Fault diagnosis of multi-channel data in a forging process using the linear support higher-order tensor machine

摘要

During a forging process, the pressure data collected at different positions of the forging machine constitute a group of multi-channel data. Most existing research cannot use these multi-channel data to detect the variation of product quality due to the correlation between different channels. This paper investigates the complex tensor structure and characteristics of the multi-channel data. A fault diagnosis model of the forging process is then built. In the fault diagnosis model, the multilinear principal component analysis is used to reduce the dimension of the multi-channel data without altering the tensor structure, and the linear support higher-order tensor machine is adopted to construct classifier for fault diagnosis. The hyper-parameter of the model is optimized by using the cuckoo search algorithm. The performance of the proposed diagnostic method is compared with existing methods in both a simulation study and a real-world case study. The results show that the proposed method is more effective in processing multi-channel data from the forging process.