Skip to main content

DEVELOPMENT OF AN ADVANCED MEDIUM RANGE ULTRASONIC TECHNIQUE FOR MOORING CHAINS INSPECTION IN WATER

Final Report Summary - MOORINSPECT (DEVELOPMENT OF AN ADVANCED MEDIUM RANGE ULTRASONIC TECHNIQUE FOR MOORING CHAINS INSPECTION IN WATER)

Executive Summary:

Introduction
As offshore oil and gas exploration goes into ever increasing depths of water, so greater emphasis is placed on monitoring the structural integrity of the Floating, Production, Storage and Offloading vessels, which extract the product from well-heads on the sea bed. Integral to the structural integrity of these vessels is the condition of the mooring chains that anchor them to the sea bed. The mooring chains can be 2km long and because they are fixed can only be inspected in situ, except in the very rare occasions when the vessel is moved to a new well-head.
Current inspection is restricted to diver or ROV ‘swim-bys’ from below or rope access from above. This is particularly hazardous in the ‘splash zone’. MoorInspect has addressed this problem by developing a unique chain climbing robot that is able to deploy a guided wave ultrasonic non-destructive test technique through the splash zone. This technique has been used previously to inspect pipelines and for use on chains it has been made sensitive to the cracks and corrosion that can occur in the wear surfaces between the chain links.

Objective
To develop a system for inspecting mooring chains in situ using a guided ultrasonic wave technique deployed from an autonomous robot that can climb the chain from 24m below water, up through the splash zone to the chain hawse and which uses signal processing and neural network techniques to aid in the interpretation of inspection data.

Work scope
The work has been split between Plant Integrity, which developed the guided wave ultrasonic technique for chains and the marinsied transducer collar, InnotecUK, which developed the chain climbing robot and IknowHow, which developed the automatic defect detection software.
Guided wave ultrasound propagation in solid cylinders and around bends has been studied experimentally and using numerical models. The dispersion curves that are used to select wave modes and test frequencies have been derived. Problems in low signal-to-noise were encountered when changing from 110mm to 160mm chain sizes, which were only resolved at a particular ultrasound frequency-transducer ring separation combination. Although alternative electro-magnetic acoustic transducers were investigated in place of lead zirconium titanate ones, there use in rings to propagate symmetrical waves was not evaluated.
The chain climbing robot was more sophisticated than the one envisaged in the project Description of Work, in order to accommodate inspection of chains from turrets around which the vessel weather-vanes. A unique chain climbing mechanism was developed and demonstrated on a suspended chain in air. Trials in water have had to be organised for a date after the project completion.
The software for recognising signals from defects in the A-scans collected in the ultrasonic inspections incorporates a neural network

Summary
A prototype chain inspection has been developed as far as technology readiness level 6. The next step is to take it into an operational environment for which financial support is being sought from the offshore oil & gas industry.


Project Context and Objectives:
MoorInspect is a two-year collaborative project, sponsored by the European Commission’s Framework Programme (FP)7 to develop a Medium Range Ultrasonic Technique (MRUT) for inspecting mooring chains.
Moorings chains are a critical component of floating structures in offshore oil and gas production, which include drilling rigs, spar platforms and Floating, Production, Offloading and Storage vessels (FPSOs) and their liquid natural gas variant. There is also a growing use floating structures in the renewable energy sector, where towers for wind turbines in deep water can no long longer be fixed on the sea-bed. However, mooring chains for FPSOs have received the most attention and are the target application for this project.
Unlike ships, FPSOs stay at fixed positions year after year without regular dry docking for inspection and repair. Since they cannot move off station they must withstand whatever weather comes their way. Hence, depending on location, at times their mooring systems need to withstand high storm loadings. Typically, during design for harsh environments, mooring systems do not have much reserve capacity above what is required to withstand survival conditions. Therefore, deterioration of the lines over time can increase the likelihood of single or multiple line failures. A FPSO may have 12-18 mooring lines, but multiple line failure could conceivably result in a FPSO breaking away from the moorings and freely drifting into the middle of an oil field as has been seen in the past.
The primary purpose of a mooring system is to maintain the FPSO on station within a specified tolerance, typically based on an offset limit determined from the configuration of the risers. The mooring system provides a restoring force that acts against the environmental forces which want to push the FPSO off station.
Where the sea current flow is directional, the mooring lines can be tied directly to the vessel (Figure 1) in so-called ‘spread moorings’. Where it is multi-directional, the mooring lines are tied to a single point around which the FPSO rotates, rather like a weather vane. Single point moorings are a speciality of one of the MoorInspect project partners, Single Buoy Moorings (SBM) and therefore the MoorInspect specification concentrated on this application. For larger FPSOs the single point mooring is in the form of a ‘turret’, which can either be external (Figure 2) or internal (Figure 3) to the FPSO’s hull. Figure 4 shows the complexity of the turret. The close proximity of the moorings chains to the risers that bring oil and gas up from the sea bed adds to the criticality of their structural integrity.
FPSOs can operate in very deep water. The current record is 2600m for the Cascade and Chinook oil field in the Gulf of Mexico. For deep water moorings, the steel chain links are to be found in the top 20-30m near the turret and the bottom 20-30m near the anchor. The rest of the mooring is of polyester composite or wire rope. The MoorInspect project could only aim at developing a prototype for use on the chain links near the surface, where marinisation of GUW collar, robot, cabling and connectors is relatively straight forward and instrumentation can be kept at the surface. A greater added value could be added to the MoorInspect inspections, if they could be carried out near the mooring touch-down point on the sea bed. Hopefully this will be a future development for MoorInspect.
The first important feature of MoorInspect is its ability to inspect chains in situ. Chains on drilling rigs maybe lifted when the rig moves station and can be sent on-shore for inspection using conventional Non-Destructive Testing (NDT). This is mainly a close visual examination, accompanied with magnetic particle inspection (MPI) to reveal surface breaking cracks. This is not possible on FPSOs that remain fixed on station for several years. It is possible conduct visual inspection by rope access (Figure 5) or underwater by a simple ‘swim-by’, or even a more detailed survey by checking for elongation of the chains using callipers.
The second important feature of MoorInspect is its ability to detect cracks and corrosion/erosion in the intrados area of the chain’s inner circumference, where the chains link and rub together and which is hidden from view (inserts in Figure 5). Here a combination of fretting between the surfaces, erosion and corrosion and fatigue can lead to cracks radiating out into the chain. However the reasons for chain failure are complex and are an active area of research, particularly now that higher strength steels are being introduced to reduce the size of the chain. Cracks as much as 50% through wall can be tolerated without failure due to high design redundancy, but current NDT is limited in its ability to measure crack size without removing the chain. Currently MoorInspect is only able to screen chains for defects. It is able to assess the severity of the defect from the amplitude of the reflected signal, but to measure crack size or corrosion depth requires a ultrasonic technique, possibly using a phased array ultrasonic probe on a robotic arm in place of the GUW transducer collar.
The third and perhaps the most unique feature of MoorInspect, is a robot that is able to climb the chain from 25m below surface up through the splash zone to the chain hawse. The splash zone is a very hazardous zone to reach, either by rope access from above (Figure 6) or by divers from below. By its ability to climb through this zone MoorInspect offers a unique solution to a recurring problem. Moreover MoorInspect is designed so that its payload of GUW transducer collar can be replaced by a cleaning tool or even a robot arm (Figure 7), to which a conventional phased array probe can be attached to allow a detailed evaluation of chain cracks and corrosion.
The decision to build an autonomous climbing robot instead of a simpler capsule with GUW collar that could be lowered to the chain was made by the project consortium because of the concerns of project partner SBM. They believed inspection would otherwise interfere with operation of the FPSO. Their FPSOs ‘weather-vane’ about the turret in the prevailing currents and there may not be time to lower MoorInpsect from a boat on the surface to the chain, collect GUW data from that chain and raise MoorInspect back to the surface before the boat would have to move away from the FPSO. Instead the robot would need to operate autonomously. The decision to build a chain climbing robot was a very bold one, because of the high level of risk. A completely autonomous robot would not be feasible within the project resources, but a proof-of-concept with a prototype that was tethered with an umbilical to provide power and collect data was. However the project did incur cost and schedule over-runs.
The overall objectives of the MoorInspect project were included in two statements:
O1. The MoorInspect project will bring a step change to the current in-water moor chain inspection systems through the development and introduction of new method for detection of fatigue cracks in the large chain links used in Deepwater offshore facilities.
O2. The project will introduce a novel scientific methodology of new Non Destructive underwater Testing (NDT) using Medium Range Ultrasonic Transducer collar, coupled by a technology step change through a deployable prototype device
The level of achievement of O1 was high. The MoorInspect system brings a step change in current mooring chain inspection, which has been enhanced by providing an in-air as well as an in-water inspection capability, with unique capability for inspection in the splash-zone. Moreover, the autonomous nature of the climbing robot means that it can deployed to inspect chains to the more advanced FPSO’s that are attached to turrets around which they weather-vane, rather than just the spread moorings used in older FPSO’s.
The level of achievement of O2 was moderate. The original plan was to use higher frequency GUW in order to increase resolution to defects over current GUW techniques used on pipes. Unfortunately, the complexity of GUW propagation around chain links was not anticipated and low frequency (25-30KHz) ultrasound was used. However adequate sensitivity was demonstrated to use the technique as a screening tool, that detects the presence of fatigue cracks or corrosion and determines which side of the chain link should be evaluated further with another NDT technique, such as phased array ultrasonics.

The technical objectives of the MoorInspect project were:
T1. A two tier software that would be able to (i) interpret the results of the local In-water chain link scan, thus identifying defects signals and cracks, (ii) a spatial universal coordinates locator that would be able to identify the location of the relevant tested chain link and pinpoint its location within the moor chain system environment. This software would be essential step in providing an underwater deployable system with minimal underwater Human supervision.
T2. A prototype of the suggested system that can be deployed for underwater inspection. The prototype would be mounted on a deployment mechanised vehicle to ensure flexibility in application and reduction of development costs.
T3. Development of new safety standards and procedures to the current In-water inspection standards and codes.
The level of achievement of T1 was moderate. The second tier of the software was not achieved totally because of the low resolution power of the long wavelength ultrasound. Instead the defective chain link can be identified and the side (top or bottom) of the chain where the defect is located. The deploy ability of the system with minimal human intervention is achieved through use of automatic defect detection software.
The level of achievement of T2 was high. The flexibility of the system allows replacement of the GUW collar with other devices.
The level of achievement of T3 was low. There are currently no in-water inspection standards and codes and their development will take several years after acceptance of the MoorInspect technology.

In WP-1, the objectives were:
1.1 To agree among partners and in consultation with end-users, upon a specification for the system, which is both feasible in terms of project resources and desirable in terms of test piece geometry’s, defect sought, test sensitivity levels and inspection speed.
1.2 To obtain representative test samples for mooring chain links.
The level of achievement of 1.1 was moderate. The desirability of an autonomous chain climbing robot for deploying the GUW technique on the part of the end-users was compromised by its feasibility within the project resources.
The level of achievement of 1.2 high – a good set of 110mm and 160mm chain samples was obtained from SBM, Vicinay and affiliate TWI.

In WP-2 the objectives were:-
2.1 To understand the propagation of ultrasonic guided waves in mooring chain links through theoretical modelling work, and development of an ultrasonic guided wave technique for testing mooring chain links.
2.2 To design and manufacture a marinised transducer housing.
The level of achievement of 2.1 was moderate. The modelling work failed to explain the good sensitivity of the test at specific GUW frequencies. This was partially due to the complexity of wave modes generated in solid bars and the degree of distortion around bends. But it was also due to the complexity of the pulses generated by a 2-ring transducer ring that uses phase interference to give direction to the pulse. It is the intention to pursue this problem with further fundamental research work.
The level of achievement of 2.2 was high. Despite a manufacturing fault in the bladder that spoiled the final in-water demonstration, the hydraulically operated housing showed itself to be a robust design for sub-sea use, which project partner Sonomatic intends to develop for a number of applications

In WP-3 the objectives were:-
3.1 To develop a database for recognition of defects in the mooring chain links.
3.2 To produce a software for control, analysis and automated defect detection.
The level of achievement of 3.1 was high. The database will allow test data collected from future inspections to be stored in a structured way to help refine the automated defect detection and provided a ‘finger-print’ of chain integrity that can be monitored over periodic inspections.
The level of achievement of 3.2 was moderate. The automated defect detection software is still a work in progress, that requires more test data for operational inspections.

In WP-4 the objective was originally:-
4.1 To develop a deployment system for an inspection capsule that provides access to each link, in sequence, of a mooring chain immersed in sea water. The inspection capsule will deploy a cleaning system to remove marine growth on the link and then deploy transducer housings to inspect each link from a single position. It will also include a vision system.
This was revised to become:-
4.1r To develop a deployment system for an inspection capsule that provides access to each link, in sequence, of a mooring chain immersed in sea water. The inspection capsule will deploy transducer housings to inspect each link from a single position. The inspection capsule will also be capable of taking an off-the-shelf cleaning system in place of the transducer housing. It will also include a vision system.
The level of achievement of 4.1r was high. The inspection capsule will deploy the transducer housing above water as well as in water, and critically through the splash zone. An off-the-shelf cleaning system that can be mounted on the inspection capsule to clean the small area on the side of the chain, where the transducer collar is clamped, has not been identified, but the SME-partner is confident it can adapt one of its own pipe-cleaning tools to the task.

The objective of WP5 was:-
5.1 To assemble, integrate and test the MoorInspect prototype system.
The level of achievement of this objective was high.

The objective of WP6 was:-
6.1 To perform ultrasonic guided wave testing on real mooring chain links in-water.
The level of achievement of this objective was moderate. The MoorInspect system was demonstrated satisfactorily in-air, but due to a collar leakage caused by a manufacturing fault in the bladder could only be demonstrated in water with the collar removed. The collar was demonstrated in water separately.

The objective of WP7 was:-
7.1 To disseminate the project results (and techniques) as widely as possible within the target industry and to maximise exploitation opportunities for the SMEs.
The level of achievement of this objective was high. Due to the affiliate status of TWI to the project, which has a wide Industrial Member constituency in the offshore oil and gas industry, the project results have been disseminated widely to FPSO operators, insurance companies and authorities. This has been supplemented with dissemination to the international FPSO-forum.

The objective of WP8 was:-
8.1 To plan, organise and review all consortium and participant activities. To co-ordinate activities such that a durable structure is established, to report to the REA and the Project Steering Committee, monitor plans and operational performance of all activities. To assist participants and WP leaders to achieve their objectives. To review, control and report financial information (budget and costs) to the consortium and the REA.
The level of achievement of this objective was moderate. The project consortium worked enthusiastically together, with a high level of confidence in what each participant was contributing. This was perhaps slightly misplaced towards the end of the project, when it was disclosed that the marinisation trials before the final in-water demonstration could not be completed in time.

Project Results:
The research and development is divided between 3 Work Packages (WPs), each allocated to a Research and Technology Developer (RTD) with relevant expertise:
1. Development of marinised transducer housing. – Plant Integrity Ltd.
2. Development of MoorInspect software for control, data collection and automated defect detection. – IknowHow.
3. Development of prototype marinised inspection capsule for deployment of transducer housing in-water. – InnoTechUK.

Development of marinised transducer housing
The MRUT technique requires a transducer housing that can be clamped around the chain link by the climbing robot above or below water. To begin, a fundamental investigation of guided wave propagation around chain links was undertaken, before moving onto procedure and marinised transducer collar development.

Technique development
The MRUT is a variant of the Long Range Ultrasonic Technique (LRUT) that has been applied to pipelines for over a decade. LRUT uses a mode of ultrasonic wave propagation called the Guided Ultrasonic Wave (GUW) to distinguish it from the bulk ultrasonic wave used in conventional ultrasonics NDT. GUW can propagate distances of tens of metres. LRUT is used to screen pipes for corrosion and in exceptional cases for cracks. In order to propagate the GUW the pipe is encircled by a ring of ultrasound piezo-electric transducers or as an alternative method of generating ultrasonic stress waves, by a magneto-strictive strip encircled by an AC carrying coil. The magneto-strictive method is impracticable as an inspection technique since the magneto-strictive strip has to be bonded to the test piece. However a third alternative, using a ring of Electro-Magnetic Acoustic Transducers (EMATs) is practicable and so was investigated in the project.
The piezo-electric transducers are normally of Lead Zirconium Titanate (PZT) polarised longitudinally to produce a shear motion against the surface of the pipe. If transducers in the rings are aligned with the pipe axis, the resulting GUW will be longitudinal (L-wave). If aligned circumferentially, the resulting GUW will be torsional (T-wave). The ultrasound energy is transmitted from the PZT through the test surface by friction. Therefore as long as there is sufficient pressure applied on the transducer it will ‘couple’ to the test piece without the need for a liquid couplant. A force of 200Netwons is generally regarded as sufficient on each transducer and this is achieved by surrounding the rings of transducers in the collar with a bladder that can be inflated to a pressure of about 60psi (similar to a bicycle tyre). To propagate GUW in pipes, the transducers vibrate at certain critical frequencies, dependent on the wavelength and pipe wall thickness relationship.
As an alternative to shear-wave transducers, the MoorInspect project also investigated the use of compression-wave transducer. These are used in conventional ultrasonics NDT to propagate bulk waves. The PZT transducers are polarised through-thickness to produce a compression motion (L-wave). By placing these transducers on a wedge, the incident L-wave will refract a T-wave in the test piece. At specific angle of incidence – ultrasound frequency combinations, they propagate Lamb waves (a GUW in a plate).
In the MoorInspect project this alternative method of propagating GUW was investigated because placed on a wedge the GUW propagation is uni-directional, requiring only one ring of transducers. Shear-wave transducers are bi-directional, propagating shear stress waves forwards and backwards. The rings therefore have to be in pairs to which the driving pulses are phased so as to destructively interfere with the backward going GUW.
The compression transducers were mounted on a variable-angle (VA) wedge to propagate guided waves around the intrados of the chain (Figure 8). The angle was adjusted to generate the required guided wave by first maximising the echo-signal off the end of a bar, the same diameter as the chain. The broad band pulses from conventional ultrasonic flaw detectors gave very poor results and special equipment to produce pulses modulated with a ‘Han window’ to restrict pulse width had to be used. In this way good sensitivity was obtained to notches on the intrados, but only as far as the opposite side of the chain and not the whole chain circumference. This was because the ultrasound frequencies could not be reduced below 500KHz without making the wedge very large to accommodate a wavelength of ultrasound, a necessary pre-requisite for refracting the required GUW. However the possibility remains for using these higher frequencies with VA probes in an alternative transducer collar in the MoorInspect collar, which might be used to evaluate defects at short range.
The third alternative for propagating GUW was investigated in MoorInspect. EMATs were included, firstly because the rigid ceramic transducer might not sit properly on the tight curvature of the chain link, secondly because the chain’s surface roughness might make the inductive ultrasound coupling of the EMAT more efficient than PZT and thirdly because EMATs can be used to generate the Shear-Horizonatal (SH) wave. SH-waves are less affected by surface roughness than other GUW.
Chains are made from ferro-magnetic low carbon steels. Therefore the EMATs can rely on the magneto-strictive effect to generate ultrasound waves. By applying an alternating magnetic field from a coil across a permanent ‘bias’ magnetic that aligns the magnetic dipoles in the chain, stress waves can be generated. With careful selection of the bias field direction and strength, the coil turns, applied current and AC frequency, the stress waves can be ordered to produce specific GUW.
Trials were made with a number of EMAT configurations including encircling coil, pancake coil, meander coil and Periodic Permanent Magnet (PPM) coil (Figure 9).
The encircling coil EMATs were the simplest and most effective, but they were ruled out because of the impracticability of attaching to the chain in a clamp. However, as with the magneto-strictive transducers, encircling coils could be used in permanently installed sensors. Of the other EMATs, the PPM proved the most effective as a transmitter of GUW. It generates the SH-wave that is very sensitive to surface breaking cracks. However, the properties of the EMATs as receiver were found to be poor. Indeed the design specification for an optimum performance receiver EMAT (number of turns, coil size, magnetic field strength etc) are quite different from those for a transmitter EMAT. They should be separate ana a pitch-catch configuration adopted. Additional rings of transducers would therefore be needed.
Consequently, it was decided to progress with the marinised collar design using conventional PZT transducers. The coupling conditions for a prototype MoorInspect transducer collar are relatively benign and PZT transducers would suffice. With the MRUT procedures developed for PZT transducers it should be possible to modify these for EMATs at a later date.
To begin the technique development, the shear-wave PZT transducers were investigated using a modified LRUT pipe inspection tool on a straight 3m long solid cylindrical rod of the same stock as used in the chain links (110mm for the initial experiments). A notch at 1m from the end was cut to give a target reflector. The signals from the notch and bar ends could be identified in collected A-scans, where their time of flight could be measured. Knowing the actual distance of the reflector from the transducers allowed the group velocities could be calculated. These were important measurements, because the group velocities and corresponding phase velocities of GUW can vary with frequency. GUW are ‘dispersive, and if reflectors are to be positioned from the transducer the bulk velocity of the pulse must be known. Moreover, GUW exist in different modes, each with a different motion and a different velocity. The simplest wave modes in cylinders of interest in GUW are Torsional (T) wave, Longitudinal (L) wave and Flexural (F) wave. (Figure 10). The presence of wave modes and there velocity-frequency dependence can be derived from so-called ‘dispersion curves’ (Figure 11). These can be generated numerically using information about geometry and mechanical properties about the object through which the GUW are propagating.
When inspecting pipes, T-waves or L-waves can be propagated from the transducer rings. These are symmetrical around the pipe. When reflected from an area of corrosion, flexural waves can be expected, that are not symmetrical about the pipe. The GUW propagate in 6-10 cycle pulses long in the 20-100KHz ultrasound frequency band. The test frequency is selected so they are non-dispersive i.e. a flat region of the dispersion curve. T(0,1), the fundamental T-wave is non-dispersive, but the corresponding reflected flexural waves (F(1.2)) is very dispersive. L(0,1), the fundamental L-wave is highly dispersive and cannot be used, whereas the next order wave mode, the L(0,2) is only dispersive below certain frequencies, while the corresponding reflected flexural waves (F(1.3)) are less dispersive than the F(1,2) waves.
Clearly wave mode and frequency selection are important in developing a GUW technique. It is also important in transducer design. From one ring of T(0,1) transducers, GUW pulses will propagate in both directions along the pipe. To make them directional a second ring has to be introduced, where the same voltage pulse is applied 180deg out of phase and with a delay equal to the time it takes the wave at the first ring to reach the second ring. Thus the T-wave pulse in one direction is destructively interfered with, but reinforced in the other direction. The process is reversed to send T(0,1) pulses in the opposite direction. Obviously it is critical to know the phase velocity of the T(0,1) wave to give the necessary time delay to the second ring, if directionality is to be provided. As explained later, this directionality optimisation at certain frequency-ring spacing combinations proved critical in achieving high signal-to-noise when testing chains. From one ring of L-wave transducers there is additional complication of two wave modes being present in the frequency band used; L(0,1) and L(0,2). The highly dispersive L(0,1) has to be removed, leaving the L(0,2) wave to produce the signal. This is accomplished by adding a third ring, this time to destructively interfere with the L(0,1) going in either direction.
To detect F-waves, the transducer tool is divided into quadrants, the amplitude of the wave in each quadrant is compared. If it is the same in each quadrant, then the wave is symmetrical. If the wave is higher in the two top quadrants than it is the two lower quadrants then the flexing is in the vertical plane, and if it is higher in the left two quadrants than it is in the right two quadrants the wave is flexing in the horizontal plane. They are useful in GUW testing of pipes because symmetrical waves incident on corrosion become mode converted to flexural waves, unless in the rare cases the corrosion is symmetrical around the pipe circumference, when they remain as symmetrical waves.
When testing the solid rod, the A-scans collected with the T-wave and L-wave transducer collars were far more complex than those collected from pipes of the same diameter, because of the presence of additional wave modes. The least complicated on the 110mm bar were with T-waves at around 60KHz. The A-scans from L-wave tests were too complex to interpret.
The reasons for this became evident when the theoretical dispersion curves were constructed for GUW in two bar diameters (Figure 12). 110mm was used in the initial tests, 160mm in the subsequent tests after it was decided this would be the chain size for demonstrating the MoorInspect system. These showed additional wave modes were present at frequencies used in GUW testing (<100KHz). Of particular significance for the project was the shift in the dispersive high order T-waves to lower frequencies, when the diameter of the rod increased from 110mm to 160mm. The curves show that for the larger diameter the non-dispersive T(0,1) is accompanied by the highly dispersive T(0,2) and T(0,3). Finite element modelling showed the energy distribution of the T(0,1) is at the surface and of the T(0,2) is mid-way through. The project had set itself the objective of using higher frequencies for better sensitivity to small cracks, but now it appeared that this higher sensitivity would be undermined by the presence of additional wave modes. Eventually good results were obtained on the 160mm chain with T-waves, but in a low very narrow frequency band that corresponded with maximum output from the two-ring collar.
Chains are not straight rods. The GUW have to travel around two 180deg bends in one circuit of the chain. These will obviously distort the guided waves giving rise to flexural modes (F-waves) that have their own frequency-velocity dependency. The amount of distortion around bends was found by numerical modelling (Figure 13). Surprisingly, the experimental results showed that their remained a strong symmetrical content in the GUW that had encircled the chain. The experiments were done on several 110mm and 160mm chains using the modified two ring T-wave pipe-testing transducer collar (Figure 14). The collar was placed around the chain on the straight piece opposite to the weld in each test and the GUW pulses fired in both the forward and backward directions.
More surprising however, was that when slots were introduced into the intrados of the chain to represent fatigue cracks or machined flats to represent corrosion, the signals from the slots were visible only within a narrow frequency range, as previously explained as corresponding to the peak amplitude of the T(0,1) for a specific ring spacing. The sensitivity of this test was confirmed by observing an increase in the height of the defect signal when the slot or machined flat depth was increased, though not when its lateral length was extended (Figure 15).
An understanding of these phenomena is part of an on-going discussion. One hypothesis is that the symmetrical wave in the straight cylindrical sides of the chain becomes a torroidal version of the wave in the rounded chain ends (Figure 16). This reconverts to the cylindrical symmetrical GUW after it passes the bend. Moreover some of the flexural wave modes formed in the bend combine when entering the straight part of the chain to form symmetrical GUW. The numerical modelling did support this. The result is that at certain frequencies there is a regular pattern of symmetrical signals, each time the GUW pulse circuits the chain (Figure 17).

Procedure qualification
The introduction of any new NDT procedure requires its qualification and eventual validation. Validation is beyond the scope of this project as its requires extensive blind trials with samples that are subsequently sectioned to reveal nature, size and location of any implanted or natural flaws. The qualification was also limited to samples that were available of mooring chains taken out of service into which machined slots were implanted to simulate fatigue cracks on the intrados of the chain and machined flats to simulate corrosion. The various procedure parameters that influence the GUW technique performance on chains include chain size, ultrasound frequency, transducer ring separation, GUW mode, surface roughness, transducer loading. For reasons already mentioned, the chain size had far greater influence than was anticipated and a more procedure development work had to be done on the larger 160mm chains, which had been selected for demonstrating the MoorInspect prototype.

Marinised transducer collar development
The shear-wave PZT transducers for GUW collars are fixed in a housing to the electrodes and this must contain sufficient damping material to reduce transducer ringing after it has been fired by a voltage pulse. Keeping the so-called ‘dead-zone’ is important for testing chains because of the short test length. For underwater use the transducer housings had to be redesigned to prevent water ingress at either the connector or the transducer face. The housings were pressure tested to 35bar. To propagate symmetrical GUW, T(0,1), there is a minimum density to the transducers surrounding the chain. If reduced below this, high order flexural waves will be included. Moreover, the number of transducers must be dividable by 4, if flexural waves are to be distinguished from symmetrical ones in reception, as described earlier. For a 160mm chain-link 24 transducers were used in each ring. In common with conventional LRUT collars, it was modularised so that the transducers could be placed in collars of different diameter (Figure 18).
Marinisation of the cable connections to each transducer is problematic, because the integrity of the co-axial cable that transmits signals back to the instrument has to be maintained. A simple low-cost solution was followed with each transducer having its own integral cable. For the trials, 24m of cabling was used, which had to be reeled up. This is impracticable in an operational system and a connector should be included to detach the cabling from the tool.
There are two straight portions of chain link on which to place the transducer collar. However closer examination revealed that these had a slight curvature, which had to be accommodated by tilting of the individual transducers. One straight side has the flash butt weld upset and although slight, this was enough to reduce coupling efficiency when the transducers were placed down on it and was avoided in the tests. Fortunately in mooring chains, the weld in each link is always aligned down one side of the chain, so a simple solution was possible for the robot to place the transducer collar on each link.
To achieve the downward force on each transducer of 200Newton in water at depths down to 30m, the transducer modules are surrounded by a bladder which is filled with oil in place of air. The replacement of the pneumatic system used on pipes with a hydraulic one was costly. The hydraulics require hoses, valves, pumps and a power, controlled from a separate panel at the surface. Sucking oil out of the bladder became as much an issue as pumping oil into it. Each bladder therefore has its own inlet and outlet house, giving rise to a relatively large collar (Figure 19).
For clamping around the chain, the collar is divided into three. The middle section is pushed against the chain and the two outer sections are brought together and clamped. A proximity switch confirms that the lamp is engaged before the collar is inflated. Clearances between the arms of the collar and the chain sides were a critical issue. The space inside the chain link is limited and a simulation had to be conducted before manufacture. The angle of approach of the collar to the chain is at 45 deg to the face of the chain. This is to allow the collar carriage to remain on the same side of the robot as it goes from one chain link to the next. Otherwise the collar would have to be moved across from one side to the other.
Development of MoorInspect software for control, data collection and automated defect detection
The signals from GUW tests are displayed in A-scans. The signal from a pulse of GUW received by the transducers is seen on a time base that starts when the pulse is transmitted. By calibrating the time base for the group velocity of the pulse, the distance travelled by a signal can be found. In pulse-echo testing this allows the location of a defect to be measured from the transducer. The amplitude of the signal will be a measure of the size of the defect. The complexity of A-scans generated from GUW pulses propagating around a chain link however require a higher level of processing than is currently necessary for LRUT of pipes. This led to the development of a neural network to aid the interpretation of A-scans.
The first problem with GUW of chains is that the group velocity is determined by the wave mode and its frequency. The dispersion curves can provide this information, but only if we can identify the wave mode. Dispersion curves were developed, but showed that there were many wave modes present and all but the T(0,1) were dispersive, i.e. their velocity changes with frequency. Moreover, dispersion curves for all the wave modes in chains are not available yet. However, as with conventional ultrasonic testing, reference echo-signals were present with which to calibrate the A-scans. These were the repeat signals as the GUW pulse passed the transducer rings on each circuit of the chain link.
The second problem is that GUW pulses circle around the chain repeatedly and the transducer collar is a receiver of not only pulse-echo signals, but also of through transmission signals. These can be distinguished by carefully examining the A-scan of each ring individually to distinguish which ring the pulse is received at first.

Processing of signals from chains.
The A-scans are collected from each quadrant in each of the two rings, resulting in 8 A-scans in total. Standard processing for LRUT of pipes is able to cancel any backward coming pulse by adding a delay to the signal in the 2nd ring equal to the time taken for the pulse to travel from the 1st ring to the second and inverting it. The process is a reverse to the one used in transmission to give the test directionality. In order to do this, the phase velocity of the wave in the pulse must be known. For chains, this is not known. Practically however this was not a problem. Using the same processing for the chains as is used for pipelines, the pulse-echo signals from the defects and the through-transmission signals from circuits of the chain were clearly visible (Figure 20), but only at specific frequencies. This is a clear indication that the symmetrical signals observed from circuits of the chain by the GUW pulse do not have the same phase velocity as transmitted T(0,1) wave mode. This phenomenon and others encountered in qualifying the GUW procedure for 160mm chains have yet to be explained.
What was also apparent was the change in pattern between chains with and without defects (Figure 17). A pulse-echo signal was clearly visible from the defect and the presence of a defect also affected the circuiting through-transmission signal and its repeats. The amplitude of the signal also increased with the defect depth, but not with its circumferential length (Figure 15). A pattern recognition method was therefore investigated for incorporation in a neural network for automatically recognising the presence of a defect.
Figure 21 gives an outline of the software used in MoorInspect to automatically detect defects in chains. It comprises a defect identification module and a defect location module.
Training a neural network requires large amounts of data. This was accomplished by using 20-120KHz frequency sweeps of chains without defects, with slots and with machined flats. This produced a total of 1440 A-scans from the 160mm chains alone. A-scans collected ‘raw’ from each of the two rings and each octant independently or combined (Figure 22), A-scans collected after using the two rings as a phased array to transmit in each direction independently (Figure 23) and A-scans processed using the standard processing routine used on pipelines (Figure 24) were all used to train the neural network. The principle for processed data collection on each chain is illustrated in Figure 25.
The data was collected in both the forward and backward directions from the transducer collar placed opposite the weld. The transducer collar was a modified version of a pipe testing collar, adjusted to fit the diameter and with 2-ring torsional operation only. From the dispersion curves it was evident that the non-dispersive T(0,1) wave used in pipe testing would be accompanied in all but the very low frequencies, by the dispersive T(0,2) and T(0.3) wave modes. The GUW A-scan data was collected in frequency sweeps from 20KHz to 100KHz and in 6cycle or 10cycle pulses. As already mentioned, on the 160mm chain the processed data gave very poor A-scans with a high level of noise except within a narrow band of frequencies near the peak of the transducer output curve. Therefore it was decided to investigate the use of a number of special signal processing algorithms with which to ‘clean-up’ the raw data at other GUW frequencies before feeding it into the neural network.
Using the 1440 ‘raw’ A-scans from the phased directional collar on the seven 160mm chains, the neural network was a little over 70% successful in distinguishing defective from defect-free chains.
The first processing algorithm used correlation coefficients and resulted in an improvement in detection to about 80%.
The second processing algorithm used ‘Power Spectral Density’ (PSD) calculations of each A-scan. ‘Fast Fourier Transforms’ (FFTs) were used to extract the frequencies within bands defined through a number of methodologies. The results were not as good as with the raw data.
For the third processing algorithm, the PSD method was augmented by a (Spectrogram) analysis. It provides information in both the time and frequency domain. A classification success rate of over 99% was achieved.
The defect location part of the software is activated when the NN identifies a defect. It detects peaks in the A-scans and uses the repeat through-transmission signals as the pulse circulates the chain link as the ‘base peaks’ for calibration purposes. Even using standard processed A-scans on defect free chain links, it was found that the A-scans became too complex after the second circuit of the chain to be usable. As with the development of the defect identification part of the software, the use of ‘raw’ A-scan data was investigated as well as A-scans processed with the standard method.
Again useful standard processed A-scans were limited to a narrow frequency bandwidth at the maximum output form the two-ring array. The alternative raw data A-scans were de-noised using a Stationary Wavelet Transform (SWT), and this resulted in improved signal to noise separation over the standard processed A-scans (Figure 26).
Because of the essential symmetry between GUW pulses propagating forwards and GUW pulses propagating backwards from the transducer collar, it may be possible to identify and locate defects with raw data from just one ring. This would be very advantageous for a continuous monitoring system, where only changes in defect size need be identified without the need for positioning. The cost of the transducer collar could be reduced significantly for permanently installed collars, wired to a data gathering centre or the data is collected periodically. A permanently installed variant of the MoorInspect system could be a future development.

Defect recognition database
The ultimate performance of the neural network will depend upon the building of a database with A-scans over a long period from many different chains, with and without defects. This can only be done with an operational system. The framework for this database is ready using .NET programming framework and MS SQL server into which data is fed in ASCI files from the GUW data recording instrument. Presently this is a manual transfer, but an operational system would need direct link to the GUW instrument.
A Human-Machine Interface (HMI) has been developed in MS Visual Studio 2010 for the test operator to record essential reporting information, such as data and time of inspection, chain link identification, results for defect type, severity and location (Figure 27).

Summary of software development
Automatic defect detection software has been developed that achieved a 90% success rate in defect signal recognition on the samples of 160mm chain available in the trials. Moreover it was able to distinguish between two levels of defect severity and between slots and machined flats. These results were obtained with the optimum frequency used for identifying the defect signals visually, but progress was also made in developing custom signal processing algorithms that distinguished the defect signals in much higher frequency (91kHz) A-scans. This needs further investigation, because higher test frequencies afford better defect resolution.
The automatic detection software was also able to detect the presence of defects in the unprocessed A-scans from one ring only ie. GUW pulses propagating in both directions from the transducer ring. This is a significant development for single-ring operation for permanently installed sensors for monitoring mooring chains with known defects.
The defect location software is only partially developed because of uncertainties in the wave modes present and their bulk velocities. However as a chain screening tool, this is not so important because defective chains would be evaluated further with different NDT techniques to determine the nature,
Development of prototype marinised inspection capsule for deployment of transducer housing in-water
The aim of this part of the work programme was to develop a robot that can climb the chain, above and below the water-line, to deploy the inspection collar at individual chain links. Climbing is complicated by the fact that the plane of each chain link is at 90° degree to the next.
The transducer collar is carried on top of the chain-climbing robot. The chain climbing mechanism is based on two retractable beams that thread the chain. The first beam is placed on the crown of the first chain while the second moves up to the second chain using a ball screw. Once this is in place the first beam is lifted past the second beam to the next link using another ball screw, lifting the frame with it. The maximum payload in air is 40Kg. The robot itself weighs about 450Kg.

Among the considerations in its design were:
• To reduce its weight the frame is constructed in aluminium, but to maintain sufficient strength the arms, these are made in stainless steel.
• The frame is totally enclosed and contains enough rollers to guide the robot up the chain even if it is twisted.
• The frame comes in two halves with a latching system that would allow divers to clamp the robot onto the chain under water. The frame will fit a range of chain sizes.
• The welds in alternate chain links are all aligned. The collar can therefore be carried in one corner of the frame on a diagonally orientated arm that pushes the collar against the chain link, then when moved to the next link need only be displaced laterally a few centimetres to align the arm up for pushing the collar forward to the next link.
• By placing the collar on top, it can be moved up close to the hawse at the top of the chain.
• The collar can be interchanged with a special collar with VA probes or EMATs for defect evaluation, cleaning tools or cameras for visual inspection.

Initial designs
Seven traction mechanisms were considered before deciding upon the final one. These included the use of magnetic wheels and ‘grippers’. Models were built and ‘brain-storming’ sessions with technical staff held to establish the advantages and disadvantage of each.
The structure proposed finally imitated the method used by a human trying to climb up/down the chain underwater and in air (Figure 28).
The robots frame was designed to embrace the chain in order to reduce the likelihood of losing the robot and assure that the robot worked when the chain was in various inclination angles. The frame could easily be upgraded to work with different sizes and types of chains or even when two consecutive chain links are not perfectly placed at 90 degree between both of them.
The hooks could be considered as arms with two different types of movement: Up/down to climb the chain (Using stainless steel screw bars) and Close/open to support the robot structure weight over each chain link. These two movements, per hook, had to be perfectly synchronized by the control system in order to reach the next chain link. A set of motor-gearboxes ensured smooth movement and enough torque to counteract all the external forces in the worst working environment (in splash zone).
Underwater, the robot would move more easily due to the Archimedes’ principle. Therefore the possibility of including buoyancy aids in the design to reduce the weight of the robot underwater was considered and, hence, to reduce the power needs.
The set of rollers are necessary to slide the frame over the chain and make its movement easier.
The motors and the rest of electronic devices needed are supplied, monitored and controlled from the surface through an umbilical cable.
The motors, webcams, transducer collar are placed on a platform at the top of the robot.

Final designs
After starting the manufacturing of the initial design several technical issues came to light. These included:-
• Taking into account the length of the robot (around 1.5 meters), its weight (about 450kg) and the distance between the ball screws (1m) and the force generation point (Centre of the robot), the physical moment generated on the ball screws and on the arms was found to be too great and could bend and break them. The force needed to lift the robot has to take into account the inertial forces (acceleration and deceleration between moves) as well as the robot weight. This can be controlled by the operating procedure and the weight of materials used in the fabrication. However to reduce the moment on the arms it became apparent that a bent rotating arm design was too weak and should be replaced with a straight arm that went in and out between the chains.
• The presence of twists between chain links. After performing a set of 3D model simulations, on a chain with a misalignment of around 8 degrees it was found the current position of the arms and frame configuration could not assure that the robot could always move along the mooring chain. Therefore the frame of the inspection capsule had to completely enclose the chain to prevent it dropping off and several rollers would be needed to rotate the frame relative to the centre of the chain.
• The frame needed to be welded rather than of bolted construction to increase rigidity and reduce weight. However to allow the chain to be mounted on the chain underwater by divers, it would be split into two halves that would be clamped together.
• To simplify the transducer collar movements between chain links the current rotary movement could be eliminated by replacing it with a sliding mechanism angled at 45 degree to the face of the chain, such that the collar can be moved to one side or the other and then pushed forward to connect with the side of the chain link.

The final design is shown in Figure 29. The motors and the Gearboxes (Two per arm) are marinised by means of housing boxes. These are manufactured in aluminium to reduce their weight. The wall thickness has been calculated to support the water pressure in excess of that specified for final demonstration. The cables (Power and comm’s cables) are routed form the outside to inside by means of cable glands and the motor shaft is sealed by means of a mechanical seal system using a spring solution.
The transducer collar is placed on the top of the robot and all the movements performed by means of linear guides and ball screws. Three electrical motors perform the following movements.
1.- Align collar with chain side by linear movement at 45degree to chain face.
2.- Open and close the NDT collar
3.- Move the NDT collar in and out to chain link.

Robot control
The robot is controlled from a rack at the surface (Figure 30). For the prototype a 24m umbilical was considered sufficient. This resulted in an unwieldy mass of cabling and an operational system would have to consider means of reducing this, using connects and junction boxes for example, all of which would have added considerable cost for a prototype used for demonstration purposes only.

The rack contains the programmable logic controller (PLC), servos to control the electrical motors installed on the robot and a Power Distribution Unit (PDU) with a set of electrical switches to shut down the electrical power in case of a short circuit. This is very important for use in water.
In order to control the servos and the motors, control software has been developed using Labview. This is installed on a laptop. Communication with the PLC is performed through EtherCaT, through which all the in-out and up-down robot functionalities can be controlled and monitored.
The low level software to control the servos from the PLC has been coded using a KOP language provided by the PLC provider.
Prototype demonstration.
Demonstration of the complete MoorInspect system in-air was carried out for the partners at the final project meeting on 27th September 2013. The in-water demonstration is to take place in a diver-training tank at TWI’s Middlesbrough facility on 16th December 2013.
The robot was demonstrated on a 4-link chain suspended from a gantry chain (Figure 31). It was shown climbing the chain and placing the transducer collar around the chain to collect data, before returning to the ground.

OVERALL MOORINSPECT SYSTEM
1a) The pertinence and overall usefulness of the results obtained
The use of GUW as an NDT method is a recent innovation for inspecting pipes. Its use for inspecting mooring chains is entirely new. Its mode of operation is also new. Pipes can be inspected manually, but mooring chains are inherently dangerous and close-up inspection is only possible from an ROV or under very tightly controlled H&S conditions, by a diver. Therefore, robot deployment is key to applying GUW inspection to chains. Finally the GUW data obtained from chains is complex because of the additional signals caused by the ‘race-track’ effect of GUW pulses circulating the chain and the presence of several GUW modes. The defect recognition software and database can be used to alleviate this problem.

1b) progress beyond state-of-the-art
Proper in-situ inspection of mooring chains is not possible currently. Identification of major damage might be possible visually from ‘swim-bys’ and there are ROV deployed instruments for measuring elongation of individual chain links, but these measurements cannot be used to detect the presence of cracks and corrosion.
1c) expected impacts
There is major concern among FPSO operators about the structural integrity of their mooring chains and there is considerable interest in the results of MoorInspect. Moreover LE-Partner sees the results of MoorInspect as part of the process in developing a business for leasing the chains that they manufacture.
2a) adequateness of the S&T methodology followed
The S&T methodology split the R&D activity between GUW procedure/equipment development, software development and robot development, which were run concurrently, whereas a more productive approach might have had them run sequentially. This would not have been possible within the project’s time frame.
2b) validity of the conclusions drawn/results obtained
The system has achieved TRL6, a prototype demonstrated in a relevant environment. The validation of the equipment in terms of the sensitivity to defects in chains will require extensive blind trials with multiple samples of chains with and without natural defects that can be subsequently sectioned to measure and compare with GUW test results.
3a) compliance with the contract, the deviations (with explanation) and the use of resources
A major problem occurred because RTD-partner Innotech did not inform the project co-ordinator until month 23 that the testing of the chain climbing robot’s water-tightness, before the in-water demonstration would not be possible before the project end-date. The robot was at this time operating and the marinised drive mechanism and motors had been designed and built. Innotech wanted a thorough testing programme to prove there were not any leaks. The problem was compounded by the unavailability of a slot for underwater trials at TWI’s diver training facility until December month 27.
3b) general lessons learned from the implementation of the project that may be useful for future projects.
There was slippage in several deadlines during the project, which, if they had been challenged earlier might have high-lighted the marinisation problem.
The effort to meet the specific requirements of LE-partner SBM over-ruled the equipment specification in the DoW, leading to conflict with the REA.

INSPECTION CAPSULE
1a) The pertinence and overall usefulness of the results obtained
The most novel component of the MoorInspect system is the Inspection capsule. It is designed to climb through the critical ‘splash zone’, where neither diver inspectors from the sea below nor ‘rope access’ technicians from the air above can operate.
1b) progress beyond state-of-the-art
There is no equivalent equipment for use on chains.
1c) expected impacts
Autonomous operation on mooring chains attached to turrets around which FPS0’s weather-vane is critical for in-situ mooring chain inspection.
2a) adequateness of the S&T methodology followed
The development of an autonomous chain climbing robot that could operate in-air as well as in-water within WP4 proved a heavy drain on resources. Having designed and built the prototype, not enough time was allocated to proving the sealing of the motors used to power the climbing mechanism.
2b) validity of the conclusions drawn/results obtained
The robot was demonstrated climbing a chain in water within the controlled environment of TWI’s diver training tank. Although this was a relevant environment, in an operational environment there would be additional stresses placed on the equipment from waves and wind. This aspect was considered at some length by SME-partner Sonomatic, who have experience in subsea inspections and led to modifications to the frame of the robot, which they believe would minimise the additional loads. However this cannot be proven without an offshore demonstration.
3a) compliance with the contract, the deviations (with explanation) and the use of resources
The original contract envisaged an inspection capsule that would be lowered to the chain on a tether, where it would be attached by divers. It would operate only in-water. The change to an autonomous robot that could climb the chain from in-water to in-air was a major effort that used up resources that had been allocated to the cleaning system.
3b) general lessons learned from the implementation of the project that may be useful for future projects.
A better analysis of the costs of developing an autonomous chain climbing robot might have led RTD-partner to limit its efforts to a tethered in-water inspection capsule, but this would not have been of limited use to LE-partner SBM.

TRANSDUCER DESIGN
1a) The pertinence and overall usefulness of the results obtained
The design of the transducer allows sub-sea use and use in environments where ingress of liquids, soils or dust will lead to loss of performance either as transducers (propagation and reception of guided waves, or as receivers (acoustic emission). This was accomplished by simplifying the design to allow ‘potting’ of the transducer housing with a filler, use of an integral cable to reduce problems with marinised connectors and the use of coatings to seal the transducer. The design also allows for use of cross-polarised piezo-electric plates in place of conventional longitudinally polarised ones. This is a major advantage when generating T-waves from the internal bore of pipes.
1b) progress beyond state-of-the-art
Transducers used in GUW inspection need to operate in rings that act as array encircling the pipe. They must perform uniformly and consistently. The transducers are able to do this environments that would otherwise cause failure.
1c) expected impacts
Use of shear wave transducers underwater and in heavily soiled environments for generating Lamb waves and Rayleigh waves as well as GUW to inspect structures in environments that are currently inaccessible. More rugged and durable transducer housing design for use in permanently installed monitoring systems.
2a) adequateness of the S&T methodology followed
The transducers were designed within WP-2 of the project and were only part of the development of a marinised transducer collar used in the GUW procedure for chains that provided data with which to build the defect detection software in WP-2. The project worked on EMATs as an alternative to piezo-electric transducers, and a meander coil design was used to propagate GUW, but their performance was found not to equate with piezo-electrics and the their development had to be curtailed.
2b) validity of the conclusions drawn/results obtained
The transducers performed on the test bed within a pressure chamber without any deterioration in performance, but the project provided only limited opportunity to assess long-term performance in water.
3a) compliance with the contract, the deviations (with explanation) and the use of resources
The transducers developed were in compliance with contract. A decision had to made between using peizo-electric or EMAT transducers. Piezo- electric transducers are more difficult to couple ultrasonically to the chain, so more resources had to be diverted to the hydraulic system for pressing the transducers onto the chain surface.
3b) general lessons learned from the implementation of the project that may be useful for future projects.
When two alternative technologies are proposed (piezo-electric and EMAT) as a solution to a problem (propagating GUW around chain links), then the risks and costs associated with each should be defined in the proposal. Clearly there were not enough resources to develop both and a decision point, perhaps a milestone should be included in the project plan

MARINISED TRANSDUCER COLLAR
1a) The pertinence and overall usefulness of the results obtained
When the decision was made to adopt piezo-electric transducers over EMAT transducers, the design of a collar that could adequately press the rings of transducers down onto the roughened and slightly curved sides of the chain link became critical. The adoption of an hydraulic inflation system in place of the pneumatic one used in GUW inspections of pipes and the associated control system has produced an equipment that can be modified to test a range of pipe as well as chain diameters. Moreover the collar can be adopted for diver or ROV deployment.
The hydraulic system uses inflation bladders, but these could be replaced with micro-cylinders on individual transducers, giving much greater control on transducer pressure. This would benefit current in-air GUW inspection of pipes as well underwater GUW inspection and could be used in high-temperature inspections where polymer bladders cannot be used.
1b) progress beyond state-of-the-art
Although the basic collar design, two rings of transducers in modules that can be detached and inserted in a frame that can be sized to fit other chain-link or pipe diameters is current state-of-the-art, the hydraulic inflation mechanism is unique. Unlike a pneumatic inflation mechanism, the hydraulic one includes deflation as well as inflation pumps on the bladders. This process could be adapted for use with micro-cylinders on individual transducers, widening the application of the GUW collar.
1c) expected impacts
The hydraulically operated GUW collar will find use in diver and ROV deployed inspections as well as from the chain climbing robot.
2a) adequateness of the S&T methodology followed
The S&T methodology employed, which involved developing and qualifying the GUW test procedures on a range of chain link samples in air, with and without machined defects, uncovered a set of influencing test parameters, including the very significant one of chain diameter. Understanding this influential parameter played an important role in developing the GUW system.
2b) validity of the conclusions drawn/results obtained
Although in-air demonstration of the transducer collar was successful, the in-water demonstration was incomplete due to failures in the inflation bladder, the first caused by a design fault, which was rectified, the second by a manufacturing fault that could not be rectified in time for the final in-water demonstration. Although a replacement bladder was used in the final demonstration, H&S concerns about possible contamination of TWI’s diver training tank meant that the collar could not be demonstrated with the robot. It had to be demonstrated separately in a small tank used by TWI for equipment demonstration purposes only, without the presence of divers. Moreover, because of dangers from potential oil leaks the demonstration had to be made at half normal pressure, ruding the signal-to-noise separation of the resulting test data.
3a) compliance with the contract, the deviations (with explanation) and the use of resources
EMAT development was abandoned at an early stage, because of poor sensitivity and the resources transferred to developing an hydraulic inflation system that could be used under-water with piezo-electric transducers.
3b) general lessons learned from the implementation of the project that may be useful for future projects.
The development of GUW procedure and the transducer collar (WP2) was planned to run concurrently with the development of the automatic defect detection software (WP3). Since most of the work in WP3 has to follow WP2, the delays in developing the GUW procedure had a large impact on WP3 the possibility of which was not accounted for in the DoW.

AUTOMATED DEFECT DETECTION SOFTWARE
1a) The pertinence and overall usefulness of the results obtained
The GUW test data obtained from chain links are more complex than those obtained from pipes and an automated defect detection software is needed to aid the interpretation of this data to provide information about the nature, size and location of defects in the chain link. The expert system at the core of this system has been shown using data processed using a variety of algorithms developed in the project . This data, which is in the form of A-scans, has come from the GUW instrument in three ways; firstly processed using standard procedures; secondly as raw data from a 2-ring GUW collar; thirdly as from 1-ring collar. This last is significant for developing a permanently installed version of MoorInspect. A database has been developed, which will be built up with GUW inspection data overtime in order to train the expert system. The database also has the potential purpose of providing a ‘fingerprint’ of each chain link for monitoring its condition over a series of in-situ inspections from its installation until decommissioning.
1b) progress beyond state-of-the-art
There are no current automated defect detection systems for GUW test data.
1c)expected impacts
The software will help in the validation of the GUW technique. The process, which uses the presence of multiple repeat signals in the A-scans as the GUW pulse circles around the chain can be adopted for other applications where defects causes changes in the pattern of multiple repeat signals, such as the ultrasonic inspection of spot welds in automotive panels.
2a) adequateness of the S&T methodology followed
Training of expert systems to automatically detect defect signals in ultrasound A-scans requires large amounts of data. This must extend beyond the time horizon of the MoorInspect project. For this reason, the development of a database for storing GUW test data from chains is important. The investigation of new algorithms for processing the GUW test data was also beneficial in opening-up possibilities defect location and single-ring operation.
2b) validity of the conclusions drawn/results obtained
Although probability-of-detection success rates in excess of 95% were achieved with the software, this was on a limited number of chain link samples. The defect location measurements cannot be validated until the group velocities of GUW pulses in chain links is known. For this a better understanding of GUW propagation around chain links is needed, taking into account the distortion of pule shape around the bends at each end of the chain.
3a) compliance with the contract, the deviations (with explanation) and the use of resources
The work complied with contract.
3b) general lessons learned from the implementation of the project that may be useful for future projects.
Automated defect detection software development is limited prior to development of the test procedure and test equipment hardware.

Potential Impact:
To discuss the potential impact of the MoorInspect system, we must consider its Technology Readiness Level (TRL) at the end of the project. In terms of the graph (Figure 32), the MoorInspect project has seen a concept at TRL1 progress to a prototype demonstrated in a relevant environment at TRL6.
The next step will be to demonstrate the MoorInspect in an operational environment. In an offshore environment this is impossible without the support of an end-user. To this end contacts have been secured with the FPSO-Forum of FPSO operators and specific communications have been had with two major oil companies and with an offshore engineering contractor with sub-sea expertise. Moreover, there is the possibility of taking part in a technology assessment being proposed in group sponsored project circulated to industrial members of TWI to develop a best practice guide for the inspection and structural health monitoring of mooring chains.
A key feature of the TRL graph is the development of an operational system must be ‘pulled’ by the market rather than ‘pushed’ by the technology. It also requires a different set of skills, aligned with the aims of production engineering rather than research and development. Finally it illustrates that the rate of spend increases as the technology passes to TRL8 and TRL9. It shows that none of the SME-partners of the MoorInspect project can proceed on its own. The development can only continue as a collaboration with support from one or more large enterprises. Moreover, the collaboration can only succeed with each SME-partner providing a part of the supply chain, Sonomatic for underwater engineering, Bytest for system builds, Orme for software development and Robotnik for expertise in robotics.

Other potential impacts of the MoorInspect project include:
• A transducer design for use in harsh environments including the internal bores of oil-field tubulars for in-situ inspection.
• A marinised transducer collar that can be adapted for diver or ROV use and inspection of underwater pipelines and risers.
• Signal processing software and neural network, which can be used to aid the detection of defect signals, where there is a strong repeat pattern in the A-scans, for example the ultrasonic testing of spot welds.
• A robot that is able to climb a chain through the hazardous splash zone carrying one of a variety of payloads including a cleaning system, a close visual system or a robot arm with phased array probe to evaluate defects detected by GUW.

Summary
As with many FP7 projects, the objectives of MoorInspect have been very ambitious. To develop an operational inspection robot for inspecting mooring chains on a €2m budget is a ‘tall order’, but the MoorInspect project has resulted in a working prototype that will shortly be demonstrated in a relevant environment. In terms of technology readiness level (TRL) this is TRL7. The next step is to demonstrate the prototype in an operational environment. This will be a very expensive activity and support is being sought from operators or owners of FPSOs. If the funding needed to take the MoorInspect prototype forward to an operational system is not available in one project, then a ‘step-by-step’ approach is being advocated. Although the whole system may be at prototype stage (TRL7), there are components that are a higher TRL and could be operational now. For example the marinised transducer modules could be incorporated into a light-weight collar with hydraulic inflation that could be applied by a diver just below the surface in an internal turret. These are the chain links that more commonly fail. Practical experience could then be used to make modifications and take the design forward.

Acknowledgments
MoorInspect is collaboration between the following organisations: ByTest, Orme, Robotnik Automation, Sonomatic, InnotecUK, iKnowHow, Plant Integrity, Vicinay Cadenas and SBM. The Project is partly funded by the EC under the Research for the Benefit of SMEs programme in Grant Agreement 286976.


For Dissemination Activities and the explotation of results please refer to attachement "4.1 Final Publishable Summary Report The explotation of results and dissemination activities"

List of Websites:

www.moorinspect.eu
Contact List attached.