Final Report Summary - MOLD4PRODE (Intelligent Mold for Productivity Enhancement)
Executive Summary:
The constant demand for diversified/personalized goods leads to decrease time-to-market, and to plastic injection moulds with smaller batch production, with no stock. This generates huge productivity losses in the moulding industry:
- Tool Tuning is too long for tool makers before delivery of molds.
- New production run set-up time is too long and generates too much scraps, due to different “operator and machine” parameters.
Many approaches are currently adopted to reduce mold design and production time, but none of them addresses directly the parts productivity enhancement, where “trial error” is still the most employed engineering approach. The innovative approach of Mold4ProdE was to develop and assess a methodology to tool makers, for them to propose and deliver quicker “turnkey intelligent molds”, to clients with a reasonable extra cost, and offering them significant productivity enhancement.
The main purpose of the project is to:
- optimize the number and position of sensors in the mold via a knowledge approach, and also via the development of a virtual approach OSLO (Optimal Sensor Location) (WP2)
- implement and assess the concept “Mold Signature” by using the data driven models, built from experimental data collected during production runs with sensors embedded in process equipments (WP3).
These 2 points have been achieved by two different ways, the first based on experience via the implementation of a Knowledge Base System, the other via a complete virtual simulation approach. It aims at applying modern data mining techniques for modeling polymer injection molding process, based on virtual data coming from virtual sensors implemented in a CAD model of the mold.
Mold4ProdE has also provided dedicated and optimized components for tool instrumentation, data collection and treatment (WP4). The project delivers the basis to efficiently use “intelligent moulds”.
The project was organized in 3 successive loops. Each loop Integrates work on the theoretical aspects (RTD Work Packages 2, 3 and4) and work on demonstration cases with an update of the related deliverables.
It means that French, Italian, Spanish and German Technical Center have managed 3 different cases during the project, so 12 moulds, with 12 different mould makers, have been instrumented, tested and evaluated during the project (WP5).
The assessments of sensors embedded in moulds have evaluated benefits in mould tuning time and mould new production set up time. The benefits on tuning time can reach 30%, but needs trained people on data collection and treatment. This learning curve has been highlighted during the 3 years of the project. On new production set up time, the benefits are quicker and can reach more than 30%.
16 training sessions have been organized at regional level to Mold4ProdE mould makers and converters, to allow them to integrate the approaches developed in the project, and also the existing knowledge(WP6).
This knowledge is collected in 5 training modules available on the project website for all partners of the project. This will help the European mold makers to move to the knowledge based service industry, identified as a key factor to increase their competitiveness.
Project Context and Objectives:
WP2: Methodology for optimized sensors implementation in tools (RTD)
The main objective of the Mold4ProdE project is to develop and assess methodologies that help tool makers deliver quicker ‘turn key intelligent’ tools to end users. These tools will allow: (1) selecting the most proper type of sensors and the best location inside the mold, and (2) providing optimisation algorithms for each phase of the whole injection molding process. This addresses the mold signature and holds significant benefits for increasing productivity, cost/scrap reductions and improved quality of the resulting part.
WP2 has been carried out in parallel with WP3 in order to provide a global solution for sensors implementation in injection moulds. The common goal of WP2 and WP3 is to converge together to a practical definition of “mould signature” associated to the concept of intelligent turn-key injection mould. To look for a solution to optimal sensor implementation problem, two ways are considered.
- The first way is based on knowledge approach. This approach consists in accumulating knowledge to build a relevant use cases database and to provide tools to exploit it. The objectives of this first period were to built, test and provide a first implementation of the Knowledge tool.
- The second approach consists in exploiting numerical simulation tools (polymer injection moulding simulation software) to define numerically the best sensor configuration. The objective of such an approach is to provide information on sensors location at early stage during the development of the mold.
WP3: Algorithms and software for mold assessment
This WP contributes to the project challenges by:
1. Broadening the conception of mould signature to address current industry needs, and then describes some needed developments to convert that vision to reality. This vision pull together the whole injection moulding story, that is, the big picture, integrating existing fragmented ideas of machine, material, process, production and information control with the development of tools to produce a fully optimised manufacturing strategy. In this paradigm, each phase of the whole process will have consistent setup/optimisation procedures. The main objective is to provide a specific direction for improving product quality during the whole life cycle of the product: from the design to the production. A general scheme of the whole injection moulding process was defined and a software application was created in order to help mold makers and converters to know what information is necessary to gain a feasible mold signature definition.
2. Providing optimised algorithms and mathematical treatments for data processing in order to build a predictive model capable of detecting the quality of the molded plastic part. The predictive model calculates the mathematical distance between the related processed data and the mold signature. The creation of the predictive model needs to gain the following specific objectives:
• Decide what variables characterise the process (data collecting).
• Data dimensionality reduction.
• Selection of the best variables (best predictors).
• Building the predictive model.
• Evaluate the quality of the model by using different indicators.
Different existing data-driven modelling approaches were analysed and adapted for this specific problem. The predictive model was built by using real data collected during production runs with sensors embedded in process equipments.
WP4: Sensors integration technology and data acquisition and control system
The WP4 is devoted to the identification of existing solutions in the field of mold monitoring, either sensors and Data Acquisition and Control System systems, as well as the identification of best practices in deploying those solutions and the implementation of algorithms developed in WP3.
The main objectives are:
1. To create a catalogue of sensing technologies and good implementation practices in the field of injection
molding monitoring
2. To identify potential improvements in existing solutions
3. To implement the algorithms developed in WP3 in a real processing unit
According to these objectives, the WP4 is split into 3 main tasks:
Task 4.1 Sensors and Data Acquisition & Control Systems (DACS) in molding monitoring
Sub task 4.1.1 Survey on existing solutions
Sub task 4.1.2 Practical implementation of sensors and data transmission
Sub task 4.1.3 Data acquisition and processing systems
Task 4.2 Definition and specifications for Molding machine interoperability
Task 4.3 Algorithm integration in selected DACS (data acquisition & Control System)
WP5: Demonstration Case Studies
The WP5 moves from the necessity to evaluate and to test the M4PE approach that is developed, mainly, in the frame of WP2, WP3 and WP4.
Several aspects were developed in those workpackages from the implementation of a standardised methodology for choosing the best sensors configuration (based on knowledge, experience and software simulations) for developing the electronic devices for acquisition of the information and the software for its mining and elaboration.
The intent of the real experimentation is to practice the global approach and assess what are the main outcomes from real cases.
The main interest for the Consortium is acquiring experience and data about Mould Tuning time reduction (mainly addressed to mould makers) and about the reduction in time for the Setup of a New Production (of main interest for end users).
A powerful tool that links the job of the mould maker to the end-user is the definition, at the stage of the Mould Tuning, of the Mould Digital Signature which can be provided by the toolmaker and used by the end user each time he run a production.
Finally, the economical evaluation of the Intelligent Mould is addressed to all the partners and is the key point to tell if the approach is successful or not.
WP6: Dissemination, Training, Exploitation and Technology Transfer
The purpose of the work package was to ensure the project was disseminated to as wide an audience as possible by various means including the use of dissemination material developed during the project and also through a number of training sessions. These training sessions were used to pass on the technical knowledge to SMEs throughout Europe in order to ensure that they were fully conversant with the advantages of the productivity enhancements of intelligent moulds.
During the project, it was necessary for the technical centres and mould makers to work together in order to become familiar with the new technologies and techniques developed in the project.
Lastly, plans and procedures were implemented to enable the technologies developed to be effectively exploited.
Project Results:
WP2: Methodology for optimized sensors implementation in tools (RTD)
Knowledge Based System (KBS) for sensor implementation for injection molding has been developed as a hybrid system based on Case-based and Rule-based reasoning.
Both standard phases of knowledge-based systems (knowledge acquisition and knowledge extraction) are implemented and accessible though the web of the project (www.mold4prode.org):
The rule-based reasoning process has been defined following the recommendations that exits in the state-of-the-art. Some of them extracted from this project and others from scientific papers. The case-based reasoning (CBR) has been implemented by modelling the cases as a vector of numerical and categorical attributes over which a distance metrics is defined.
We have implemented and tested several metrics for the CBR. The preliminary validation results (done internally) state that the best metrics are: the Euclidian distance for numerical attributes and the Jacobs’ distance for binary and quantitative attributes. The simple sum of both distances for each comparison case base – new problem is satisfactory.
A user web guide of the web based application has been developed. The guide described the two different web-based applications: to acquire knowledge from experts (only accessible from the website of the project) and to give recommendations to end-users (http://mold4prode.ascamm.com/kbs.php#results):
In parallel, a virtual methodology for Optimal Sensor Determination OSLO) has been developed in 3 steps:
- Process features extraction
- Optimal subset calculation
- Optimal sensor combination calculation
During the period, 4 versions of Algorithms provides by UCBL have been tested to improve the results. The different use cases tested with the methodology are summarized below
Number of sensors Type Good Precision Comments
BOX (PEP) 4 2 Mold Temperature sensor
1 Part température sensor
1 Part Pressure sensor 100% Surface Result are not satisfactory
HCOVER (KIMW- KM) 3 2 Mold Temperature sensor
1 Part Pressure sensor 93,7%
CAP (ASCAMM- Matrix) 2 1 Part Temperature sensor
1 Part Pressure sensor 91%
Tank (PEP- Compose) 3 3 Part Pressure sensor 97.4%
Depending of the case, it is possible to predict the position of 2 to 4 sensors (pressure and/or temperature) to reach 91% to 100% of theoretical prediction of geometrical quality of the parts
By working on more sophisticated feature selection methods and classification algorithm, we managed to build a consistent methodology. The methodology is totally new and continued effort is required to make methodology more robust.
Moreover, as that methodology gives the best location of sensor to recover data and measures, sensor location is included into the mold signature definition.
Last efforts have been focussed on the integration of the Knowledge Based System (KBS)
with the Optimal Sensor Location (OSLO) system to improve the time of calculation.
KBS was used as a starting point to reduce the computational time of numerical simulations by enclosing the surface of the mold to be analyzed by the numerical simulation tool:
The integration of both methods for sensor implementation was analysed by using a real case, the Hcover (Part of the German first run):
The results on good precision reaches 87% with 2 sensors: a part pressure sensor and a mold temperature sensor. Previously, with the largest nodes selection, good precision attained 93,7%. Here, precision is less high because there are less possibilities of location. Nevertheless, results are acceptable with less computation time.
WP3: Algorithms and software for mold assessment
Motivated by significant progresses done in the field of fault detection and diagnosis of batch processes based on the information directly gathered from sensors, it was essential to investigate if these concepts could be applied entirely to calculate the mould signature. In this WP, the mould signature is defined as the ideal configuration of parameters that consistently yield an optimal injection. Here we were especially concerned about obtaining a robust feature set capable of representing the main characteristics that describe major trends in terms of data variability.
The first step was to identify the set of parameters that affect the injection process. We realised that not only parameters involved during the production phase determine the quality of the process. Therefore, the mold signature concept was spread out to all phases involved in the whole injection process since it was detected that a mold signature cannot be defined in a reliable way without guaranteeing that the key information of previous phases are within acceptable ranges. This key information determines the creation of the mold signature.
As a result, a fully optimised manufacturing strategy based on identifying and predicting the best indicators involved during the whole life cycle of the product from the design to the production was described. Since the main causes of defect in injection molding can be because of mould design, process parameters, machine, operator or material, this vision pull together the whole injection moulding phases, integrating existing fragmented ideas of machine, material, process, production and information control with the development of tools to produce a fully optimised manufacturing strategy. In this paradigm, each phase of the whole process has consistent setup/optimisation procedures.
This global methodology provides consistent machine set-up and optimisation procedures through the adaptation of expert knowledge and computer simulation. “Mould Signature Concept” can be defined as a set of information describing moulding conditions (viewed in the mould) required to make good parts. Notions of “mould process set-point” and “mould processing window” have been introduced to specify moulding conditions. The following figure shows the global methodology described:
Special attention was given to the ‘process optimising and monitoring’ phase since the quality of the final part is identified by monitoring both a number of parameters from inside the cavity of the mold and from the machine. In order to evaluate part quality, we extract a set of salient statistical features given a number of injection moulding process parameters and then match these features against a statistical model of the Normal Operating Conditions (NOC) or mould signature. Deviations from the NOC model are clear indications of process fluctuations leading to part defects.
This work was concerned with applying the techniques and modelling methodologies that have shown a good performance for fault detection and diagnosis of batch processes based on the information directly gathered from sensors. In particular, we present a study that aims to create the mould signature using statistical approaches that reflect the structure of data obtained from sensors within a mould and data obtained from the injection machine. Experiments were carried out in real environments.
A number of statistical methods were analysed in order to build a robust classifier capable of determining the quality of the piece in real time which demands to gain an optimal performance in terms of computation time. As trials demonstrated, the main benefit of using these mold signature tools is to reduce the tuning time and setting time by an average of 15%.
A number of process supervision models were implemented, evaluated and improved. An in-depth evaluation was intended to provide detailed, empirical information on the effectiveness and impact of different model parameters on the performance of the classifiers applied during the process optimising and monitoring phase. This work was devoted to complement the mold signature tool; in particular, information acquired both from the cavity of the mold and machine parameters that define the state of the process during the production phase.
Different classifiers trained on real data obtained from trials were tested in several conditions and their performance was studied. Different approaches based on feature extraction optimisation and best feature representation were proposed for improving the robustness of the decision support system to identify the quality of an injected plastic part. Different ways to extract feature vectors were analysed. Training set sizes were analysed in order to decide the minimum training set size to guarantee a good performance.
According to the results reported, it is possible to conclude that the method based on Standard Deviation can be applied successfully as a data reduction method for each type of sensor. Data from different sensors give most information about the state of the whole injection process which led us to improve the performance of the control and monitoring system.
As a summary, the main Scientific and Technology (S&T) results were:
• A fully optimised manufacturing strategy (new concept of mold signature).
• Software tool to help mould makers and converters control the whole injection process from design to production. This tool helps to define the mold signature.
• Improvement of classification and feature selection techniques in data mining to create a mould signature.
• A predictive system capable of detecting the quality of the plastic part based on information obtained inside the mould and machine parameters.
Interesting insights can be gained from the foregrounds described in this section. These can be summarised as follows:
• Supervised classifiers can lead to successful decision support systems for identifying the quality of the final part.
• Closer to a more realistic problem of classification. Most of tests were carried out with real data obtained from different trials. Furthermore, there was no test data overlap with training data (cross-validation technique was used).
• Not limited to two-class condition classification problem. A seven-class condition problem was analysed. Six different types of defects were identified.
• Assumptions made about injection molding process. Some assumptions about the whole injection process were appropriate in the creation of mold signature; especially, those based on identifying key information.
• Support Vector Machine as a potential classifier.
• Classifiers independent of the Data Acquisition Control System (DAC)s.
There may be several possible directions in which the work presented here could be extended or improved. Some of the possibilities include:
• Increasing the complexity of cases. Since the initial objectives of this work concerned with creating the mold signature, it could be interesting to check the performance of the proposed tools with more real cases.
• Reducing the computational time of classifiers by reducing the data dimensionality.
• Providing corrective measurements when a default is detected. This approach would imply to find the correlation between defects and causes. The Case-Based Approach (CBR) could be analysed for this goal.
• Adaptation of the proposed solution to other manufacturing processes such as die cast.
WP4: Sensors integration technology and data acquisition and control system
The main generated results and foreground obtained from this workpackge, are the catalogue of sensing technologies and good implementation practices in the field of injection and molding monitoring, the identification potential improvements in existing solutions and the implementation of the algorithms developed in WP3 in a real processing unit.
The catalogue is an in depth study of the different sensors available for the mold industry, attending to the two principal physical variables to measure, Temperature and Pressure. It was not intended for this catalogue to go beyond the industrial state of the art on applied sensors to the mold in plastic injection processes. This limits the presented catalogue to pressure and temperature sensors.
It has a chapter defining the typology of the sensors as well as the physical principles that explain their way of working.
Due to its inherent interest for plastic injection, there is an in depth study and comparative between direct and indirect ways of measuring pressure of the process. Indirect sensors placed behind the ejectors have the potential to be suitable substitutes for direct sensor because it does not cause any marks on the pieces. A comparative test study is provided in this chapter taking into account these two ways of performing the measurements.
After presenting the typology of the sensors and the way of performing measurements, application notes are given in terms of sensor placing within the mold and wiring. This last aspect becomes fundamental to avoid breakages and malfunctioning problems of the sensors used.
The catalogue of sensors takes into account magnitude to measure, way of measuring, operating conditions, etc. They were considered the main producers of sensors for plastic injection, that the industrial partners stated in the questionary surveyed during the first half of 2011. The criterion was to select the pressure sensors. Temperature technology is much more common and only were selected the temperature sensors offered by vendors of pressure sensors specific for plastic injection process. It also provides guidelines to be able to select a concrete sensor based on concrete needs.
In this catalogue it is included an analysis on infrared sensors, more specifically it will reported a set of tests done on Goizper with a FOS infrared sensor due to its inherent interest that relays on its speed time response. This reason makes it an interesting process monitoring sensor for the plastic injection.
Also wireless data transmission mechanisms are analyzed, focusing on those taking into account data wideband and reliability in industrial working conditions. This technology is very interesting for the monitoring of the mold because reduces the risk of damaging sensory system due to the continuous manipulations that the mold suffers during its life
Also it analyzes the possibilities that ultrasonics has for plastic injection monitoring. After an introductory explanation of this technology, its sensing capabilities are shown.
A deep analysis about the status of research of the academia on the molding process has been also done. Plastic Injection Molding is an active field of research that embraces many different engineering. First, different congresses and scientific journals where these works are published are shown. Its complexity and enormous amount of process parameter manipulation during real time production create a very intense research effort to maintain the process under control. The state of the art techniques applied to the process are documented.
Also a market analysis on DACs is done. Once the sensor converts the measured parameter into an electrical signal, this has to be conditioned (filtered, amplified) and acquired by the acquisition hardware. There are different commercial solutions ranging from specific data acquisition and processing systems specific for molding injection available on the market and several generic purpose acquisition systems. First, it will document the different plastic injection acquisition systems. Then it will document on generic purpose acquisition systems. Finally it has been documented the possibilities that the market offers to condition sensors to obtain their output to be standard voltage or current outputs.
During the project an analysis of improvements for acquisition systems has been done. A first functional analysis was performed based in the horned animal concept. This analysis was made aimed to help discovering new domain of service and product for plastic injection data acquisition systems in general and the PSP in particular.
The used functional analysis is an engineering process to generate concept for a system. It is a heuristic approach to identify main actors in the environment, and show interrelations and leads to the construct of result of guideline. The proposed method follows the following steps:
a. Validation of the system.
b. Identification of environment
c. Identification of functions
d. Characterization of functions
In the project, two DAQ systems has been used to integrate the developments done within the project under WP3
a. PSP system from SISE
b. Compact DAQ from National Instruments
It is documented the possibility of integration of the algorithms developed during the project in these platforms. In the case of PSP it is limited to a type of algorithm even though after the analysis done about potential improvements to DACs equipments it is planned in the future to give wider integration possibilities to this equipment.
In the case of using National Instrument, or another generic purpose Data Acquisition System, it is designed a platform able to integrate and test a wider set of algorithms as those developed under WP3.
Using demonstration as a basis, and working closely with the end users involved in the project, served to record process data in real environments that was lately used for the creation of several machine learning algorithms. These analyzed algorithms may be used as a powerful tools for quality control in fabricated plastic parts..
WP5: Demonstration Case Studies
The main results obtained by the three years experimentation are:
• Reduction of Mould Tuning time. The actual reduction has increased from the first run to the third being approximately coincident to the target. In particular the first run of demo mould closed with an average reduction of Mould Tuning time of 9% (target of 1st year: 10%); at the end of the second year of experimentation it was increased to 29% (target of 2nd year: 30%); and at the end of the third year it was 33% (target of 3rd year: >30%).
• Reduction of New Production Setup time. After the first year of experimentation the reduction of the time spent for the setup of each new production averaged (over the 4 case studies) 43% (to be compared with the target of the 1st year of 7%); at the end of the second year the Consortium obtained an average reduction of the New Production Setup time of 30% (target of 2nd year: 15%); and at the end of the third year it increased again to 37% (target of 3rd year: >30%). For this assessment key point the actual results have always been highly above the expectations showing thaht the system is very powerful and easy to use in reproducing the Digital Signature that is obtain during the Mould Tuning phase.
• A learning curve could be identified showing that the best results can be obtained with the acquisition of specific skills and competence about moulding with sensors. The trend of n the experimentation over the three years is shown in the following charts:
• Digital Signature.
It is a set of curves coming both from moulding machine and sensors signals that condensate all the information related to the moulding of 100% in specs parts. An example of digital signature is given in the following picture:
WP6: Dissemination, Training, Exploitation and Technology Transfer
Five electronic training modules were created on the following subjects:
1. Design phase and sensor implementation
2. Practical implementation of sensors
3. DACs for injection moulding
4. Process validation through sensors
5. Process setting with sensors/DACs
Each of the above included a ‘quiz’ in order to test the user’s understanding of the module.
The modules were based on the knowledge developed throughout the course of the project and refined over the series of Regional Workshops and included key information such as demonstration activities like economic advantages together with technical details The purpose of these training modules was to transfer the results of the methodology and logistic developments of the mould makers and end users of the projects.
They have been uploaded onto the project website for future use.
Potential Impact:
The worldwide market for injection moulded plastics is continuing to grow.
Modern day injection moulding machines are controlled by a built in computer based on sensor fed information. In plastic injection moulding, the primary objective is to manufacture dimensionally and structurally consistent parts, independent of the moulding machine being used. To accomplish this, moulders can benefit from increased process optimization and improved control methods. Turnkey intelligent moulds can offer such an advantage. With the vast sums of money involved; maximising profitability is obviously at the forefront of people’s minds.
Computer modelling and software simulation have become key to meeting these objectives. The first commercial injection moulding simulation software only came to market in 1978. These basic original 2D “layflat” simulations have since moved on to incorporate 3D, shrinkage, warpage, gas-assist, coinjection, multi-shot injection, injection-compression, and thermoset moulding.
Injection moulding simulation software helps manufacturers verify and optimize part and mould designs by providing visual and numerical feedback on injection moulding production. Using optimisation and simulation software helps reduce the need for costly physical prototypes, avoid manufacturing defects and reducing the product’s time to market.
Simulation and optimisation software has become big business in an industry where accuracy and efficiency is king. The global CAE market is expected to post revenue of €2521 million Euro by 2016 (TechNavio 2013). The main players in the market include Mathworks, Ansys, MSC Software, Dassault systems, Siemens PLM, ESI, Autodesk. There is a shift towards global licencing of software which can be used anywhere in the world as vendors push to avoid price differentiation in each region. The major drivers for software market include improved speed to market, reduced analysis time, reduce design cycle, reduced scrap and reduced down time.
Global trends in plastics processing are moving towards lightweight design, increasing functional integration and efficiency of the whole production process. These are posing a significant challenge for injection moulders. Often demands cannot be filled by conventional materials or manufacturing processes. As a result there has been a growth in the market for simulation software which can evaluate different scenarios without incurring the real world costs associated with testing and set up.
With an increasing number of product versions being required by customers the order quantities per product have reduced and the need for a greater number of moulds changes has increased. A direct consequence of this shift in practice has seen the need for reduced turnaround times for tool changes and tool tuning.
While software’s such as Autodesks MOLDFLOW, Siemens CADFLOW, Dassualt SIMPOE Mathworks Matlab and MOLDEX among others have focused on anticipating problems in the process, there has been limited investigation into optimisation of moulds into turnkey solutions to enable faster turnarounds for tuning and tool changes. Predicting the physical behaviour of a system based on so many different parameters is a complex task.
Simulation and optimization algorithms which can meet this requirement have the potential to disrupt the market and bring significant cost savings to organisations. This gap in the market should provide strong growth potential along with the possibility of partnering or licencing the software to key players in the marketplace.
The aim of this research was to provide an insight into the injection moulding market and opportunities for software to facilitate turnkey moulds. The study was based on a SWOT and PEST analysis. This approach allowed identification of significant aspects which can affect a positive outcome to the project being achieved.
Key items of note to date include:
The political scenario always has the potential to affect this industry. Increasing concerns about the chemicals in plastics has led to tightening of environmental regulations; there is also a push towards improving end of life modelling of products and improving fire safety. Stricter regulations can have a big effect on the injection moulding industry increasing costs. This may result in reduced costs being allocated for R&D or updating existing facilities affecting the uptake of new systems.
The fact turnkey intelligent moulds are becoming a growing focus suggest the way the injection moulding business operates is changing. Increased product versions are leading to reduced order sizes with flexibility now a significant factor for success. Products which can tap in to this need for flexibility are going to have a high growth rate potential.
Increasing energy costs are becoming the norm and with injection mouldings having such a high energy dependency any products which can help reduce these energy requirements are going to be giving a warm reception within industry.
The key threat will be through open source software with many commercial enterprises now adopting a mixture of both licenced and open source.
The potential for job creation both direct and indirect will be limited as this is more an add-on software solution. Jobs will in general be limited to specialist within the field.
Quality, price, and time to market are critical factors which will determine whether or not the product is successful. The rewards will be relative to the algorithms developed, and its effectiveness. Protection for the algorithm will need to be a primary focus. It will be important to ensure the name chosen is can be trademarked and consideration will need to be giving as to whether to patent or take the trade secret route.
The prospects which will arise from a successful project outcome are significant. Choosing the correct path to follow to maximise returns on investment will be critical. Consideration should be giving to market exploitation at an early stage to maximise any potential opportunities.
The final results of Mold4ProdE total 5, in different fields, to help the plastic industry increase its competitiveness:
- a Knowledge based system for recommending the selection and location of sensors inside a mould as a WebTool application available on the project website
- a Numerical Optimal Sensor LOcation (OSLO) methodology for polymer injection moulding process
- a User Case Assessment Metrics to evaluate Tuning time reduction, Set up time reduction and the learning curve of operators when embeded sensors are used
- Algorithms for control and monitoring in injection molding process with:
• a tool to provide “mould signature” and collect related information,
• a versatile integration platform (from National Instrument) to test different algorithms
• an optimization of data treatment algorithms
• Improvements of PSP, the Data Acquisition and Control System from SISE
- Training Modules available on the project website
List of Websites:
http://www.mold4prode.org/
The constant demand for diversified/personalized goods leads to decrease time-to-market, and to plastic injection moulds with smaller batch production, with no stock. This generates huge productivity losses in the moulding industry:
- Tool Tuning is too long for tool makers before delivery of molds.
- New production run set-up time is too long and generates too much scraps, due to different “operator and machine” parameters.
Many approaches are currently adopted to reduce mold design and production time, but none of them addresses directly the parts productivity enhancement, where “trial error” is still the most employed engineering approach. The innovative approach of Mold4ProdE was to develop and assess a methodology to tool makers, for them to propose and deliver quicker “turnkey intelligent molds”, to clients with a reasonable extra cost, and offering them significant productivity enhancement.
The main purpose of the project is to:
- optimize the number and position of sensors in the mold via a knowledge approach, and also via the development of a virtual approach OSLO (Optimal Sensor Location) (WP2)
- implement and assess the concept “Mold Signature” by using the data driven models, built from experimental data collected during production runs with sensors embedded in process equipments (WP3).
These 2 points have been achieved by two different ways, the first based on experience via the implementation of a Knowledge Base System, the other via a complete virtual simulation approach. It aims at applying modern data mining techniques for modeling polymer injection molding process, based on virtual data coming from virtual sensors implemented in a CAD model of the mold.
Mold4ProdE has also provided dedicated and optimized components for tool instrumentation, data collection and treatment (WP4). The project delivers the basis to efficiently use “intelligent moulds”.
The project was organized in 3 successive loops. Each loop Integrates work on the theoretical aspects (RTD Work Packages 2, 3 and4) and work on demonstration cases with an update of the related deliverables.
It means that French, Italian, Spanish and German Technical Center have managed 3 different cases during the project, so 12 moulds, with 12 different mould makers, have been instrumented, tested and evaluated during the project (WP5).
The assessments of sensors embedded in moulds have evaluated benefits in mould tuning time and mould new production set up time. The benefits on tuning time can reach 30%, but needs trained people on data collection and treatment. This learning curve has been highlighted during the 3 years of the project. On new production set up time, the benefits are quicker and can reach more than 30%.
16 training sessions have been organized at regional level to Mold4ProdE mould makers and converters, to allow them to integrate the approaches developed in the project, and also the existing knowledge(WP6).
This knowledge is collected in 5 training modules available on the project website for all partners of the project. This will help the European mold makers to move to the knowledge based service industry, identified as a key factor to increase their competitiveness.
Project Context and Objectives:
WP2: Methodology for optimized sensors implementation in tools (RTD)
The main objective of the Mold4ProdE project is to develop and assess methodologies that help tool makers deliver quicker ‘turn key intelligent’ tools to end users. These tools will allow: (1) selecting the most proper type of sensors and the best location inside the mold, and (2) providing optimisation algorithms for each phase of the whole injection molding process. This addresses the mold signature and holds significant benefits for increasing productivity, cost/scrap reductions and improved quality of the resulting part.
WP2 has been carried out in parallel with WP3 in order to provide a global solution for sensors implementation in injection moulds. The common goal of WP2 and WP3 is to converge together to a practical definition of “mould signature” associated to the concept of intelligent turn-key injection mould. To look for a solution to optimal sensor implementation problem, two ways are considered.
- The first way is based on knowledge approach. This approach consists in accumulating knowledge to build a relevant use cases database and to provide tools to exploit it. The objectives of this first period were to built, test and provide a first implementation of the Knowledge tool.
- The second approach consists in exploiting numerical simulation tools (polymer injection moulding simulation software) to define numerically the best sensor configuration. The objective of such an approach is to provide information on sensors location at early stage during the development of the mold.
WP3: Algorithms and software for mold assessment
This WP contributes to the project challenges by:
1. Broadening the conception of mould signature to address current industry needs, and then describes some needed developments to convert that vision to reality. This vision pull together the whole injection moulding story, that is, the big picture, integrating existing fragmented ideas of machine, material, process, production and information control with the development of tools to produce a fully optimised manufacturing strategy. In this paradigm, each phase of the whole process will have consistent setup/optimisation procedures. The main objective is to provide a specific direction for improving product quality during the whole life cycle of the product: from the design to the production. A general scheme of the whole injection moulding process was defined and a software application was created in order to help mold makers and converters to know what information is necessary to gain a feasible mold signature definition.
2. Providing optimised algorithms and mathematical treatments for data processing in order to build a predictive model capable of detecting the quality of the molded plastic part. The predictive model calculates the mathematical distance between the related processed data and the mold signature. The creation of the predictive model needs to gain the following specific objectives:
• Decide what variables characterise the process (data collecting).
• Data dimensionality reduction.
• Selection of the best variables (best predictors).
• Building the predictive model.
• Evaluate the quality of the model by using different indicators.
Different existing data-driven modelling approaches were analysed and adapted for this specific problem. The predictive model was built by using real data collected during production runs with sensors embedded in process equipments.
WP4: Sensors integration technology and data acquisition and control system
The WP4 is devoted to the identification of existing solutions in the field of mold monitoring, either sensors and Data Acquisition and Control System systems, as well as the identification of best practices in deploying those solutions and the implementation of algorithms developed in WP3.
The main objectives are:
1. To create a catalogue of sensing technologies and good implementation practices in the field of injection
molding monitoring
2. To identify potential improvements in existing solutions
3. To implement the algorithms developed in WP3 in a real processing unit
According to these objectives, the WP4 is split into 3 main tasks:
Task 4.1 Sensors and Data Acquisition & Control Systems (DACS) in molding monitoring
Sub task 4.1.1 Survey on existing solutions
Sub task 4.1.2 Practical implementation of sensors and data transmission
Sub task 4.1.3 Data acquisition and processing systems
Task 4.2 Definition and specifications for Molding machine interoperability
Task 4.3 Algorithm integration in selected DACS (data acquisition & Control System)
WP5: Demonstration Case Studies
The WP5 moves from the necessity to evaluate and to test the M4PE approach that is developed, mainly, in the frame of WP2, WP3 and WP4.
Several aspects were developed in those workpackages from the implementation of a standardised methodology for choosing the best sensors configuration (based on knowledge, experience and software simulations) for developing the electronic devices for acquisition of the information and the software for its mining and elaboration.
The intent of the real experimentation is to practice the global approach and assess what are the main outcomes from real cases.
The main interest for the Consortium is acquiring experience and data about Mould Tuning time reduction (mainly addressed to mould makers) and about the reduction in time for the Setup of a New Production (of main interest for end users).
A powerful tool that links the job of the mould maker to the end-user is the definition, at the stage of the Mould Tuning, of the Mould Digital Signature which can be provided by the toolmaker and used by the end user each time he run a production.
Finally, the economical evaluation of the Intelligent Mould is addressed to all the partners and is the key point to tell if the approach is successful or not.
WP6: Dissemination, Training, Exploitation and Technology Transfer
The purpose of the work package was to ensure the project was disseminated to as wide an audience as possible by various means including the use of dissemination material developed during the project and also through a number of training sessions. These training sessions were used to pass on the technical knowledge to SMEs throughout Europe in order to ensure that they were fully conversant with the advantages of the productivity enhancements of intelligent moulds.
During the project, it was necessary for the technical centres and mould makers to work together in order to become familiar with the new technologies and techniques developed in the project.
Lastly, plans and procedures were implemented to enable the technologies developed to be effectively exploited.
Project Results:
WP2: Methodology for optimized sensors implementation in tools (RTD)
Knowledge Based System (KBS) for sensor implementation for injection molding has been developed as a hybrid system based on Case-based and Rule-based reasoning.
Both standard phases of knowledge-based systems (knowledge acquisition and knowledge extraction) are implemented and accessible though the web of the project (www.mold4prode.org):
The rule-based reasoning process has been defined following the recommendations that exits in the state-of-the-art. Some of them extracted from this project and others from scientific papers. The case-based reasoning (CBR) has been implemented by modelling the cases as a vector of numerical and categorical attributes over which a distance metrics is defined.
We have implemented and tested several metrics for the CBR. The preliminary validation results (done internally) state that the best metrics are: the Euclidian distance for numerical attributes and the Jacobs’ distance for binary and quantitative attributes. The simple sum of both distances for each comparison case base – new problem is satisfactory.
A user web guide of the web based application has been developed. The guide described the two different web-based applications: to acquire knowledge from experts (only accessible from the website of the project) and to give recommendations to end-users (http://mold4prode.ascamm.com/kbs.php#results):
In parallel, a virtual methodology for Optimal Sensor Determination OSLO) has been developed in 3 steps:
- Process features extraction
- Optimal subset calculation
- Optimal sensor combination calculation
During the period, 4 versions of Algorithms provides by UCBL have been tested to improve the results. The different use cases tested with the methodology are summarized below
Number of sensors Type Good Precision Comments
BOX (PEP) 4 2 Mold Temperature sensor
1 Part température sensor
1 Part Pressure sensor 100% Surface Result are not satisfactory
HCOVER (KIMW- KM) 3 2 Mold Temperature sensor
1 Part Pressure sensor 93,7%
CAP (ASCAMM- Matrix) 2 1 Part Temperature sensor
1 Part Pressure sensor 91%
Tank (PEP- Compose) 3 3 Part Pressure sensor 97.4%
Depending of the case, it is possible to predict the position of 2 to 4 sensors (pressure and/or temperature) to reach 91% to 100% of theoretical prediction of geometrical quality of the parts
By working on more sophisticated feature selection methods and classification algorithm, we managed to build a consistent methodology. The methodology is totally new and continued effort is required to make methodology more robust.
Moreover, as that methodology gives the best location of sensor to recover data and measures, sensor location is included into the mold signature definition.
Last efforts have been focussed on the integration of the Knowledge Based System (KBS)
with the Optimal Sensor Location (OSLO) system to improve the time of calculation.
KBS was used as a starting point to reduce the computational time of numerical simulations by enclosing the surface of the mold to be analyzed by the numerical simulation tool:
The integration of both methods for sensor implementation was analysed by using a real case, the Hcover (Part of the German first run):
The results on good precision reaches 87% with 2 sensors: a part pressure sensor and a mold temperature sensor. Previously, with the largest nodes selection, good precision attained 93,7%. Here, precision is less high because there are less possibilities of location. Nevertheless, results are acceptable with less computation time.
WP3: Algorithms and software for mold assessment
Motivated by significant progresses done in the field of fault detection and diagnosis of batch processes based on the information directly gathered from sensors, it was essential to investigate if these concepts could be applied entirely to calculate the mould signature. In this WP, the mould signature is defined as the ideal configuration of parameters that consistently yield an optimal injection. Here we were especially concerned about obtaining a robust feature set capable of representing the main characteristics that describe major trends in terms of data variability.
The first step was to identify the set of parameters that affect the injection process. We realised that not only parameters involved during the production phase determine the quality of the process. Therefore, the mold signature concept was spread out to all phases involved in the whole injection process since it was detected that a mold signature cannot be defined in a reliable way without guaranteeing that the key information of previous phases are within acceptable ranges. This key information determines the creation of the mold signature.
As a result, a fully optimised manufacturing strategy based on identifying and predicting the best indicators involved during the whole life cycle of the product from the design to the production was described. Since the main causes of defect in injection molding can be because of mould design, process parameters, machine, operator or material, this vision pull together the whole injection moulding phases, integrating existing fragmented ideas of machine, material, process, production and information control with the development of tools to produce a fully optimised manufacturing strategy. In this paradigm, each phase of the whole process has consistent setup/optimisation procedures.
This global methodology provides consistent machine set-up and optimisation procedures through the adaptation of expert knowledge and computer simulation. “Mould Signature Concept” can be defined as a set of information describing moulding conditions (viewed in the mould) required to make good parts. Notions of “mould process set-point” and “mould processing window” have been introduced to specify moulding conditions. The following figure shows the global methodology described:
Special attention was given to the ‘process optimising and monitoring’ phase since the quality of the final part is identified by monitoring both a number of parameters from inside the cavity of the mold and from the machine. In order to evaluate part quality, we extract a set of salient statistical features given a number of injection moulding process parameters and then match these features against a statistical model of the Normal Operating Conditions (NOC) or mould signature. Deviations from the NOC model are clear indications of process fluctuations leading to part defects.
This work was concerned with applying the techniques and modelling methodologies that have shown a good performance for fault detection and diagnosis of batch processes based on the information directly gathered from sensors. In particular, we present a study that aims to create the mould signature using statistical approaches that reflect the structure of data obtained from sensors within a mould and data obtained from the injection machine. Experiments were carried out in real environments.
A number of statistical methods were analysed in order to build a robust classifier capable of determining the quality of the piece in real time which demands to gain an optimal performance in terms of computation time. As trials demonstrated, the main benefit of using these mold signature tools is to reduce the tuning time and setting time by an average of 15%.
A number of process supervision models were implemented, evaluated and improved. An in-depth evaluation was intended to provide detailed, empirical information on the effectiveness and impact of different model parameters on the performance of the classifiers applied during the process optimising and monitoring phase. This work was devoted to complement the mold signature tool; in particular, information acquired both from the cavity of the mold and machine parameters that define the state of the process during the production phase.
Different classifiers trained on real data obtained from trials were tested in several conditions and their performance was studied. Different approaches based on feature extraction optimisation and best feature representation were proposed for improving the robustness of the decision support system to identify the quality of an injected plastic part. Different ways to extract feature vectors were analysed. Training set sizes were analysed in order to decide the minimum training set size to guarantee a good performance.
According to the results reported, it is possible to conclude that the method based on Standard Deviation can be applied successfully as a data reduction method for each type of sensor. Data from different sensors give most information about the state of the whole injection process which led us to improve the performance of the control and monitoring system.
As a summary, the main Scientific and Technology (S&T) results were:
• A fully optimised manufacturing strategy (new concept of mold signature).
• Software tool to help mould makers and converters control the whole injection process from design to production. This tool helps to define the mold signature.
• Improvement of classification and feature selection techniques in data mining to create a mould signature.
• A predictive system capable of detecting the quality of the plastic part based on information obtained inside the mould and machine parameters.
Interesting insights can be gained from the foregrounds described in this section. These can be summarised as follows:
• Supervised classifiers can lead to successful decision support systems for identifying the quality of the final part.
• Closer to a more realistic problem of classification. Most of tests were carried out with real data obtained from different trials. Furthermore, there was no test data overlap with training data (cross-validation technique was used).
• Not limited to two-class condition classification problem. A seven-class condition problem was analysed. Six different types of defects were identified.
• Assumptions made about injection molding process. Some assumptions about the whole injection process were appropriate in the creation of mold signature; especially, those based on identifying key information.
• Support Vector Machine as a potential classifier.
• Classifiers independent of the Data Acquisition Control System (DAC)s.
There may be several possible directions in which the work presented here could be extended or improved. Some of the possibilities include:
• Increasing the complexity of cases. Since the initial objectives of this work concerned with creating the mold signature, it could be interesting to check the performance of the proposed tools with more real cases.
• Reducing the computational time of classifiers by reducing the data dimensionality.
• Providing corrective measurements when a default is detected. This approach would imply to find the correlation between defects and causes. The Case-Based Approach (CBR) could be analysed for this goal.
• Adaptation of the proposed solution to other manufacturing processes such as die cast.
WP4: Sensors integration technology and data acquisition and control system
The main generated results and foreground obtained from this workpackge, are the catalogue of sensing technologies and good implementation practices in the field of injection and molding monitoring, the identification potential improvements in existing solutions and the implementation of the algorithms developed in WP3 in a real processing unit.
The catalogue is an in depth study of the different sensors available for the mold industry, attending to the two principal physical variables to measure, Temperature and Pressure. It was not intended for this catalogue to go beyond the industrial state of the art on applied sensors to the mold in plastic injection processes. This limits the presented catalogue to pressure and temperature sensors.
It has a chapter defining the typology of the sensors as well as the physical principles that explain their way of working.
Due to its inherent interest for plastic injection, there is an in depth study and comparative between direct and indirect ways of measuring pressure of the process. Indirect sensors placed behind the ejectors have the potential to be suitable substitutes for direct sensor because it does not cause any marks on the pieces. A comparative test study is provided in this chapter taking into account these two ways of performing the measurements.
After presenting the typology of the sensors and the way of performing measurements, application notes are given in terms of sensor placing within the mold and wiring. This last aspect becomes fundamental to avoid breakages and malfunctioning problems of the sensors used.
The catalogue of sensors takes into account magnitude to measure, way of measuring, operating conditions, etc. They were considered the main producers of sensors for plastic injection, that the industrial partners stated in the questionary surveyed during the first half of 2011. The criterion was to select the pressure sensors. Temperature technology is much more common and only were selected the temperature sensors offered by vendors of pressure sensors specific for plastic injection process. It also provides guidelines to be able to select a concrete sensor based on concrete needs.
In this catalogue it is included an analysis on infrared sensors, more specifically it will reported a set of tests done on Goizper with a FOS infrared sensor due to its inherent interest that relays on its speed time response. This reason makes it an interesting process monitoring sensor for the plastic injection.
Also wireless data transmission mechanisms are analyzed, focusing on those taking into account data wideband and reliability in industrial working conditions. This technology is very interesting for the monitoring of the mold because reduces the risk of damaging sensory system due to the continuous manipulations that the mold suffers during its life
Also it analyzes the possibilities that ultrasonics has for plastic injection monitoring. After an introductory explanation of this technology, its sensing capabilities are shown.
A deep analysis about the status of research of the academia on the molding process has been also done. Plastic Injection Molding is an active field of research that embraces many different engineering. First, different congresses and scientific journals where these works are published are shown. Its complexity and enormous amount of process parameter manipulation during real time production create a very intense research effort to maintain the process under control. The state of the art techniques applied to the process are documented.
Also a market analysis on DACs is done. Once the sensor converts the measured parameter into an electrical signal, this has to be conditioned (filtered, amplified) and acquired by the acquisition hardware. There are different commercial solutions ranging from specific data acquisition and processing systems specific for molding injection available on the market and several generic purpose acquisition systems. First, it will document the different plastic injection acquisition systems. Then it will document on generic purpose acquisition systems. Finally it has been documented the possibilities that the market offers to condition sensors to obtain their output to be standard voltage or current outputs.
During the project an analysis of improvements for acquisition systems has been done. A first functional analysis was performed based in the horned animal concept. This analysis was made aimed to help discovering new domain of service and product for plastic injection data acquisition systems in general and the PSP in particular.
The used functional analysis is an engineering process to generate concept for a system. It is a heuristic approach to identify main actors in the environment, and show interrelations and leads to the construct of result of guideline. The proposed method follows the following steps:
a. Validation of the system.
b. Identification of environment
c. Identification of functions
d. Characterization of functions
In the project, two DAQ systems has been used to integrate the developments done within the project under WP3
a. PSP system from SISE
b. Compact DAQ from National Instruments
It is documented the possibility of integration of the algorithms developed during the project in these platforms. In the case of PSP it is limited to a type of algorithm even though after the analysis done about potential improvements to DACs equipments it is planned in the future to give wider integration possibilities to this equipment.
In the case of using National Instrument, or another generic purpose Data Acquisition System, it is designed a platform able to integrate and test a wider set of algorithms as those developed under WP3.
Using demonstration as a basis, and working closely with the end users involved in the project, served to record process data in real environments that was lately used for the creation of several machine learning algorithms. These analyzed algorithms may be used as a powerful tools for quality control in fabricated plastic parts..
WP5: Demonstration Case Studies
The main results obtained by the three years experimentation are:
• Reduction of Mould Tuning time. The actual reduction has increased from the first run to the third being approximately coincident to the target. In particular the first run of demo mould closed with an average reduction of Mould Tuning time of 9% (target of 1st year: 10%); at the end of the second year of experimentation it was increased to 29% (target of 2nd year: 30%); and at the end of the third year it was 33% (target of 3rd year: >30%).
• Reduction of New Production Setup time. After the first year of experimentation the reduction of the time spent for the setup of each new production averaged (over the 4 case studies) 43% (to be compared with the target of the 1st year of 7%); at the end of the second year the Consortium obtained an average reduction of the New Production Setup time of 30% (target of 2nd year: 15%); and at the end of the third year it increased again to 37% (target of 3rd year: >30%). For this assessment key point the actual results have always been highly above the expectations showing thaht the system is very powerful and easy to use in reproducing the Digital Signature that is obtain during the Mould Tuning phase.
• A learning curve could be identified showing that the best results can be obtained with the acquisition of specific skills and competence about moulding with sensors. The trend of n the experimentation over the three years is shown in the following charts:
• Digital Signature.
It is a set of curves coming both from moulding machine and sensors signals that condensate all the information related to the moulding of 100% in specs parts. An example of digital signature is given in the following picture:
WP6: Dissemination, Training, Exploitation and Technology Transfer
Five electronic training modules were created on the following subjects:
1. Design phase and sensor implementation
2. Practical implementation of sensors
3. DACs for injection moulding
4. Process validation through sensors
5. Process setting with sensors/DACs
Each of the above included a ‘quiz’ in order to test the user’s understanding of the module.
The modules were based on the knowledge developed throughout the course of the project and refined over the series of Regional Workshops and included key information such as demonstration activities like economic advantages together with technical details The purpose of these training modules was to transfer the results of the methodology and logistic developments of the mould makers and end users of the projects.
They have been uploaded onto the project website for future use.
Potential Impact:
The worldwide market for injection moulded plastics is continuing to grow.
Modern day injection moulding machines are controlled by a built in computer based on sensor fed information. In plastic injection moulding, the primary objective is to manufacture dimensionally and structurally consistent parts, independent of the moulding machine being used. To accomplish this, moulders can benefit from increased process optimization and improved control methods. Turnkey intelligent moulds can offer such an advantage. With the vast sums of money involved; maximising profitability is obviously at the forefront of people’s minds.
Computer modelling and software simulation have become key to meeting these objectives. The first commercial injection moulding simulation software only came to market in 1978. These basic original 2D “layflat” simulations have since moved on to incorporate 3D, shrinkage, warpage, gas-assist, coinjection, multi-shot injection, injection-compression, and thermoset moulding.
Injection moulding simulation software helps manufacturers verify and optimize part and mould designs by providing visual and numerical feedback on injection moulding production. Using optimisation and simulation software helps reduce the need for costly physical prototypes, avoid manufacturing defects and reducing the product’s time to market.
Simulation and optimisation software has become big business in an industry where accuracy and efficiency is king. The global CAE market is expected to post revenue of €2521 million Euro by 2016 (TechNavio 2013). The main players in the market include Mathworks, Ansys, MSC Software, Dassault systems, Siemens PLM, ESI, Autodesk. There is a shift towards global licencing of software which can be used anywhere in the world as vendors push to avoid price differentiation in each region. The major drivers for software market include improved speed to market, reduced analysis time, reduce design cycle, reduced scrap and reduced down time.
Global trends in plastics processing are moving towards lightweight design, increasing functional integration and efficiency of the whole production process. These are posing a significant challenge for injection moulders. Often demands cannot be filled by conventional materials or manufacturing processes. As a result there has been a growth in the market for simulation software which can evaluate different scenarios without incurring the real world costs associated with testing and set up.
With an increasing number of product versions being required by customers the order quantities per product have reduced and the need for a greater number of moulds changes has increased. A direct consequence of this shift in practice has seen the need for reduced turnaround times for tool changes and tool tuning.
While software’s such as Autodesks MOLDFLOW, Siemens CADFLOW, Dassualt SIMPOE Mathworks Matlab and MOLDEX among others have focused on anticipating problems in the process, there has been limited investigation into optimisation of moulds into turnkey solutions to enable faster turnarounds for tuning and tool changes. Predicting the physical behaviour of a system based on so many different parameters is a complex task.
Simulation and optimization algorithms which can meet this requirement have the potential to disrupt the market and bring significant cost savings to organisations. This gap in the market should provide strong growth potential along with the possibility of partnering or licencing the software to key players in the marketplace.
The aim of this research was to provide an insight into the injection moulding market and opportunities for software to facilitate turnkey moulds. The study was based on a SWOT and PEST analysis. This approach allowed identification of significant aspects which can affect a positive outcome to the project being achieved.
Key items of note to date include:
The political scenario always has the potential to affect this industry. Increasing concerns about the chemicals in plastics has led to tightening of environmental regulations; there is also a push towards improving end of life modelling of products and improving fire safety. Stricter regulations can have a big effect on the injection moulding industry increasing costs. This may result in reduced costs being allocated for R&D or updating existing facilities affecting the uptake of new systems.
The fact turnkey intelligent moulds are becoming a growing focus suggest the way the injection moulding business operates is changing. Increased product versions are leading to reduced order sizes with flexibility now a significant factor for success. Products which can tap in to this need for flexibility are going to have a high growth rate potential.
Increasing energy costs are becoming the norm and with injection mouldings having such a high energy dependency any products which can help reduce these energy requirements are going to be giving a warm reception within industry.
The key threat will be through open source software with many commercial enterprises now adopting a mixture of both licenced and open source.
The potential for job creation both direct and indirect will be limited as this is more an add-on software solution. Jobs will in general be limited to specialist within the field.
Quality, price, and time to market are critical factors which will determine whether or not the product is successful. The rewards will be relative to the algorithms developed, and its effectiveness. Protection for the algorithm will need to be a primary focus. It will be important to ensure the name chosen is can be trademarked and consideration will need to be giving as to whether to patent or take the trade secret route.
The prospects which will arise from a successful project outcome are significant. Choosing the correct path to follow to maximise returns on investment will be critical. Consideration should be giving to market exploitation at an early stage to maximise any potential opportunities.
The final results of Mold4ProdE total 5, in different fields, to help the plastic industry increase its competitiveness:
- a Knowledge based system for recommending the selection and location of sensors inside a mould as a WebTool application available on the project website
- a Numerical Optimal Sensor LOcation (OSLO) methodology for polymer injection moulding process
- a User Case Assessment Metrics to evaluate Tuning time reduction, Set up time reduction and the learning curve of operators when embeded sensors are used
- Algorithms for control and monitoring in injection molding process with:
• a tool to provide “mould signature” and collect related information,
• a versatile integration platform (from National Instrument) to test different algorithms
• an optimization of data treatment algorithms
• Improvements of PSP, the Data Acquisition and Control System from SISE
- Training Modules available on the project website
List of Websites:
http://www.mold4prode.org/
