Combining dynamic neural networks and image sequences in a dynamic model for complex industrial production processes

Abstract

This paper describes how to build a quality prediction model for complex industrial production processes using dynamic neural networks. It is known that it is difficult to analyze the mechanisms of many complex industrial production processes and build models by employing classical methods. In this study, based on the image sequences obtained by using the computer-vision-detection-system, the features of image sequences are extracted, and a dynamic neural network model is built to predict and judge the product qualities. The neural network with recurrent architecture consists of two blocks and is controlled by a switch function. The performance evaluation shows that the proposed method achieves a prediction rate of 87.5% accurate, and provides evidence that the method is feasible, effective and promising in its future applications.