IRS has developed a new algorithm for the detection of defects in production lines together with the University of Verona, which will be published in "Engineering Applications of Artificial Intelligence". To view and download the full article, just click: https: //authors.elsevier.com/c/1eciH3OWJ90~i8
IRS is a company that develops test measurement systems using innovative technologies. As part of a research project called Premani (Digital Manufacturing Solutions) IRS has developed Artificial Intelligence algorithms for the detection of defects in production lines.
In the paper, an unsupervised algorithm based on Machine Learning and neural networks is proposed for defect detection in a production line. This method is applied to real data by means of a test station implemented on the line. Both thermal images and current and power measurements from refrigerators are collected. The dataset considered is highly unbalanced with the vast majority of healthy samples. The thermal images are processed using a Deep Convolutional Neural Network. Features extracted from the thermal images are then merged with structured power, current and temperature data.
Next, an algorithm called Deep Auto-Encoder is trained on the dataset to report anomalies corresponding to faults in refrigerators. Three different methods are trained and compared: (1) an automatic method in which an expert extracts relevant features from the thermal images without using the image recognition module; (2) a semi-automatic method in which the convolutional neural network is applied to regions of interest within the thermal images selected by an expert operator; (3) a fully automatic method in which the deep convolutional network processes the entire thermal image without any human intervention. The three methods show comparable results with, however, slight differences.