system a work package to integrate the systems was needed and this was Work Package 6. The objective was system integration and it was necessary to do this with several different technologies, mechanical, electrical and software. It was expected that each method would require a mechanical adaptation to the robot, and that each system would require a different path to that required by deposition, and that these would themselves be different. The electrical compatibility would be needed for driving the circuit and acquiring which would be transferred to a master computer. Software to drive each system would also be needed to link to the robot movement.

It was intended that each system would then be tested for functionality and calibration within the integrated system.

Work Package 7 was the testing the various NDT techniques and producing detection (POD) Curves, followed by production trials. The NDT tests would be carried out on the test samples developed in Task 2, which would then be sectioned, and the compared with the system outputs.

Finally the tests would be carried out with the complete system deployed from the robot. This would be a complete operating prototype which could be tested for each method.

Exploitation of the project to the benefit of the SMEs is intended at the outset, and the plan and preparation for this would be carried out in Work Package 8. The draft plan would be produced in two stages at the 9 month and 24 month points of the project and would include the main dissemination activities as well as the exploitation plan. Other tasks within this work package included production of publicity material and a suitable training course, and another task was to ensure that links to relevant websites were made. Also a project video was intended to be produced.

REFERENCES

1 www.madeforone.com/Articles/index.php/technology/rapid-prototyping market
2 The Engine Yearbook 2010, published by UBM Aviation Publications Ltd. Pp 2-7
3 B. Morey, ”Aerospace leads in Additive Manufacturing”, Aerospace Engineering, August 2012 pp18-21.

Project Results:

The project was divided into seven technical work packages. These were WP-1, Specification and reference samples; WP2 Test Samples:WP3 Laser Ultrasonic System;WP4 Eddy Current System;WP5 Laser Thermography system, WP6 System Integration and WP7 Laboratory and Site Trials.

Note that the Figure Numbers, Table Numbers and Equation Numbers refer to the pdf attachment to this document: These are Figures 1-107 (pp1-61), Tables 1-11 (pp62-66), Equations 1-31 (pp67-69).

Specification and Reference Samples

The first aim of Work Package 1(Task 1.1) was to get together background knowledge of the project partners and specify and/or confirm with the input of the project end-users the parameters to be focussing on during the course of the project. Amongst others, the most important ones to be defined in the early stages of the project are presented below.

• The type of materials, powders, sample geometries and flaws that need to be produced for test samples to be manufactured (this is directly linked to Work Package 2 which consists in the manufacture of test samples).
• Target sensitivities and geometrical/mechanical restrictions for the NDT systems. Also, the mechanical details of the laser deposition system should be obtained for each NDT Work Packages and for final future reference at the integration stage (Work Package 6).
• Samples taken as reference standards for the NDT techniques using sheet materials were designed and manufactured. These samples were different from the test samples because they will contain known artificial defects such as flat bottom holes that were machined instead of being introduced by variation of the process parameters. This allowed development and validation of the NDT techniques to be developed so that they could better detect real defects that may be caused in the Laser Metal Deposition (LMD) process.

Task 1.1 of Work Package 1 also required the specification of test samples and test objects. This topic was approached at the kick off meeting on the 13th and 14th October 2011. It was suggested that the initial work should be kept as broad as possible (i.e. to use a material of interest to both the automotive and aerospace sectors initially). Therefore, it was agreed to use an Inconel type material to satisfy this requirement. Inconel alloy 600 was chosen for to produce sets of reference samples. Inconel alloy 600 is usually well-known for its high corrosion resistance to a variety of corrosive conditions and its wide use in the industry.

Test samples with simple geometries and flaws should then be made. Typically the deposit would be 0.3mm wide by 0.5mm thick.

For aerospace applications the flaw size of interest is in the range around 100 µm, although some sources quoted 10 µm but according to studies already carried out there is little change in mechanical properties for such small flaws.

It was agreed with the participants that later on in the project additional test samples would be produced and more specific to the Toyota valve seat application hence, for instance, involving materials such as copper nickel alloy onto to an aluminium substrate or some other application of interest to Toyota.

Figure 1 below gives a very schematic idea of the valve seat part involved with the powder deposition process at Toyota. The 1.2 mm refers to the width of the engagement area of the valve seat on the powder deposit after machining it down to the correct shape. The copper layer is deposited against circular groove and a lot of copper cladding material in the depth direction is preserved through the machining stage meaning that only a limited amount of the cladding is machined away. The machining is done in the vertical (red dashed line) and horizontal (blue dashed line) planes cutting through the powder deposit (in yellow). The engagement surface of the valve corresponds to what is left after the machining process was carried out and located at the top of the powder deposited region.

It was agreed that the technique for flaw production in WP2 should be developed on single straight passes on to a substrate. Porosity and lack of fusion, either to the substrate or adjacent deposits were the flaws to be considered. The surface finish (as deposited or machined) also needed to be specified and it was suggested that both should be considered. However, note that the inspection stage is preferred to occur before the machining process, therefore the” as deposited” surface is of more significance.

Therefore the inspection requirements are as shown in Table 1.

Task 1.2 of Work Package 1 determined the specification of the inspection systems for the 3 non-destructive testing techniques developed i.e. laser ultrasonics, eddy currents and laser thermography. For this purpose, the ideal performance requirements of the NDT equipment in terms of flaw detection had to be related to the detection requirements above. This task also involved the understanding of how to combine and integrate the laser deposition system with the NDT systems. This meaning that the mode of operation of the laser deposition system and the potential effects and constraints generated by this latter on the NDT system had to be identified. The laser deposition system was examined and the constraints of access together with operational, mechanical, electrical and software requirements that would be placed on the installation of the NDT systems were investigated.

The general requirements concerning the inspection unit containing all or separately the 3 NDT systems were specified. It was initially thought that the delivery system for inspection should be within 100mm x 100mm x 200mm box but this was relaxed later in the project, it was noted that supporting equipment should fit into the robot bay (5m square, approximately) and control should be from the outside. The weight of the delivery system could not exceed 10Kg.

It was expected that the inspection system will be deployed after the deposition system by rotation of the head on the robot.

Consideration of the specification for the individual inspection systems was carried out as far as was possible at the time but more detail is given in individual WPs. Table 2 gives the expected constraints of each system.

The purpose of the reference samples produced for the project is threefold (1)to enable an initial estimation of the feasibility of each NDT technique and its sensitivity to be obtained (2)to enable optimisation of the techniques to be achieved(3) to enable designs of calibration samples for individual applications to be developed

There are several designs of reference samples now available to the project.

Type 1 (as agreed at the consortium meeting 10/9/12 + 1 supplementary design): These represent a general thin structure, and the design is shown in Figure 2. A group of these samples is shown in Figure 3. A typical flaw produced by the laser machining process is shown in Figure 4 and the summary of flaws produced in this design is shown in Table 3.

Type 2 (as agreed at the kick off meeting): The intention of the Type 2 samples is to establish capability at the first critical deposition layer, and so includes a substrate as part of the design. It was found that laser machining could not produce flaws in these locations due to access problems. Type 2 samples were therefore made by various methods (a) high intensity electron beams (b) micro EDM drilling (c) diffusion bonding with machining.

Figure 5 gives the basic design and Figure 6 is a group picture of the electron beam manufactured samples. Figure 7 gives an example of the flaw produced by the electron beam, Figure 8 a group of the EDM produced samples and Figure 9 is an example of a flaw produced by micro-edm drilling, and Figure 10 one produced by diffusion bonding. Tables 4 and 5 show the range of flaw sizes and depths produced.

Types 3 and 4 were made by the diffusion bonding process (to represent the valve seat applications). Figures 11 and 12 show the designs for these.

Test Samples

The production of test samples was carried out in WP2. Task 1.1 established the deposition parameters and equipment to be used. The deposition equipment used in these trials incudes (a) Kuka KR30HA High Accuracy robot,(b) IPG 7000 7kW fibre laser, (c) ILT Fraunhofer 3 beam powder delivery nozzle (d) Sulzer Metco powder feeder system.

The materials used include (i) Inconel 718 powder (size 20 – 53 µm),(ii) Inconel 718 x 3mm thick substrate material,(iii) Inconel 600 powder (size 44 – 105 µm),(iv) Inconel 600 substrate material,(v)Cast aluminium blocks grade AC2C (supplied by Toyota),(vi) Copper-Nickel alloy powder (supplied by Toyota),(vii) Aluminium alloy (Al-Mg-Zr) powder

In Task 2.2 a number of methods for manufacturing reproducible defects were considered, for example development of porous structures, production of cracked deposits, development of geometries and toolpaths to promote defects.

Previous work has demonstrated that the deposition of poor quality powder can reliably produce porosity within the deposit. Powder which has been manufactured with internal pores within the individual powder particles has been shown to produce significant porosity within the deposit (Figures 13). However, most of this porosity is below 50µm in size and is therefore too small to meet the specifications of the proposed NDT systems.

The production of porous deposits by using non-ideal deposition parameters is also a possible strategy to develop defects. This involved the monitoring of the effect of changing deposition parameters to produce defects. A series of blocks were produced by building multi-track, multi-layer deposits which were subsequently sectioned and assessed for porosity (Figure 14).

The development of geometries and toolpaths to promote defects initially used In 718 material rather than the In 600 material. The trials involved momentarily interrupting the beam during deposition to promote a lack of fusion defect and then developing the parameters to vary its size and shape. The final procedure chosen was to simulate a momentary laser failure by depositing a single track width but interrupting the deposit and producing gaps of 1.0mm 0.8mm 0.6mm 0.4mm and 0.2mm along the length of the deposit (Figure 15). Further deposits of a number of continuous layers on top then mask the evidence of a discontinuity. In this way, defects were successfully produced (Figure 16).

Work then continued using Inconel 600 powder and substrate, to do this process parameters were developed for the new material. A number of single track deposits containing defects were successfully manufactured.

In addition to single line tracks, further deposits with defects were produced, but in this case the defects were designed to be encased within specimen blocks of deposited material.

Three adjacent track widths of interrupted deposit were laid down, with a number of full deposited tracks at each side to form a 25mm square block of material. 2 layers were deposited to this design (Figure 17).

2 full continuous layers were then deposited over the top to form a 4 layer high block of 25mm x 25mm square containing 2 defects.

It was noted however that it was required ultimately to carry out the NDT on a single layer, and that the methods developed were visually detectable (see Figure 15). Therefore further tests were carried out to try and produce voids inside the deposit without them showing on the surface. Various inserts were tried but eventually it was found that a small hole suitably drilled in the substrate produced the required voids, although there was a limit to the hole size with which this could be done. Figure 18 shows a flaw produced by this method and Figure 19 shows a radiograph of a similar flaw. Radiography was ultimately preferred to sectioning as a method for confirming the presence of a flaw as sectioning to find such small flaws was difficult and the sample was also retained for further tests.

A range of test samples in Inconel 600 was then produced, and these were machined to give different depths of the flaw below the surface.

To produce flaws in the aluminium samples initially aluminium powder was deposited onto aluminium substrates and it was demonstrated that porous deposits could be produced as the density of the deposited material could be varied between 95% and 70%. However, the results were random and difficult to predict and other ways of developing defects were attempted.

With these trials, the copper-nickel alloy and the aluminium alloy A2C2 as supplied by Toyota were used to produce cracked deposits. The A2C2 cast blocks were machined to improve the surface finish and also some of the blocks were pre-machined with grooves to depths of between 1.0 and 2.5 mm.

A number of deposits were produced (Figure 20) using the copper-nickel alloy powder and the deposition parameters were varied to try to produce a consistent and reproducible defect. The results varied from good deposits with very fine surface cracks as shown in Figure 21, through to cracks of approx. 0.5 mm length at the interface between the deposit and substrate, to deep cracks of up to 1.5 mm long into the base of the deposit. The deeper and longer cracks occurred when depositing into the bottom of the deeper grooves. There was also evidence of gas porosity in the base of the grooves, probably caused by the entrapment of gases due to the groove geometry (Figure 22).

SEM and EDX investigations were undertaken to identify the precise metallurgy of the defects produced to give a better understanding of the cracking mechanisms involved with these two materials (Figure 22).
In all these cases, the copper-nickel deposit was crack and porosity free, indicating that the best opportunity to produce a defect would be at the interface between the two materials. It was therefore decided to add a second track of deposited copper-nickel at the side of and parallel to the first track to attempt to trap the defect between the deposits.

Further work was also carried out using different process parameters to produce flaw conditions. Deposition of the phosphor bronze onto milled aluminium substrates was performed in using same deposition equipment. Following standard procedure for deposition onto highly reflective materials, the laser optics were tilted 11° along the long axis of the substrate, to prevent back reflection of laser light into the optical cavity.

Substrates were prepared with 2mm x 2mm and 4mm x 4mm grooves running the length of the substrate edge as shown in Figure 23. These substrates were angled at 45° so that the laser beam hits the corner of the milled groove, as well as both faces.

Calibration of laser power was performed using OPHIR power meter and calibration of spot size was conducted using Spiricon digital beam profiler. Deposition was performed using localised shielding, delivered coaxially to the laser beam through the central aperture of the nozzle (Figure 24).

The fixed parameters for the deposition were (a) Spot size = 2mm (b) Scanning speed = 0.008 m/s, (c) Carrier gas pressure = 1.5 bar, (d) Carrier gas flow rate = 7 l/min argon (BOC cryopure),(e) Nozzle shielding gas flow rate = 20 l/min argon (BOC Cryopure) and Deposit length = 30mm.

The variable parameters and the results obtained were as follows

1. 2 KW, 40% powder feed rate – filled the groove well, but was slightly concave at the start, most likely due to the rapid heat absorption at the start of the deposit, which grew less as the laser heated up the surrounding material. A cross section of this deposit is shown in Figure 25.
2. 2 KW, 50% powder feed rate – didn’t filled the groove as well as deposit one, but had a more bronze colour to it. Got a slight undercut and centreline ridge effect. A cross section of this deposit is shown in Figure 26.
3. 2 KW, 60% powder feed rate – Same as with deposit 2. Higher powder feed rate may be causing melt pool shielding and loss of power to the substrate. A cross section of this deposit is shown in Figure 27
4. 2.2KW 50% powder feed rate – Better infill than deposit 3, caused slight melting of the groove edges due to higher power. Deposit was silvery in colour, most likely due to good deposit shielding and better melting of alloying constituents. A cross section of this deposit is shown in Figure 28
5. 2.2 KW, 60% laser power – Deposit appeared lumpy along the length, possibly due to power pulsing or melt pool surface tension. A cross section of this deposit is shown in Figure 29.

All deposits were sectioned, polished, etched and subjected to optical microscopy before further trials were performed.

Based on the limited optical microscopy performed, sample 5 displayed the best deposit integrity and shape and was therefore chosen as a basis for further tests.

A test sample was prepared with 30mm long sections and different parameters as shown in Figure 30. The sample parameters are given below:.

6. Accidental deposit- not used
7. Accidental deposit-not used
8. Layer 1 - 2200W, 60% powder feed rate, Layer 2 – 2200W, 60% powder feed rate. Sample had good shape and surface finish.
9. Layer 1 – 2200W, 60% powder feed rate. Layer 2 – 2200W, 50% powder feed rate. Sample had slightly lumpy surface, and still does not properly wet the edges.
10. Layer 1 – 2200W, 60% powder feed rate. Layer 2 – 2200W, 40% powder feed rate.Still lumpy.

The cross sections are given in Figures 31-33.

A further group of samples with different parameters was also made.

11. – 0.008m/s. 2000W laser power, 40% powder, 10l/min nozzle gas, 30mm long. – Note: forgot to seal around cover slide holder, so gas delivery to melt pool was lower than anticipated. (Figure 34)
12. 0.008m/s. 2000W laser power, 40% powder, 20l/min nozzle gas, 30mm long. Note: sealed around cover slide holder, deposit looked much better (Figure 35).
13. 0.008m/s. 2000W laser power, 50% powder, 20l/min nozzle gas, 30mm long (Figure 36).
14. 0.008m/s. 2200W laser power, 40% powder, 20l/min nozzle gas, 20mm long (Figure 37).
15. 0.008m/s. 2200W laser power, 50% powder, 20l/min nozzle gas, 20mm long (Figure 38)
16. 0.006m/s. 2200W laser power, 40% powder, 20l/min nozzle gas, 20mm long(Figure 39)
17. First half same as 13, second half same as 14. Sample 14 produced a fractured deposit bead.

On completion of this work a further two samples were made, one used 30mm of procedure as used in Sample 13 followed by 30mm of procedure 14. This is shown in Figure 40

A final sample T2 was made with procedures 13 and 14. 30mm of this was ground away as shown in Figure 41.

Development of Laser Ultrasonic Prototype

The purpose of Work Package 3 Task 3.1 was to carry out some preliminary feasibility studies for the laser ultrasound methods. The feasibility of the technique was studied by modelling of the ultrasonic waves in the thin sections. This was carried out to assist in the determination of some of the required laser parameters (e.g. pulse duration, energy, beam shape). Preliminary numerical studies were useful to decide system specifications for the generating and the receiving laser equipment; they are developed considering the reference sample geometries. A practical feasibility study was carried out on the reference samples.

Firstly it was necessary to establish the physical properties of the material for the numerical study. INCONEL® (nickel-chromium-iron) alloy 600 is a standard engineering material for applications which require resistance to corrosion and heat. The alloy also has excellent mechanical properties and presents the desirable combination of high strength and good workability. The high nickel content gives the alloy resistance to corrosion by many organic and inorganic compounds. Chromium confers resistance to sulphur compounds and also provides resistance to oxidizing conditions at high temperatures or in corrosive solutions. The alloy is hardened and strengthened only by cold work. The versatility of INCONEL alloy 600 has led to its use in a variety of applications involving temperatures from cryogenic to above 2000 °F (1095 °C). The mass density is and the melting range between 1354 °C and 1413 °C. The following table shows the thermal and the elastic proprieties of this material [2].

To prepare the model it is important to take into account the rules of explicit numerical analysis, regarding the size of elements in the mesh. In order to make some preliminary calculations, it is convenient to take in account the physical parameters of INCONEL for a single temperature.

Considering the values of physical and thermal properties for T=20 °C, the velocities of the longitudinal (c1), transversal (c2) and superficial wave (cR) are given in Equations (1)-(3):

In ultrasonic testing, it is necessary to make a decision about the frequency that will be used. Changing the frequency when the sound velocity is fixed will result in a change in the wavelength of the sound. The wavelength of the ultrasound used has a significant effect on the probability of detecting a discontinuity. A general rule of thumb is that a discontinuity must be larger than one-half the wavelength to stand a reasonable chance of being detected. Sensitivity and resolution are two terms that are often used in ultrasonic inspection to describe a technique's ability to locate flaws. Sensitivity is the ability to locate small discontinuities. Sensitivity generally increases with higher frequency (shorter wavelengths). Resolution is the ability of the system to locate discontinuities that are close together within the material or located near the part surface. Resolution also generally increases as the frequency increases. The wave frequency can also affect the capability of an inspection in adverse ways. Therefore, selecting the optimal inspection frequency often involves maintaining a balance between the favourable and unfavourable results of the selection. Before selecting an inspection frequency, the material's grain structure and thickness, and the discontinuity's type, size, and probable location should be considered. As frequency increases, sound tends to scatter from large or course grain structure and from small imperfections within a material. Cast materials often have coarse grains and other sound scatters that require lower frequencies to be used for evaluations of these products. Wrought and forged products with directional and refined grain structure can usually be inspected with higher frequency transducers.

Since more things in a material are likely to scatter a portion of the sound energy at higher frequencies, the penetrating power (or the maximum depth in a material that flaws can be located) is also reduced. In the traditional technique of ultrasonic testing, where ultrasound are generated by piezo-electric transducers with narrow spectrums, the frequency also has an effect on the shape of the ultrasonic beam. Beam spread, or the divergence of the beam from the center axis of the transducer, it is affected by frequency. Unlike traditional ultrasound generators, a thermoelastic laser source generate ultrasound in a large range of frequencies. Now it is important to understand what features a laser source must have, in order to maximize the relief of flaws with size minor than 100 μm.

Proposing a minimum size of ϕ=50µm for the flaws that we aspire to detect, in order to stand a reasonable chance of being detected, the ultrasound field have to contain waves with wavelengths smaller than twice this size:

Considering the waves velocity, indicated in Eq. (1-3), in accordance with the main rule of sound propagation, Eq. (4) gives the inferior limits for the frequencies shown in Eqs (5-7)

In explicit dynamics procedures temporal and spatial resolution of the FE simulations are of fundamental importance for the stability and the accuracy of the solution. An approximation to the stability limit is often written as the smallest transit time of a dilatational wave across any of the elements in the mesh as in Eq (8):

Laser generated ultrasound wave requires a very small time step in order to accurately resolve their high frequencies components. Accurate solutions can be obtained provided that the conditions of Eq (9) are met. This shows that to resolve the ultrasound propagation with the main component at 60MHz, a stable time increment is of the order of 10-10s is needed.

A reasonable spatial resolution of the propagating waves can be obtained when the size of the finite elements is at least 1/10 of the shortest wavelength to be analysed (Eq (10).

So, looking at Eq. (4), a reasonable spatial resolution can be obtained when the size of the finite elements is 0.01mm as shown in Eq (11).

In explicit dynamic simulation a thermoelastic source is replicated, applying a surface heat flux to a certain portion of a model with the same shape of the laser beam profile. So we need to estimate a reasonable value of power density, applicable to the INCONEL sample surface without causing any damage of it. In other terms we need to find the limit between the thermoelastic and the ablative regime of generation. We can calculate the temperature distribution resulting from the absorption of a laser pulse at the surface of a metal. We assume that the thermal properties of the metal are independent of temperature, that a local thermal equilibrium is established during the pulse, and that negligible energy is lost from the surface radiation. The differential equation for heat flow in a semi-infinite slab (half-space) with a boundary plane at z=0 is given by Eq (12).

The boundary condition are that T(x,y,z,0)=0, and T→0 as z→∞, and no heat flux crosses the z=0 plane. For most practical situations in which a short-pulse laser irradiates a metal over a limited area, the depth to which heat is conducted during the duration of the pulse is much less than the area, so that a 1D treatment is appropriate [1]. Thus Eq 13 applies.

If the absorbed laser flux density is I0, uniform across the irradiated area, and the laser is switched on instantaneously at zero time, then, for high absorption at the surface the solution of equation (13) is given in Eqs (14-16).

Suppose the laser generates a square pulse of duration t0, i.e. is switched on at t=0 and switched off at t=t0. Eq. (14) must still hold for t<t0. However, for t>t0 a second term must be added to represent switching off as shown in Eq (17). It can also be seen that for z=0 Eq 18 applies.

Replacing t=t0 (pulse duration) and T=Tf(melting temperature) in Eq. (16), we can find (Eq (19)) the maximum absorbed laser flux density I0max for thermoelastic sources and supposing a laser pulse duration of t0=5.1ns and using the parameters given in Table 2, for INCONEL alloy 600 gives Eq (20).

This is an important result. Applying this value of surface heat flux to the numerical model, the entity of the simulated displacement will be comparable with the real displacement, under the condition of thermoelastic regime. Figure 42 shows a comparison between the theoretical temperature rise, in agreement with Eq. (16) and (18), and the simulated pattern.

The numerical calculated temperature pattern is in good agreement with the theoretical prevision, especially for the first 20 ns in which there is the temperature rise and the major fall after the laser pulse.

In order to consider sources of different shapes a further analysis was carried out. The area of the beam, as it leaves the laser, is not very important because it can readily be changed by standard optical means. The area of the incident radiation on the sample surface is, however, much more important, because it affects both the power density (and hence the threshold for ablation and surface damage) and the dimension of the ultrasonic source. Many commercial lasers produce beams with diameter of the order of a centimetre, which coincidentally is comparable with the diameter of many ultrasonic transducers. Particular ultrasonic beam characteristics can readily be obtained by changing the irradiated area from a circular region, to a point, to a line, to a circular line. Phased arrays of small irradiated areas offer the maximum flexibility in ultrasonic directivity. Here we want to compare the ultrasound fields produced by punctual, linear or multi-linear sources, in order to increase the generated frequency range and find the best solution for detecting defects of 50 μm.

Figure 43 shows the model we used for the comparison of the ultrasonic fields produced by different type of sources. The displacement is registered in three points: P1, P2 and P3. P1 is 1 mm far from the centre of source on the same surface, P2 is at the same distance but on the opposite side and P3 is on the epicentral direction. The used value for the surface heat flux was that indicated in Eq. (20). Three types of source were simulated: a punctual source with circular perimeter and a diameter of 100 μm, a linear source 100 μm wide and a multi-linear source composed by 4 lines with width of 50 μm and 50 μm spaced. The same surface heat flux, of 12.4 MW/cm2, was applied to the three sources.

The dimensions of the multi-linear array were calculated using the conclusions presented in [3-4]. Consider a laser light line array, in which w and 2h are the width and length of each line, respectively, and d is the spacing between lines. A non-dispersive single mode guided wave generated with a laser light focused to a single point may be represented by a simplified triangular wave. The frequency of the generated signal depends on different parameters: the laser pulse length, shape of the illuminated region, and the physical properties of the material. For a light pulse geometry of finite width and length, such as a line, there is an infinitesimal delay between the elements that comprise the illuminated region. A continuous series of signals arrive between the first and last arrivals. In the case of a line array, the signal produced by a single line is repeated at intervals corresponding to the spacing between lines d and the number of lines NL Three fundamental frequencies emerge from this arrangement, depending from the length of the lines, their width and spacing, but the main frequency can be valuated with Eq (21).

As was shown in Eq. (3) and (7), the velocity of Rayleigh waves is cR=2.88mm/µs and the necessary frequency to detect defects is fR =28.8MHz; so the separation distance between the lines must be given by Eq (22). Choosing 2w=d, we obtain in Eq (23) the width of the lines of the multi-linear source:

The explicit simulations were developed using 3D elements with sides of10µm, in agreement with Eq. (11), and simulating the ultrasound propagation for a period of 1µm. The displacements and the frequency spectrums are showed in the results shown in Figures 44-48.

The frequency spectrum relative to the ultrasound generated by the multi-linear source and received at P1 shows an interesting peak at about 29 MHz and a certain attenuation of lower frequencies. This fact is in agreement with the previous calculations and means that a linear array increases the energy of surface waves with wavelength of the order of λ = 0.1mm improving the detection of superficial flaws. It is known that surface waves penetrate in the volume only for a wavelength, so further researches have been developed in order to understand if, thanks to the conversion of modes, this type of source could result convenient also for the detection of flaws inside the volume of samples.

With the 3D model it was not possible with current resources to use elements with sides smaller than 10µm. The limits of our calculations were confirmed, but frequencies higher than 51.6MHz 30.9MHz and 28.8MHz respectively for longitudinal, transversal and surface waves, were not as accurate as the lower frequencies. Moreover the wavelength of 10µm is only the greatest value to have possibility of finding flaws of 0.05mm. In order to have a good accuracy of the simulations along the range between 0-100 MHz, the possibility of using a plane model was valuated. Creating a plane-strain model, we succeed to use elements with sides of 1µm (an order smaller than 10µm). This fact means that a greater number of higher frequencies can be followed with a great accuracy; using the 2D model, the new limits are given in Eqs(24-26).

With the 2D model, the ultrasound fields produced by the linear and the multi-linear source were replicated. The displacements and the frequency spectrums are showed in the following tables.

The results obtained by the simulations with the plane model show again that the frequency spectrum relative to the ultrasound generated by the multi-linear source and received at P1 has an interesting peak at about 29 MHz and a strong attenuation of lower frequencies. The following table shows the map of the displacement magnitude at different values of time, both for the linear source and the multi-linear source, obtained from the simulations with the plane model.

Definitely, by the previous preliminary calculations and the study of different types of sources, some important information was obtained. A laser pulse with length of the order of some nanoseconds can generate broadband ultrasound. Considering a length of 5.1 ns, the limit of the absorbed thermal power for a thermoelastic ultrasound generation was estimated to be equal to 12.4 MW/cm2. It is likely that the optical power emitted by a laser can be greater than this value, because of the material surface reflectivity. Simulating the ultrasound field produced by sources of different shapes it is possible to assert that all of them generate waves in the field of interest (0-100 MHz), but only the multi-linear array source allows concentration of the ultrasound power in a more narrow range. Choosing an appropriate spacing between lines, a consistent part of the Rayleigh wave power has been concentrated around the theoretical value of 28.8 MHz, suitable to find defects (intended as porosities) with dimensions greater than 50 μm.

The first concept for laser ultrasonics was to have a source and receiver close together (see Figure 49), the NDT technique could be linked to the LMD nozzle (Figure 50) and check the absence of flaws of the deposition soon after its production.

Various configurations of laser source and receiver can be used when inspecting samples, which closely resemble those used for conventional probes [1]. Thus the first configuration has the source and receiver on opposite surfaces of the sample so that the probing ultrasonic wave field samples the material in transmission. This configuration is advantageous for laser ultrasonics since the sample isolates one optical system from the other. The transmission configuration is particularly useful for detecting the presence of planar defects (such as lamellar defects and inclusions in hot rolled steel) which cause a loss of signal when interposed between source and receiver. It is also a good geometry for inspection in the ablation regime, since this source radiates high-intensity compression waves normal to the source surface. In the thermoelastic regime little energy is radiated normal to the surface so that this first configuration would give disappointing results. Shifting the receiver, according to the directivity of the source, would therefore be preferable for a transmission measurement in the thermoelastic regime. Both of these configurations, however, require access to both sides of the specimen. This is not possible in LMD process, so that the ability to inspect from one surface is essential. Two single-sided probe configurations are possible. Firstly, the source and receiver could be coincident in the same point, to emulate a pulse-echo conventional arrangement. Whereas this is the simplest way to use a conventional probe, it is more difficult for laser ultrasound, since separate laser systems must be employed for generation and reception, and care must be taken to minimize breakthrough from the high-power source laser to the detection system. The second (see Figure 49) is obviously ideal for detecting surface-breaking defects, since it permits Rayleigh waves to be used in the inspection, and it is also useful for time-of-flight measurements of the ultrasound diffracted by volumetric defects.

Adopting the conceptual configuration of Figure 50, the resultant NDT system is adapted to obtain a LMD-NDT equipment. In particular, we can imagine the NDT system moving together with the LMD nozzle and checking the presence of flaws in the deposited material at a fixed distance from the deposition point. If this distance isn’t too small the temperature of the checked zone can be also sensibly minor than in the deposition point, helping the generation of laser ultrasound based on the thermal expansion of the irradiated surface. In the numerical model the sample will be supposed at 20 °C, but further more precise simulations can be obtained studying the temperature curve in function of the distance from the deposition point. Figure 46 shows the described system depositing and checking a layer of material above a thick substrate. The generating laser beam hits the deposit at a distance of L from the LMD nozzle and the receiver registers displacements at a distance of d from the ultrasonic source. Supposing d=1 mm, the numerical feasibility study consisted in evaluating the capability of this system in giving a signal about the presence of small porosity across the deposit.

The geometry of the sample, designed in Work Package 1, has been considered to develop the feasibility study. Using the symmetry condition it is possible to build a numerical model of a half of the real sample; Figure 5 itself shows half of it. As discussed in the previous paragraph, in explicit simulations a reasonable spatial resolution of the propagating ultrasonic waves can be obtained when the size of the finite elements is at least 1/10 of the shortest wavelength to be analyzed. It was calculated that a reasonable spatial resolution can be obtained when the size of the finite elements is le ≤ 0.01mm In order to simulate the inspection procedure showed in Figure 5, the minimum volume that has to be meshed with such small elements was estimated to be that of Eq (27).

Dividing it for the volume of a single regular tetrahedral element, with sides of 0.01 mm, the resultant number of elements is 222 millions as given in Eq (28):

This great number of element is not sustainable by current computers unless we accept a very long time of analysis, for the great quantity of memory and time required at each temporal increment. So we studied a way to simplify the model without decreasing the accuracy of the numeric result, for our purposes. The problem can be simplified modelling the section lying on the symmetry plane with a 2D model. The planar representation of the problem has been the solution to minimize the number of elements; surely it adds some approximations due to the plane strain model that can’t give any information about the wave reflections on the lateral faces of the deposit, but they seems not to be very relevant for ours goals. Moreover, the renounce at obtaining a 3D model is tolerable in comparison with the advantages descending from the use of planar model. Using square elements with side of 0.01 mm, the mesh of the entire section would be composed by only 1 million of elements. It was possible to use smaller square elements with sides of 0.002 mm, resulting a very large bandwidth for the obtainable frequencies. With a 2D model it is possible to simulate both a linear and a multi-linear source, but not circular (or punctual) source; in order to develop a feasibility study, this isn’t a great problem being the ultrasonic field produced by the linear source very similar to what is produced by a circular source, as was seen in the 4th paragraph.

Simulating the ultrasound propagation produced by a linear source in a standard model (without flaw), the displacement showed in Figure 51a was collected at the receiving point. Figure 51b reports its frequency spectrum. Thus, inserting a hole with diameter of 50 μm in the mid-length of the deposit, the propagation of the ultrasound waves was simulated at different distances from it in order to measure the changes in the received displacement. For the sake of clarity, we insert a datum axis (x) in the direction of the NDT system progress and place its origin in correspondence of the flaw. Taking the midpoint between the ultrasound source and the receiver as point of reference, the inspection procedure was simulated from x=-1.5mm to x=1.5mm to, with increment of 0.5mm between two consecutive inspections. In this way, a total of 7 simulations were developed across the flaw and each displacement patter has been compared with the displacement collected by the simulation in the standard model.

Figure 52 shows the comparison between the displacement simulated in the model without defect and the displacement collected at seven different points across the defect (a hole with diameter of 50 μm). The information about the presence of the flaw arrives at different values of time, depending on the position of the NDT system and the distance covered by ultrasound waves. For each position, the difference between the two patterns is plotted on the right. At ±1.5 mm from the defected section (see Fig. 52a and Fig. 52g), it is yet possible to notice a small difference that increases in value when the inspection is simulated in the positions between ±1.0 mm. In this last interval the difference pattern assumes an amplitude greater by some nanometres.

Following the same procedure, was simulated the ultrasound propagation produced by a multi-linear source in a standard model (without flaw). The displacement showed in Figure 53a was collected at the receiving point. Figure 53b reports its frequency spectrum, exhibiting a peak of the ultrasound power in the narrow range around 29 MHz, as previously discussed.

Thus, inserting the 50 μm hole in the mid-length of the deposit, the propagation of the ultrasound waves was simulated at different distances from it in order to measure the changes in the received displacement. Adopting the same reference system used in the previous analysis with the linear source, even in this case, a total of 7 simulations were developed across the flaw.

So, considering a laser light line array, in which is the width of each line and d is the spacing between lines, and wishing to concentrate most of the energy around the fundamental frequency of fR=28.8MHz the same multi-linear source of the Paragraph 4 was simulated. The separation distance between the lines has been set to d=100µm and, choosing 2w=d, the width of the lines was w=50µm. The density of surface heat flux used for the multi-linear source was the same that was previously applied on the single line source; this means that the total energy increased twofold thank to the double irradiated area. Under these conditions, Figure 54 shows the comparison between the displacement simulated in the model without defect and the displacement collected at seven different points across the defect (a hole with diameter of 50 μm). As it was noticed in the previous set of plots, the information about the presence of the flaw arrives at different values of time, depending on the position of the NDT system and the distance covered by ultrasound waves. For each position, the difference between the two patterns is plotted on the right. Now it is possible to state that the amplitude of these difference patterns appears to be equal to the amplitude and, despite of the double energy, in terms of signal magnitude there is no great advantage in using a multi-linear source.

As discussed in the previous paragraph, the great number of element required by 3D simulations is not sustainable by current computers unless we accept a very long time of analysis. For this reason the planar representation of the problem was adopted as solution to minimize the number of elements. With a 2D model it was possible to simulate both a linear and a multi-linear source, developing two sets of analyses to consider the NDT system passing by the defected section. Comparing the standard displacement (obtained in the model without defect) and each displacement obtained at a certain distance from the defected section, it was seen that consistent differences are present when the inspection is simulated in the range between ±1.0 mm. In this interval the difference pattern assumes an amplitude of some nanometres.

However, even if the planar representation of the problem was useful to develop our feasibility study in a reasonable time, surely we don’t forget that it adds some approximations due to the plane strain model that can’t give any information about the wave reflections on the lateral faces of the deposit; before was stated that it seems not to be very relevant for ours goals, but it has been recognized of indisputable importance to conclude this report with a three-dimensional proof proving the feasibility of the laser ultrasound NDT system. Having determined, by the 2D models, that there is no great advantage in using a multi-linear source, we decided to simulate the single-line source generation. Two three-dimensional simulations were developed: the first to obtain the displacement pattern in the model without any defect, the second to collect the displacement when the hole with diameter of 50 μm is inserted along the model protrusion. All the material proprieties, boundary and load conditions, were set at the same values used for the 2D studies. The 3D numerical models were built considering the conceptual scheme of Figure. The generating laser beam hits the deposit at a certain distance from the LMD nozzle and the receiver registers displacements at a distance of d=1mm from the thermoelastic ultrasonic source. Taking into account the same reference system previously introduced, the second simulation has been developed to reproduce the inspection procedure at x=0mm, when the midpoint between the ultrasound source and the receiver is exactly aligned with the defected section (containing the hole axis).

Figure 55 shows a picture of the meshed model without defect and of the model with the hole. In both models, a portion of 2 mm along the protrusion and the below material layer, interested by ultrasound field, were meshed by means of small tetrahedral elements with sides 8 μm long. The plots reported in Figure 56 show the displacement collected at the receiving point in the standard model and its frequency spectrum.

Finally, Figure 57 compares the two displacement patterns (collected from the standard and from the defected model) giving a clear evidence of the success of the three-dimensional feasibility study. The difference pattern assumes a peak to peak amplitude of some nanometres, being its maximum value greater than 10µm.

The preliminary feasibility studies for the laser ultrasound method were successfully carried out. Numerical analysis has been used to establish system specifications for the generating and the receiving laser equipment. Simulations have been performed on the INCONEL alloy 600 models, produced by TWI. The required laser parameters were predicted (e.g. energy, beam shape). Several possible configurations for the setup of laser source and receiver have been considered and a single-sided probe arrangement has been selected. This configuration seems easy to be implemented and ideal for detecting diffraction waves. Simulations anticipated that flaws with size higher than 50 μm could be detected.

Having carried out the theoretical study it was possible to move on to a practical feasibility study and Task 3.2 the construction of the prototype. The parts needed were acquired, these were assembled on an optical bench in the laboratory as shown in Figure 58, a PC and an acquisition board were also required. The transmitting laser required a series of adjustable optical components to ensure that the beam hit the target close to the beam of the receiving laser of the interferometer.

The pulsed laser was a High power Q-switched Nd:YAG with a Wavelength 1064 nm and Max energy per pulse 300 - 400 mJ. The repetition rate at least 10 Hz and pulse duration 4-6 nsec. The Pulse-to-pulse stability was specified as ± 2 %, the output beam diameter < 8 mm and the beam divergence < 0.5 mrad.

The area of the incident laser beam on the sample surface affects both the power density (and hence the threshold for ablation) and the dimension of the ultrasonic source. The laser output diameter is first reduced to a smaller diameter and then focussed to a line source with dimensions 0.3 mm x 0.1 mm.

The beam attenuator, at the laser output, is used in the process of positioning and focusing the laser beam.

The laser receiver is intended for the optical detection of ultrasound at the surface of the sample. The detectors produce a time-varying analog voltage that is proportional to the instantaneous surface displacement. The laser receiver, purchased for the prototype, has a number of features that make it particularly effective in the detection of laser-generated ultrasound. The features are high sensitivity & low laser power, inspection of rough/porous/rusted or mirror-like surface, robust, stable & compact, user-friendly. There is no signal fading, although a 30% variation of the signal amplitude is typical, because of the random quadrature demodulation.

An infrared filter is placed after the fiber head to cut-off the infrared wavelengths while passing visible light. The laser receiver, shown in Figure 59, has a wavelength of 532 nm an output power at the laser output of 1000 mW with 880mW at the 10m fibre output. The optical head output with f=100 mm, Ø= 1 inch should be 700 mW. The detector bandwidth was 1 MHz to 65 MHz and the sensitivity noise equivalent surface displacement on aluminium target, rough surface< 9 ∙ 10-6 nm/Hz1/2 at the calibration frequency of 1 MHz.

The system was assembled and tested and operated successfully. Figure 60 shows an example of the output on a reference sample (a range of reference samples was tested and the results are shown in WP7.)

The prototype for mounting on the robot was designed considering two constraints: - there are currently no high power pulsed lasers with fibre on the market and the ultrasonic system must be adapted to the design of the attachment plate to be connected to the plate on the robot (see WP6). The final construction is shown in Figure 61.

The procedure for setting up and operating the system carried out in WP3 Task 3 was developed and is given in detail in the Deliverable Report. It aims to position the lasers as shown in Figure 62.

Development of Eddy Current Prototype

WP4 was the eddy current work and Task 1 of this did an initial feasibility with different coil types and their adaptor circuits. It was expected that it would be particularly difficult to detect flaws in the first layer above the substrate, as the probe cannot be deployed at the side of the deposit (whereas this is a possibility with subsequent layers). Therefore, the first major challenge that arises from this specific configuration. is the size of the eddy current probe to be used.

A typical size for the smallest eddy current probe is a ferrite core of 1mm diameter and an overall probe size of around 2.5mm. The relative sizes are shown for the reference block in Figure 63. The sensitivity of such a size is bound to be low, as only a small portion of the magnetic field intersects with the defect. Therefore it is a requirement to reduce the size of the probe below that which is available commercially, whilst also penetrating the thickness of the layer.

The first part of this task was to investigate different probe designs and the use of alternative geant magnetorestrictive (GMR) sensors. At the time sourcing the latter in a suitable small package (they usually are made in dual-in-line type integrated circuit packages) and the fact that a coil would be needed anyway to generate a magnetic field ruled out this type of sensor. A cross wound coil was found to be insensitive, so attempts were made to make very small coils. In a first instance, a small coil with the following characteristics 20 turns, 5mm length and 0.5mm diameter was produced in the laboratory as such small coils cannot be found available on the market. The response of such a small coil size was investigated and the aim was to compare this response with the results obtained through the modelling experiments.

The coils were wound on a very thin nylon former (see Figure 64a). This was a very fragile design so it was then inserted and electrically connected by means of a former (see Figure 64b). In order to increase the inductance and focus the magnetic field plastic tape coated with iron oxide particles was suggested to be wrapped around the core and/or the coil. Another method chosen was the use of a tapered ferrite core, with the windings continued as far as possible down the taper (Figure 64(c)).

In the event a set of probes with a very small winding and a very small core was designed and made commercially and was used for the further developments. Comparative tests such as those shown in Figure 65 were carried out and indicated the best probe and frequency to use.

The results from these are given in graphs such as that in Figure 66, which gives an idea of the flaw size and depth below the surface that can be detected, and the size of the indication (colour scale). It can be seen that although small flaws can be detected, the amplitude decreases significantly for these whatever the depth.

A theoretical investigation of the capabilities of small absolute probes was conducted, in particular the relationship between depth of penetration achievable with probe size and frequency was studied.

Using slots in a plate as shown in Table 8 (theoretical) raster scan is performed with the probe located on the top surface of the sheet and then the signal response is simulated. C-scan images of the ECT simulated signal response and impedance plane diagrams can be obtained and analysed. A comparison can be drawn between the responses obtained for three sizes of flaws and for each scenario.

Figure 67 below show the modelling results (C-scans and impedance plane diagrams) for the defect interaction simulation realised at 2MHz for the three different coil sizes investigated.

For cases 1 and 2 detectability of the three flaws is very difficult, indeed the scale on the impedance plane is in the order of 10.10-6 which is very low and suggests that the strength from the response of these flaws will not be high enough and the eddy current system involving small coils such as in cases 1 and 2 will have to be optimised to increase both depth of penetration and electric field amplitude in the Inconel sheet.

Note that the same modelling experiments were carried out at 500kHz and 1MHz but no significant improvements were observed in the strength of the signal.

An example of the results is shown in Figure 68, and further results were published [5].

Task 3.2 was the construction of a prototype. The EtherNDE Veritor instrument was chosen as the basis for the system. This is because it was very small and was controlled by a laptop, thereby enabling the automated data acquisition required when the instrument would be remotely driven on the robot. This was achieved by two interface boxes (Figure 69), one supplied by EtherNDE and one built in house using an Arduino microcontroller. The functions of each of these are described in WP6.

Development of Laser Thermography Prototype

Initial feasibility of the laser thermography system in WP5 was carried out both theoretically and experimentally. There are a number of mechanisms by which light interacts with the various types of materials. Generally, the penetration of light into an absorbing material is governed by the Beer-Lambert Law (29). According to the equation, the intensity of light travelling through the material over a distance x, diminishes logarithmically.

Equation 30 defines the relationship between the absorption coefficient and the refractive index of the material for the wavelength of interest. In metals, the imaginary part of the material’s complex refractive index k or otherwise referred to as the extinction coefficient, is related to the electron cloud density of the metal lattice as it interacts with the electric field of a propagating plane electromagnetic wave It is via this interaction that heating occurs on the material. Moreover, the penetration of light in a metal is much less than most other materials for this reason.

The heat generated in the material via this light to matter interaction then propagates into the material further than the volume that has been subjected to radiation heating. The heat dissipation pattern in high thermal conductivity materials is hemispherical around the irradiated area (Figure 70). This model is assuming that the material is homogeneous and continuous throughout its volume. When a flaw is present this is not the case and a different heating pattern is produced. In this case finite element modelling is needed.

Transient heat movements from pulsed thermography can be predicted within a homogeneous medium by applying a specific solution of the Fourier Heat Equation for a linear condition where the heat content at a specific point in time is calculated. Predictive modelling was performed to identify automated techniques to detect and to quantify the size of defects based upon findings that defected regions such as cracks and porosity have lower thermal conductivity and affect the flow of heat throughout the inspection specimen. Based upon a fine scale LPD track, heat flow was modelled for a laser pulse across a 300µm LPD track’s deposition profile. A theoretical 30µm porosity defect (95% density) was introduced from 70-80% along the length (60-90µm) of the profile.

Figure 71 (a) shows the heat profile 5ms after the laser pulse with the defect located on the left hand side. A symmetrical distribution of heat throughout non-defected homogeneous components (the blue line) is contrasted with the profile of the line profile where the defect has been introduced (the red line).

The model predicted that the defect would be detectable as heat flow reductions of over 20% were predicted. Furthermore it should be possible to identify where the defect is located and it may be possible to calibrate the size of the defect. Figure 71 (b) shows the heat content for both time and position along the line profile. The asymmetry is more clearly shown in Figure 71 (c), (d) and (e) that present heat content over the line profile, heat change over time and heat change over location.

However, while a detailed inspection of Figure 71 allows the identification of the defect, the influence of the defect is not easily visible in graphs of the full line profile and a more effective way of representing the inhomogeneity was desirable. An asymmetrical heat transfer model was therefore used. In this case, inhomogeneities within an LPD track have been characterized by a region of lower density. These volumetric discontinuities act as inhibitors of the heat movement; in effect as a partial boundary condition and causing local fluctuations in the heat content to introduce an asymmetry. While this asymmetry is detectable in the full line profile model it becomes obvious and easily quantifiable when extracted by subtraction of the defected region from other sound areas. The subtraction of the defected area against a sound region offers a large degree of flexibility and as LPD tracks are added linearly, simple algorithms such as image subtraction along the deposition path (subtraction of one side of an image from the other) effectively identifies the defected region and can be used to quantify the defect severity which can then be compared to acceptance criteria.

The temperature difference for these heat content profiles (subtracting non-defective from defective), plotted against time, temperature and position in Figure 72 shows an initial negative temperature differential followed shortly (< 10ms) by a positive temperature difference due to the heat retained for longer by the defected side of the sample.

This is more clearly seen by aggregating the temperature difference (Figure 73 (a)) for all positions along the sample where a negative temperature difference is seen in the first 5 – 7 ms followed by the net positive period until both sides have returned to the environmental temperature.

This profile and the plot of temperature over time along the locations of the LPD profile (Figure 73 (b)) are able to accurately locate the defect clearly within the 60-90µm position. However the most obvious representation is the plot of the range of temperature differences along the profile (Figure 73 (c)). This plot also offers the most accurate approach the project has identified to calibrate defect parameters to quantify the defect size.

The greatest temperature ranges (from negative to positive) are again identified at the 60-90µm position.

The extensive computational requirements for a heat transfer model that identified and quantified defects in an unbounded geometry would require comparing all radiations in three dimensions from the heat pulse impact location. This would be unnecessary in the case of thermographic examination of an LPD track as only movement of heat across the surface can be observed. Furthermore, if multiple inspection points are chosen, spaced to detect the threshold acceptable defect parameter, lateral defects that are not in the direct deposition path would also be detected.

The modelling identified an approach, termed asymmetry analysis, that suggested comparing the line profiles of the heat flow along the deposition track would be able to precisely locate a defect and allow the calibration of temperature difference ranges for a given inspection set-up to quantify defect sizes. Consequently the most computationally effective detection algorithm is the analysis of two directions along the deposition path. Analogous techniques such as subtracting opposite sides of inspection points for the entire image along the deposition track extend the detection technique to a quantification technique.

To put these theories into practice requires a suitable pulsed laser and a thermal camera. Initial experiments were carried out with the arrangement shown in Figure 74. Initial tests (Figure 75) showed that it was possible to visualize the laser heating of the surface with suitable settings.

The laser shown in Figure74 was not portable and was being used at a fraction of its power, therefore a 20W laser was substituted for this, and a series of experiments carried out on the reference samples. The arrangement with this laser is shown in Figure 76. In order to study this method image analysis is required. This was initially carried out using the Thermagram software. An example of the results is shown in Figure 77. The equipment was further improved by adding a germanium lens to the camera, to reduce the field of view and thereby focus on the incident point of the laser.

Following the modeling, analysis software was developed to automate the extraction of the region of interest (ROI), define the sequence of the LPD track that is to be compared, generate the inspection locations for a controlling robot arm to move to and to analyze the condition of the deposition and compare it to acceptance criteria.

The heat transfer model for non-defected LPD beads displayed a symmetrical heat distribution. As well as being identified by asymmetries in the heat content along the line profile, defects were also visible by differences in the heat gradient in the line profile (the spatial or positional derivative) and differences of the heat content between successive frames (the time derivative)

Initially software was developed to analyze the successive greyscale images of the cooling phase for the transient thermal event following the heat imparted by the laser impulse (Figure 78 (a)). Experiments showed that there would be four critical frames where the heat dissipation throughout the LPD track would be most visible. Extensible software was developed that presents these four frames as well as graphs of the raw line profiles of each frame and, alongside this, the asymmetry analysis.

The first process used was spatial derivative analysis is presented in the same user screen (Figure 78 (b)) showing the line profile of the gradient of the heat change for each location in the line profile and, alongside this, the asymmetry analysis of the gradient of the heat content by subtracting the two sides of the spatial derivative line profile from the centre point.

The time derivative analysis was again presented in the same user screen (Figure 78 (c)) showing the line profile of the heat change from frame to frame for each location in the line profile and, alongside this, the asymmetry analysis of the difference of the heat content by subtracting the two sides of the time derivative line profile from the centre point.

The automated software using the theory described above was however limited because it did not turn out to be possible to synchronise the data acquisition with the laser pulse.

System Integration

System Integration was carried out in WP6. The three systems were integrated together and with the deposition robot, mechanically, electrically (control signals and data acquisition), and with software control and data acquisition.

The basic mechanical requirement is shown in Figure 79. It was decided that the method of operation should be that a deposition layer would be followed by an inspection. This means that (at least for repeated layers of any geometry) that the path of the inspection could follow the path of the deposition by means of a simple translation. This means that the mechanical attachment could be achieved by simple adaptor plates. Two types of adaptor plates were planned, one which connects to the faceplate of the robot (the robot attachment, Figure 80), and the others (the NDT plates (Figure 81)) on which the equipment is mounted. This concept enables each equipment to be set up on a plate which is simply bolted to the robot plate without dismantling any of the deposition laser.

Figures 82-84 show the actual construction of the NDT plates with the equipment mounted. Figure 85 shows one of the plates mounted on the deposition robot.

In order to provide the electrical integration of the prototypes to the robot movement an electronic interface adaptor was designed. This is slightly different for each NDT system.

For Figures 86-88, the pink box represents the laser cell. Therefore all manual control must be outside this box during operation.

For the laser UT system (Figure 86) the system starts up and acquires data basically on the start and stop of the robot movement.

Figure 87 shows the eddy current system. The instrument is required to be operated basically Initially, the robot moves to position above sample with settings for material identification, the equipment is nulled, the robot moves to sample, data is recorded and new settings imported for flaw detection. The null is reset, and the robot scans sample during which the data is recorded and then exported to the main PC. This was built with two boxes as the timing of certain control actions for the instrument is related to the robot position just before engagement and after.

The laser thermography system (Figure 88) is also related only to the start and stop of the robot, but the hardware box also supplies a trigger pulse for the camera frame acquisition.

Figure 89 gives more detail of the internal structure of the hardware box. The heart of this is an Arduino unit that produces or responds to signals in a USB cable and communicates with the KUKA (under control of the KUKA software). Trigger input and output pulses are also detected and produced under PC control. A step up/step down voltage correction to enable the 24V KUKA to communicate with the 5V logic system is also included.

The software integration was slightly limited, as the KUKA robot software does not easily integrate with other software. It was therefore decided that the KUKA PC would continue to control the robot movement while outputting position indications and start and stop acquisition to the NDT PC. The laser transmitter of the laser thermography system was also kept as a separate unit as it was on loan to the project.

In order to check the successful integration a set up and calibration checks were recorded in Task 6.2. Prior to starting inspection the deposition path and the required translation co-ordinates of each of the inspection systems must be determined. This is done by using a camera on the deposition laser to record start and stop positions, and then manually setting up each inspection system over the start and stop positions and recording the differences.

The integrated software can then be implemented. The graphical user interface (GUI) is designed for automatic inspection once connected with all hardware (Laser Thermography, Laser Ultrasonic, Eddy Current, Robot and Digital IO Box). It communicates with hardware and automatically calls the three NDT techniques.

The INTRAPID graphical user interface is show below in Figure 90. The left hand side is a tree view of the functions. The right hand side is a tab control which shows currently opened tab pages. This software was designed to work with external hardware, i.e. Thermography Camera, Laser Ultrasonic and Eddy Current. Once it starts up, the system automatically detects and connects to the hardware (Figure 91). It will then load the Inspection Geometry, and Integrated tabpage, as shown in Figures 92-93. To import the component geometry, it opens a tabpage called “LPD Profile”, if no such tabpage exists, it will create one. The system currently only supports “Import Predefined Profile” which the profile is defined in an XML file.

The hardware settings are managed individually in the three NDT techniques. The software initially checks communication to the interface box and then starts the inspection process by selecting which of the three systems is to be used.

The system is then configured to adjust the inspection system settings (if such setting is available from the Software Development Kit (SDK)). This operation defines the NDT inspection system. Currently the system supports Laser Thermography Techniques (e.g Thermoteknix Miricle 307kx as shown in Figure 94), Laser Ultrasonic (UNIPA System), and Eddy Current (TWI/EtherNDE System), as the system has already integrated these device drivers and data acquisition APIs. The system can easily integrate another device once the SDK for the device is available.

An example of the output of the software is given in Figure 95. This shows the eddy current recorded data from a scan including a set threshold for flaw indication or acceptance criteria (the red box). The system automatically records and notes all the data points outside the box, while accepting those inside.

Calibration of each system is carried out with a known reference sample for example Reference Sample Type 2 No EDM 6.1 was compared between the laboratory and site trials for the laser ultrasonics (in this case the site transmitting laser was different). The result can be seen in Figure 96. For the eddy current and laser thermography the hardware is identical, so the test is only that the controlling software works and the scan is the same.

Performance Data

WP7 Task 1 was to establish the performance of each system. Normally, a probability of detection curve (a graph of detection probability against flaw size) is used to describe inspection performance. In INTRAPID, this is slightly complicated by the fact that two parameters need to be evaluated in this case, namely the size of the flaw and its depth below the surface. The two are needed together in this case because their dimensions are similar (for normal POD studies the flaw is usually much smaller than the object under test). In addition, parameters such as surface condition and flaw morphology have a significant effect and have only been partially tested. Finally it is very difficult to actually find the size of the flaw when the flaw is small. Sectioning is somewhat risky with such small flaws even when their position is thought to be known, because if there is an error in the flaw position then the section could easily miss it or partially cut it and give a wrong size. X-Ray and dye penetrant methods were used to analyse some of the deposition samples.

The use of the reference samples to establish a suitable procedure and settings has been described earlier. Because of the large number of reference samples, and the fact that good data was obtained from these (and the flaw size and position could be accurately known) the results are also used in this analysis. However the data from the deposition samples has been used wherever possible.

The degree to which the individual methods were tested also depends on the original success with the reference samples. The eddy current and laser ultrasonics methods were used on most samples whereas the laser thermography was not tested on the Inconel samples, as the use of robotic deployment was limited due to time constraints near the end of the project.

The reference and deposition samples are described earlier in this report (see for example Figures 1-40 and associated tables and text. A selection of these was machined down (to improve X-Ray resolution) to 5mm thickness and then radiographs were produced. Dye penetrant was also used to establish whether the flaws were surface breaking. The samples had the surface removed to different degrees to see the effect of the surface on the inspection method.

For laser ultrasonics the results for the Inconel samples and reference samples are summarised as a detection graph in Figure 97. In general the filled data points were detections (hits) and the line or cross data points were misses. As the mathematical function of the relationship between POD, flaw size and flaw depth below the surface is unknown, the line drawn through the points was inserted by eye and is intended to give an approximate idea of the capability of the system. It has been drawn such that there are no missed flaws to the right and below, and therefore represents at least 90% POD.

The laser ultrasonic method used in INTRAPID seems to be capable of detecting flaws in the Inconel samples down to around 0.1mm diameter near the surface, and larger deeper flaws up to around 0.6mm diameter and 0.7mm deep (Figure 97). It should be noted that these limitations were only those for this piece of equipment and test samples.

For the copper alloy deposited on aluminium, Figure 98 shows a comparison of the results obtained on the machined sample and the dye penetrant results. Apparent surface breaking flaws were detected, and these appear to correspond quite well with a dye penetrant test (Figure 98). There is a limitation in the presence of a large acoustic signal that appears in the Figures above the indications of interest. This is an acoustic signal and requires careful adjustment of the distances between equipment and focal positions to avoid it interfering with the required indications.

A test on the as-deposited surface condition was carried out and the results can be seen by comparing Figure 99. It appears to be possible to detect some surface breaking flaws even from an as-deposited surface although it cannot be expected to be as sensitive and there may be a greater restriction on sub-surface flaws that weren’t available for test here.

For eddy current testing of the Inconel samples it was noted that the substrate and the deposit were very different in their eddy current response (suggesting that each had a different conductivity, possibly due to the effect of different heating cycles). The difference can be seen in Figure 100. Note that the gains used here are less than those used for the scanning (the software system automatically adjusts this for the different tasks). This will have an effect on inspection of deposition layers close to the substrate, and also will have a strong effect if the probe is displaced sideways from the deposited layer or if this varies in width. The effect may produce false calls if pronounced.

The detection results are summarised in Figures 101(a) and 101(b) for 100KHz (Figure 101(b) is simply Figure 101(a) with the horizontal scale changed, Figures 102(a) and (b) similarly at 500KHz, Figure 103 at 2MHz and Figure 104 at 6MHz. The performance capability of the eddy current system depends on frequency and probe used. The lower frequencies tend to detect more at greater depth but the higher frequencies are more sensitive at the surface. The capability falls off as the depth increases because of depth of penetration effects discussed earlier (see Figures 67 and 68). This is broadly shown in Figures 101-104 although the 500KHz result is slightly better than might be expected.

Table 10 shows the results for diameter and depth of the deposition samples (measured by radiography) compared with the eddy current amplitude and phase. The phase is given by Cartesian quadrant only (TR = top right etc). It can be seen that for the deposited samples in the flaw indication varied in both phase and amplitude. These results are not always consistent with the radiography results, for example one might expect the 100KHz indication to be less than the 500KHz indication for a surface breaking flaw, and this is true for the one dye penetrant indication (3E2) which in addition has a different characteristic pattern, but is also true for the another flaw (5E1) which is a small deep flaw by radiography. The phase indication is also inconsistent and although it might be hoped that the phase would give a consistent relationship with flaw depth this also does not occur. It is a possibility that the eddy currents are detecting the small difference in the heating of the powder as the flaw is formed, rather than the flaw void itself (or a mixture of the two).

The results used here were on a machined surface, and the effect of surface can be seen in samples 3E6 and 5E6 which weren’t machined. In these cases the indications from the surface are typically a similar size to those from the flaw.

The result of the inspection of the copper alloy deposit was similar to that of the laser ultrasonics, with most of the surface breaking cracks consistent with the dye penetrant results (Figure 105). As for the Inconel samples (and to a greater extent) the substrate and deposit materials have different electrical conductivity, so it is important that the scan remains on the deposit.

The result in Figure 106 shows that the eddy current method appears to be capable of distinguishing the different deposition qualities (see Figures 29 and 37). It is still unclear however whether the method is truly detecting individual flaws or whether the material differences are also being indicated.

For Laser Thermography, the results for the Inconel reference samples are summarised in Figure 107. The line drawn on this graphs was added by eye as an estimation of the limits of operation for the technique. These results are on the reference samples and show a minimum detection size of around 0.2mm and a possibility of examination of depths up to 0.9mm with larger flaws. This is similar pattern to the eddy current performance at 500KHz. However it should be noted that the laser thermography method is subject to two variables (pulse power, pulse duration), and it may be possible to extend the range by increasing these, or the resolution (in both temperature and spatially) of the thermal camera. Unfortunately it was not possible to carry out the range of extensive tests with the thermography system on the Inconel deposition samples.

The results for the inspection of the copper alloy deposited on aluminium are shown in Figure 108. This shows that there is some consistency with the dye penetrant results but this is not quite so clear as the other methods. The “detection” at present, being dependent on the slope of the cooling curve, is much harder to observe than other methods.

The main conclusions that can be drawn from these tests are:

1 The performance of the three NDT methods has been broadly established for the example of an Inconel material and a copper alloy deposited on aluminium.
2 The laser ultrasonic system is the most sensitive to surface flaws, and is capable of detection down to 0.1mm in depth.
3 The eddy current methods are slightly less sensitive, around 0.2mm at the surface, but appear to be capable of detection to a greater depth. However the indication of depth (by phase) is not clear cut at this stage. There is a possibility that part of the indication is due to material change.
4 The eddy current method seems to have an indication of the characteristics of the copper alloy deposited on aluminium, but this may be a combination of material change and flaw detection.
5 The laser thermography performance seems to be able to detect down to around 0.2mm on the surface and to a depth of 0.8mm on the Inconel samples.
6 The laser thermography seems to be able to detect the surface flaws on the copper alloy deposited on aluminium, however this was dependent on the cooling curve from the laser spot, which was also affected by the surface condition.
7 The surface condition of the samples is important and to achieve high levels of sensitivity the surface needs to be better than the as deposited surface (although some inspection is possible with the latter). Further work is needed on this aspect of the testing.

REFERENCES:

1) C. B. Scruby, L.E. Drain, Laser Ultrasonics, Techniques and Applications (Adam Hilger, London, 1990).
2) www.specialmetals.com
3) Shant Kenderian, B. Boro Djordjevic, "Sensitivity of Point- and Line –Source Laser-Generated Acoustic Wave to Surface Flaws”, IEEE transactions on ultrasonics, ferroelectics, and frequency control, vol. 50, pp. 1057-1064 (2003).
4) Claudio Cosenza, Shant Kenderian, B. Boro Djordjevic, Member, IEEE, Robert E. Green, Jr., and Antonino Pasta, “Generation of Narrowband Antisymmetric Lamb Waves Using a Formed Laser Source in the Ablative Regime”, IEEE transactions on ultrasonics, ferroelectrics, and frequency control, vol. 54, pp. 147-156 (2007).
5) S Majidnia and J Rudlin “Depth of Penetration Effects in Eddy Current Testing” BINDT Conference Proceedings Daventry 2012.

Potential Impact:

It was estimated at the outset of INTRAPID that the existing casting and forging markets were now worth around €230 billion per annum globally, growing steadily at 10-15% per annum. The manufacture of parts by Laser Metal Deposition represents a highly innovative technology alternative to more specialised applications of metal casting and forging processing, serving applications in key European sectors, including the production of specialised/critical components for automotive and aerospace engine parts as well as medical implants. The additive manufacturing market has continued to grow during the course of INTRAPID, to the extent that “3D manufacturing” as it is sometimes now called, is often referred to as something akin to computer technology in the 1980’s, and an expectation that devices will be used even in a domestic environment, at least for low melting point products. However for the industrial applications methods of metal deposition will be more significant. For the automotive and the aerospace sector, LMD has found special interest since it brings technological advances in the manufacture of highly specialised and complex shapes which provides a host of benefits in designing and making automotive and aircraft engines.

In 2000 alone, almost 30% of the machines (approx 5,500 with an approximate cost of € 45K) producing parts through this technology were installed by the automotive industry. The technology has grown into this market in the last 10 years to a number of installed machines surpassing 20 000 (approximate cost of €300K per machine) and in 2008 the global market was estimated to be worth €2 billion [ ]. The automotive industry is a vital economic sector in which Europe has a large tradition and experience; therefore the consortium SMEs look forward to increasing their global market share by providing the means of improving and ensuring the quality of the engine parts produced by LMD technologies through the use of the INTRAPID system.

As Europe is introducing an innovative process that brings important benefits for the sector, it is expected that LMD technologies will have a penetration in the global casting production of automotive parts industry whose market is estimated to be about €164 billion by 2010. Importantly, Europe is the largest market accounting for 39.1% of the global value [ ]). The current situation shows Asian and Latin American markets as those experiencing a larger growth in comparison to Europe and their combined market is estimated to be approximately 4% of the global market for automotive parts production[ ]. Hence the consortium SMEs will also seek to expand the technology into the Asian and Latin American emerging markets where there is a high potential of market adoption.

Furthermore, the INTRAPID project team have been keen to engage with the aerospace industry to demonstrate the significant efficiencies players in this sector could achieve through the adoption of the INTRAPID technology. In the aerospace sector alone, there will be 24,300 new aircraft built in the next 20 years – these programmes will use 20 million tonnes of billet, yet only 2 million tonnes will actually make it on to the aircraft after machining. The market potential is clear to see, and in countries such as the UK it is estimated that potential aerospace component work could soon be worth more than €5 billion.

Wasteful manufacturing is one of the biggest challenges facing the entire manufacturing sector, but is particularly impacting aerospace. It is no longer sustainable in the aero-engine business as the cost of key materials such as titanium and nickel alloys is escalating, fuelled by the seemingly inexhaustible requirements of the world's strongly developing Asian economies. There are also tough ecological initiatives being introduced across the world aimed at reducing the impact of manufacturing on the environment, particularly in terms of energy consumption, use of resources and waste pollution. There are even tougher targets that have to be met for engine efficiencies and gas emissions such as stated by ACARE (Advisory Council for Aeronautics Research in Europe) where, for example, exhaust carbon dioxide is to be reduced to 50% by 2020.

The proportion of originally purchased material that is finally built into an aero-engine (i.e. the buy-to-fly ratio) remains as low as 10% which means that 90% is reduced to relatively low-value scrap. Thus, considerable excess time and money is used to make the raw material and even more is used to form then machine the final product shape. Although some cost can be off-set by selling purposely identified and segregated waste to be recycled, and somewhat less of the cost can be offset by selling lower grade mixed scrap to the steel industry, there is still a significant amount of fine material engrained in consumable machining waste or in the form of oxide scale that ends up in a landfill site. A crude estimate of the quantity these types of waste across the industry is 30%, 50% and 20% respectively that return less than 10% of the original purchase price in scrap value, but this is probably counteracted by the overall handling costs and disposing of the unrecoverable waste.

The obvious response to all these issues is to convert to an additive manufacturing process where fly-to-buy ratios can be around 85-90% together with the potential for superior design and application developments. LMD technology could be the solution as it has proven to produce important benefits for this industry such as almost complete elimination of material waste during manufacturing (approximately 90% on weight savings are obtained by avoiding machining metal casting parts) and easy machining of less dense materials (e.g. aluminium or titanium vs. steel). This brings important benefits such as highly improved part performance with a minimum of defects as well as process energy savings that lead to lower CO2 emissions (120 trillion Btu annual savings have been calculated for manufacturers adopting this technology, which translates to 2.3 million tons of CO2 savings [ ]). Furthermore the technology offers significant time production savings by up to 80% in comparison to the time consuming process of casting and forging [, ]. These benefits could lead to potentially vast costs savings in both labour and lead time.

A strong common driver throughout the aerospace sector is weight reduction with the ultimate goal of trying to improve fuel economy for each aircraft. This requirement is driving the uptake of lightweight materials such as titanium alloys and lightweight designs. Titanium and high-performance alloys are constantly in great demand as OEMs strive to reduce weight in new aircraft, such as the Boeing Dreamliner 787. Content of titanium in the Dreamliner is 15 percent, by weight. But the high cost of mill products and machining are major obstacles in providing cost-effective planes.

Many parts are machined from solid mill products, and many others are made from closed-die forgings or castings. The buy-to-fly ratio is as high as 15 or 20 for many flying components due to complex geometries, adding a lot of cost to the component. LMD processes provide an opportunity to produce light-weight titanium components with a buy-to-fly ratio close to 1. In addition, LMD overcomes the problems of long lead times needed for die design and fabrication.

In the case of titanium, using LMD reduces the extraction requirement for Rutile by 25 times against the same component machined conventionally. Titanium extraction is highly energy intensive; CO2 emissions for the production of 1 kg of titanium in fossil fuel dominant regions is 9 tonnes. So, if that 1 kg of Titanium becomes 1 kg of component, it costs only 9 tonnes. If we compare this with conventional manufacturing, you would need 25 kg of titanium for 1 kg of component, which would cost about 250 tonnes of carbon. The significance of these numbers becomes important when considering that, on average, an aircraft will burn about 0.03kg (.06lbs) of fuel for each kilogram (2.2lbs) carried on board per hour and given that the total commercial fleet flies about 57 million hours per year, cutting one kilogram (2.2lbs) per flight can save roughly 1,700 tons of fuel and 5,400 tons of CO2 per year.

Research has shown that reducing the weight of an aircraft by 100kg using additive manufacturing structures would save approximately 4.5 million litres of fuel over the lifetime of the aircraft, equating to one million tonnes fewer CO2 emissions. For an airline operating 30 long-haul aircraft this would mean saving €2.9m per year (source: Exeter University). Independent studies conducted by Airbus indicate that a fully ALM enabled aircraft will be 60% cheaper to make, and also 30% lighter, which again is saving energy in terms of what is needed to fuel it and get it off the ground.

Environmental impacts

There are five fundamental environmental drivers to the INTRAPID system which will impact on the automotive and aerospace sectors:

1. Reducing energy and costs during the manufacturing chain - The laser additive repair of seal segments for a high-pressure turbine aero engine is one example of an area where the environmental impact of new manufacture can be reduced. The traditional manufacturing route for such a single crystal component is to EDM the lattice that contains the abradable material into the solid base; it is this abradable lining that wears during service as the turbine blade seal fins rub into it. A conventional technique such as brazing a new lattice onto the seal segment is not feasible because the lattice walls are required to be less than 0.3mm thick (and, with 34 segments in each turbine, very costly). But by fusing a suitable powder using direct laser deposition to build up the lattice, the cost of repair can be brought down to half the OEM cost.
2. Reducing material wastage – LMD is much more efficient in terms of material wastage, which will significantly reduce the amount of scrap produced during manufacture. Aerospace machining currently generates significant volumes of scrap which then requires costly and energy inefficient reprocessing.
3. Reducing transportation, logistics and packaging - LMD has the potential to reduce many stages of the traditional supply chain. Because highly complex geometries can be constructed via this method, many parts can be consolidated, reducing both piece part count and assembly.
4. Realising optimised products - Most products are not optimised, as they are ‘designed-for-manufacture’ rather than ‘manufactured-for-design’. Because of the constraints of traditional manufacturing processes, many design objectives are curtailed. Furthermore, with LMD it is suggested that truly ‘topologically optimised’ designs could be realised, which could increase the functionality of the product, reducing wasted energy, fuel or natural resources in operation.
5. Life cycle carbon footprint - One of the most important manufacturing considerations today is life-cycle engineering, where the designer must consider both the implications of the part through its service life and the end of life disposal of the part. For both aerospace and automotive components this means that we must now consider the life-cycle implication of a component in terms of its carbon footprint and the long-term impact of the part on the environment. With LMD, by optimising the design of a product and manufacturing it with absolute minimum material, it is possible to significantly lower the weight of parts.

Due to the benefits highlighted above, and to increase the penetration opportunity of this technology, the INTRAPID system has been created to inspect the very complex shapes produced by this technology, which cannot currently be inspected with the ordinary radiography technologies used at the moment. Therefore the inspection machines currently installed and the fast growth in the acceptance of this technology represents the market opportunity in which the INTRAPID system will find penetration.

The benefits of the use of INTRAPID systems in additive manufacture at the European and International level are

• To gain competiveness in the global automotive and aerospace manufacturing parts market by improving parts quality and performance, enabling production of highly complex, specialised/critical components with known high quality levels.
• To reduce approximately 5% global material waste by a penetration of the LMD technology in the metal casting and forging industry. Hence energy savings will also be obtained. This contributes to EU policy on supporting reduction of worldwide CO2 emissions in agreement with international regulations (such as Kyoto protocol and Copenhagen strategy) to which Europe is highly committed [ ].
• To strengthen the EU’s position against manufacturers in other regions where high incidence of defects are compensated by overproduction. A penetration of at least 1% of the metal casting and forging market is expected.
• The consortium expects to improve automotive parts manufacturing reliability to above 99% (by defects and waste reduction in batch to batch production).
• The generation of approximately 300 jobs worldwide through facilitating industry growth by increased competitiveness.

The benefits to the INTRAPID consortium members include:

• To increase profit by reducing product rejections
• To gain IP from INTRAPID system development
• Increase sales of INTRAPID systems in Europe by a 4% of market penetration
• To gain a 2-3% share of the combined Asian and Latin American market by penetration of the LMD technology and sales of the INTRAPID system.

To achieve these targets, the consortium will use the INTRAPID monitoring system to carry out quality control on the automotive and aircraft engine parts production/assembly processes using LMD technology. As mentioned, it is expected that by introducing the INTRAPID system in these processes, 5% of global scrap (faulty parts) can be reduced, which will also lead to production cost savings. The overall factors that will contribute to increased performance and cost reduction are primarily 1) lighter parts, 2) material waste reduction, 3) reduction manufacturing time processes, 4) reduction in costs due to energy consumption and 5) reduction in costs due to product repair and replacement.

Implementation of this project will offer increased business opportunities for the participating SMEs and the EU automotive and aerospace parts manufacturing/assembly industry as a whole. The potential savings to both industries provide a strong incentive to use the research and SME developed NDT technology on an international basis.

Table 11 shows that the total cumulative economic benefit to the EU in the following 4 years. It has been estimated that the total investment in this project (€1.4 million) will produce significant activity immediately and facilitate direct economic benefits to the SMEs after completion of the 4th year. The INTRAPID project will generate an estimated €44m in combined cost savings and sales of the system. The return on investment ratio is estimated of 60:1, between four and ten years post project.

At the final meeting of the project it was proposed and agreed that the consortium continue to work closely together to develop a range of unique products that will be exploitable through the supply chain that includes inspection and support providers. The business that these providers generate will, in turn, provide end-users in the electronics industry with extensive cost savings opportunities outlined below.

On completion of the project, there are four main prototypes for exploitation:

1. Monitoring System INTRAPID (WP6). This will be the integrated system for automatic inspection of parts installed in LMD systems
2. Laser Ultrasonics Monitoring System (WP3)
3. Eddy Current Monitoring System (WP4)
4. Laser thermography system (WP5)

However the knowledge of the performance data (WP7) obtained will be key to the sales advice given to the customers and this was uniquely provided by INTRAPID.

In addition there will be sales from knowledge of the flaw producing process (WP2) that will be needed by users for qualification of the inspection process. The knowledge gained in production of the reference samples (WP1) will also be valuable, as similar samples will be required for most applications.

The software in each of the inspection systems may be transferable to other inspection technologies.

All the deliverables have applications in the automotive and aerospace parts and other inspection markets. The prototypes could be fully commercialised within 12 months of project completion. The complete systems will be low volume high cost items, as they will need to be tailored to each customers’ existing product or robotic system.

As a result of participation in this project, each SME expects to commercialise the system and peripheral products one year after the end of the project. To calculate profits the project targets in the first instance a penetration of 4% of the market related to LMD technologies which are serving the automotive industry, which is approximately 25-33% of the global market [iii].

An estimate of sales for the INTRAPID system, associated NDT systems and service by partner SMEs after the project completion has been carried out. The sales of systems and services by the partner SMEs can expect to be around €18M in the EU alone after project completion. A cost analysis shows that the SMEs will re-coup their initial investment after the end of year 1. It has been estimated that the total investment in this project (€1.4 million) will facilitate approximately €9M of cumulative profit and the generation of directly and indirectly approximately 300 jobs by the end of the 10th year of product commercialisation. It is expected that the market share of the participating SMEs will increase drastically as they will be able to provide an automated NDT inspection system, which will replace manual or semi-automatic inspection tasks.

The competitive system for some of the systems is likely to be computed tomography (radiographic method). This is unlikely to suitable for the most complex structures, and the cost of the equipment is in excess of €200,000. Also radiography after each layer would be impractical, as the time taken to analyse data is impossible in production lines.

It has been considered that after the 4th year in the region of 4% of the installed LMD technologies in the automotive sector will adopt the system, which will give savings of approximately €27m from materials waste reduction as this technology avoids further machining by manufacturing complex parts with almost no materials wastage. Aerospace companies have also approached the RTDs to explore the possibility of adopting the LMD technology, SMEs also expect to penetrate at least 0.5% of the aerospace market with estimated sales of the INTRAPID system reaching €20m. The project team is very conscious that aerospace companies often require a 3:1 return on its investment, meaning that for every Euro spent on a new technology or process, it must receive €3 in return to cover implementation and maintenance costs. For companies that are comparing an existing manufacturing process to additive manufacturing, it is believed that many must realize a gain of at least 30-40% when replacing the old with the new. Anything less is usually not worth the risk and hassle of replacing a proven method with one that is new and uncertain, so this is clearly a key consideration in our pricing strategy.

Furthermore, ventures into key global geographic regions will further grow the commercial opportunity of INTRAPID. Therefore sales of the INTRAPID system to the Latin American and Asian market (with a combined market value of €17m of materials sales for automotive parts [iii, ]) are expected by year 4. Considering a 1% of combined market penetration, it is estimated that INTRAPID systems could bring an extra €15m of cumulative profit.

The manufacture of critical engine parts is a very important part of the European high tech industry including aerospace and automotive; the latter has a large manufacturing tradition and economical benefits for the EU (job creation, exports, domestic market and a European brand name). The European market for auto parts is estimated at €164 billion of which cheaper manufacturers from outside the EU such as China are supplying an estimated 40%. This trend is expected to continue as greater cost pressures are being brought to bear on the automotive manufacturing industry. The INTRAPID project therefore looks forward to providing the means to ensure the quality of the LMD produced parts and increasing the acceptance and penetration of this technology in the global market of auto parts manufacturing/assemblies. This will bring many performance, costs and environmental benefits since it is well known that defects in production add significantly to the overall cost of the product. It is expected that this effort will provide to the EU a competitive edge in this market and leadership as high tech location and engine for innovation. The benefits will be perceived in the expansion of this sector to other emerging markets such as Asia. The healthy growth of the additive manufacturing industry is important to the EU due to their extensive experience in this industrial sector which supports more than 500 000 jobs and has led to the creation of new ones.

In order to achieve these sales levels it is vital that dissemination is carried out to its maximum extent. Tools such as the website, fliers, exhibition material and the video have been made available, and the completion of the project has been announced in a press release to 22 technical journals. The technical results of the project have been presented at 6 conferences with different audiences. The publication of technical papers from these conference papers will also take place, and papers from the results achieved towards the end of the project also need to be disseminated.

References

i www.madeforone.com/Articles/index.php/technology/rapid-prototyping market
ii Pheng, T.H. and Jhavar, Pragya. Autoparts Manufacturing. Business InfoBytes
iii www.thefreelibrary.com/Global+shipments+of+metal+castings+to+reach+90
iv http://www.allbusiness.com/accounting/budget/962812-1.html
v The aerospace world. Aerospace technology. Aircraft technology and engineering technology
vi Grimm T. Direct metal on the rise. Manufacturing engineering October 2004, Vol.133 no.4
vii http://offshore.windpower.org/media(929,1033)/Copenhagen_Strategy.pdf
viii www.industrialmetalcasting.com/casting-industry.html

List of Websites:

The website (www.intrapid.eu) is available and a video produced and distributed by Youtube. The partners (listed below) have agreed to continue their co-operation after the end of the project and to try to promote the inspection systems and software. For more information contact F. Baratta at Bytest.

A full list of Project Partners:-

BYTEST - Franco Baratta (Telephone: +39 (0) 119953845) / E-mail: baratta@bytest.it

TECNITEST

LPW Technology Ltd

POLKOM

TOYOTA

TWI Ltd

KCC

UNIPA