Modern manufacturing is moving towards automation, reducing or even eliminating human intervention in control mechanisms. Both for quality control and measuring the system state sensors play a vital role. Environmental factors along with aging can introduce drift faults into the sensor readings. The uncompensated drift faults can result in a wrong control system response and the system may become unstable incurring a huge loss of resources. Linear drift is well documented in the literature but nonlinear drift has not yet been properly investigated and literature is also scarce. Keeping these in perspective we propose a wavelet scattering and Artificial Neural Network(ANN) based framework for the detection and isolation of both linear and non-linear drift faults in temperature sensors. The linear and nonlinear sensor drift faults are simulated by injecting the faults into the normal temperature sensor signals. For the purpose of validating the proposed framework, the IntelLab dataset is used. MATLAB software is used for the purpose of simulating drift faults, coding, and training the models. The performance of the proposed model is compared with other ANN and SVM models. The proposed model achieved an overall classification accuracy of 96.7% in classifying the linear and nonlinear drift faults. The proposed model is found to be more efficient in classifying linear drift faults compared to nonlinear drift faults in terms of the number of misclassifications from the nondrifted signals. In critical operations, automatic diagnosis and compensation of the drift faults of sensors need to be implemented as they can avert major accidents. The proposed model can be implemented in real-time data-driven applications for drift fault diagnosis

Wavelet Scattering, Artificial Neural Network(ANN), Sensor Fault, Drift Fault