Anomaly Detection for Screw Tightening Timing Data with LSTM Recurrent Neural Network
Xiaopeng Cao, Jun Liu, Fanku Meng, Bo Yan, Hong Zheng, Hong-Yi Su · 2019
Screws are important fasteners in industrial production and are also weak points in various mechanical devices. Therefore, screw tightening detection is of great significance. The traditional tightening quality detection method mainly collects the parameters of the tightening process and draws the rotation angle-torque curves, and draws conclusions through manual analysis, which is time-consuming and labor-intensive and low in efficiency. In order to solve this problem, this paper proposes a model based on LSTM that can automatically analyze the quality of the tightening curve, which improves the timeliness and accuracy of the test. The work of this paper is mainly divided into the following two phases. First, the original data is preprocessed using a feature extraction algorithm based on traditional sampling. In the second phase, we used a classification sample as training data to train a neural network based classifier. In the experiment, we compared the model with traditional machine learning methods, such as SVM, Random Forest. The result is better than traditional machine learning methods.