A generative neural network model for the quality prediction of work in progress products

Abstract

One of the key challenges in manufacturing processes is improving the accuracy of quality monitoring and prediction. This paper proposes a generative neural network model for automatically predicting work-in-progress product quality. Our approach combines an unsupervised feature-extraction step with a supervised learning method. An autoencoding neural network is trained using raw manufacturing process data to extract rich information from production line recordings. Then, the extracted features are reformed as time-series and are fed into a multi-layer perceptron for predicting product quality. Finally, the outputs are decoded into a forecast quality measure. We evaluate the performance of the generative model on a case study from a powder metallurgy process. Our experimental results suggest that our method can precisely capture the defective products.

Artificial Intelligence in Advanced Manufacturing: Current Status and Future Outlook

“…More importantly, SPC is dependent on clear, known relationships between inspected variables and final product quality. With the vast amounts of time-series data, some ML-based approaches [30][31][32][33][34][35] are proposed to effectively overcome some of the deficiencies in conventional process control methods. In Ref.…”

“…In Ref. [34], a case study is presented on the quality control of work-in-progress products in the powder metallurgy process based on autoencoders and recurrent neural networks (RNNs). In Ref.…”

Quality Prediction and Yield Improvement in Process Manufacturing Based on Data Analytics

“…Ramana predicted quality in the plastic injection molding process as a continuous-flow process by finding data patterns and abnormal symptoms with the use of the data mining technique [5]. With the use of an artificial neural network (ANN) method, Wang predicted the quality of the product in the powder metallurgy compression process which is not a continuous-flow process [6]. Ogorodnyk conducted ANN and decision tree-based classification research in order to predict parts quality in the thermoplastic resin injection molding process [7].…”

Learning With Imbalanced Data in Smart Manufacturing: A Comparative Analysis

“…These methods are often referred to as data augmentation tools as they increase the number of samples by adding the synthetic data. For instance, authors in [10] use such a method (Synthetic Minority Oversampling TEchnique (SMOTE) [11], [12]) to mitigate data imbalance while predicting product quality.…”