G. Ø. Rønsch

Predictive and condition-based maintenance is given more and more attention to further optimize the utilization of manufacturing and production equipment. Utilizing acoustic signals for equipment monitoring and condition-based maintenance has been proven effective in many applications. Many manufacturing and production setups consist of multiple alike machines (e.g. within injection moulding) where it would be beneficial to use the same monitoring setup and configuration on all machines. Based on an industrial application within injection moulding (using five different injection moulds), a methodology is proposed utilizing acoustic signals from injection moulds combined with generative Gaussian or autoencoder modeling. To improve the generalization ability of generative modeling to moulds not seen at training time, a simple yet effective model adaptation is proposed, which only requires a few faultless moulding cycles at runtime/test time. The best results are obtained using the Gaussian model, where area under the curve values close to one are achieved when employing a model adapted to the specific mould at test time to detect abnormal situations like mechanical-defective moulds (loose latch lock) and the need for lubrication. The proposed framework is light in terms of computation and makes the setup implementation practically feasible in a real industrial context with multiple similar machines.

Various production/process equipment’s have built-in sensors allowing for continuous collection of process data. However to ease the data processing burden, it is often the case that only certain features such as aggregated measures or peak values are stored. Yet also in some cases such sensor signals can be extracted fully reflecting the dynamics of the process and utilized for process optimization. The aim of this paper is to demonstrate that such data can be utilized in an effective way to optimize an injection molding process using signals from built-in pressure and position sensors. The observational process data is combined with data from controlled experiments to observe the causal relationships between disturbance factors, process settings and the final quality of the products. We demonstrate that signals from built-in injection molding machine sensors can be used for detecting and mitigating quality issues caused by variation in raw material due to the dual sourcing from two suppliers, which cannot actually be identified during production. For this, we show that the origin of raw material can be classified using the time series profiles of dosing pressure and PLS-DA (Partial Least Squares Discriminant Analysis). Through experimental work, we conclude that this classification can be used for increasing the operating window for holding pressure and mold temperature, which ensure production of products within specifications.

This work focuses on the effective utilisation of varying data sources in injection moulding for process improvement through a close collaboration with an industrial partner. The aim is to improve productivity in an injection moulding process consisting of more than 100 injection moulding machines. It has been identified that predicting quality through Machine Process Data is the key to increase productivity by reducing scrap. The scope of this work is to investigate whether a sufficient prediction accuracy (less than 10% of the specification spread) can be achieved by using readily available Machine Process Data or additional sensor signals obtained at a higher cost are needed. The latter comprises Machine Profile and Cavity Profile Data. One of the conclusions is that the available Machine Process Data does not capture the variation in the raw material that impacts element quality and therefore fails to meet the required prediction accuracy. Utilising Machine Profiles or Cavity Profiles have shown similar results in reducing the prediction error. Since the cost of implementing cavity sensors in the entire production is higher than utilising the Machine Profiles, further exploration around improving the utilisation of Machine Profile Data in a setting where process variation and labelled data are limited is proposed.