Steinbeis software STASA QC enhances the efficiency of injection molding processes
STASA Steinbeis Applied Systems Analysis has more than 20 years of in-depth experience in the development and programming of AI-based systems, as well as systems analysis, i.e. the handling of data, its analysis, modeling, simulation, and its application to commercial AI software. For example, the AI software STASA QC developed by the Steinbeis experts in Stuttgart is being used to optimize injection molding processes.
The plastics industry is an important sector of the economy. As a manufacturing process, injection molding has more appeal than ever and it’s predestined to produce innovative components and assemblies both cost-effectively and in high quality, i.e. with ultimate precision. Questions regarding the best ways to use energy efficiently, improve material efficiency, reduce wastage, and minimize costs within process setups and production – and to reduce industrial pollution – are therefore of existential importance to the mostly medium-sized injection molding companies in the world.
Support is available in the form of artificial intelligence. AI processes have long been integral to the production of a whole host of SME products. They improve the market standing of companies by making use of groundbreaking smart technologies. This is because smart systems should be easy to use, they support users in targeted areas, and they provide solutions that go beyond simple adaptation processes.
The injection molding process is complex, making it difficult to develop deterministic simulation models for use as digital twins due to the intricate nature of the processes. This is precisely why AI methods lend themselves to modeling such complex interactions. On the one hand, injection molding machines are part of the same unit as production molds. On the other, both use sensors to generate data independently of one another and this can be evaluated and analyzed with the help of AI.
Human-AI interaction
An important factor with injection molding processes is the operator involved in the setup, the startup, and production. Building on their experience, operators work out the best possible process setup, although given that this involves an injection molding machine with a minimum of four to six variable machine parameters, it is basically impossible (from a human perspective) to match the scope of potential settings with a sufficient number of data points. Setpoints are determined by the experience of the machine technician; there is no objective criterion for making the best choice in trial-and-error processes, so it remains uncertain whether a selected machine setting actually is the best choice.
So, for example, what if an operator is entering information to determine the optimal setpoint and receives a suggestion from an AI recommender system, offering specific guidance and quickly pointing them in the right direction – just by evaluating attributive quality features, such as burring or sink marks? What if a couple of more settings, selected by using AI, were enough to create a reliably functioning model of the injection molding machine – a digital twin – with all the analysis and simulation capabilities offered by AI-based simulation software?
STASA QC: automated optimization
These were the questions investigated by the STASA experts using their software, STASA QC. The aim of the award-winning software is to save time and money – and improve efficiency. STASA QC helps efficiently determine optimal process settings, and it is capable of creating digital twins by first drawing on a small number of training settings for individual process parameters and then evaluating components. It does this by means of an AI-based process model that matches individual setting variables on the injection molding machine to quality attributes such as dimensions, weight, flash formation, sink marks, and cycle time.
The AI models in STASA QC are reliable enough for different settings to be entered to simulate their impact on the production process. This also results in the automatic determination of the best operating point to achieve the desired quality and defined optimization targets.
Quality assurance using STASA QC and ComoNeo
The quality of injection molded parts is largely determined by which production tools are used. By not just capturing quality characteristics when training the individual process parameters, but also individual pressure curves, it becomes possible to predict the quality of molded parts through the production sequence. Based on this ongoing analysis of processes, component dimensions are predicted.
Thanks to a self-generating neural network in ComoNeo, a process monitoring system made by Kistler Instruments in Winterthur, and the AI systems provided by STASA QC, faults, process fluctuations, and long-term process changes with an impact on the quality of molded parts are detected immediately. This makes it possible to assess process quality online, allowing a decision to be made for each individual cycle regarding whether a molded part that has just been produced meets all quality requirements.
Using AI to capture measurements and evaluate cavity pressures thus makes it possible to achieve zero-defect production. Combining STASA QC with ComoNeo results in a significant reduction of manufacturing tolerances from ISO grade of tolerance IT9 to – under optimal conditions – as low as IT7, resulting in the highest production volumes with maximum precision.
Symbolic diagram of AI areas surrounding an injection molding machine (Source: STASA Steinbeis Applied Systems Analysis GmbH)
The research continues
Injection molding machines contain a large number of different sensors, which are used to monitor and control internal processes. In addition to machine-related process fluctuations, there are also material fluctuations, wear, humidity, temperature fluctuations, and other disturbance variables. The large number of mechanically and thermally coupled processes and the different disturbance variables mean that every cycle and every filling process of a molded part is different.
This leads to fluctuations in melt quality and ultimately to process variations – a typical case for applying AI methods, and the central topic of an ongoing research project called KIassistsKMU[1], which Steinbeis expert Professor Dr. Günter Haag and his team are currently working on. The aim of the project is to develop an assistance system for use in the plastics industry. On the one hand, this will capture the expertise of operators and, on the other, it will use AI to tap into previously unknown relationships and develop appropriate problem-solving strategies. With the help of the assistance system, the experience and knowledge of operators will be captured through complex interactions within the plastic engineering process and these will be used to train AI-based neural networks. Recommendation algorithms will then ensure that recommended actions are provided for help with troubleshooting during live processes.
The data collected online, in both the injection molding machine and the mold, can also be used proactively for predictive maintenance. This should help maintain quality standards and reduce downtime and associated costs. In the Steinbeis team’s experience, introducing AI to processes can reduce production costs by 10 to 15% on average compared to conventional processes. “The time needed to make optimizations with the AI methods of STASA QC could even be reduced to one third or one fifth compared to conventional optimizations,” concludes Günter Haag.
Contact
Prof. Dr. habil. Günter Haag (author)
Managing Director
STASA Steinbeis Angewandte Systemanalyse GmbH (Stuttgart)
www.stasa.de