Fault detection is fundamental to many industrialapplications. With the development of system complexity, thenumber of sensors is increasing, which makes traditional faultdetection methods lose efficiency. Metric learning is an efficientway to build the relationship between feature vectors withthe categories of instances. In this paper, we firstly proposea metric learning-based fault detection framework in faultdetection. Meanwhile, a novel feature extraction method basedon wavelet transform is used to obtain the feature vectorfrom detection signals. Experiments on Tennessee Eastman (TE)chemical process datasets demonstrate that the proposed methodhas a better performance when comparing with existing methods,for example, principal component analysis (PCA) and fisherdiscriminate analysis (FDA).
展开▼
