Engineering Data Systems For Intelligent Fault Detection And Predictive Maintenance In Industrial Equipment Monitoring

Abstract

Advanced manufacturing systems produce Internet of Things (IoT)sensor data in huge volumes, which provides intelligent decision support for fault detection and Predictive Maintenance (PdM). However, current techniques mainly depend on labeled data, and it is difficult for them to handle heterogeneous, noisy, and non-stationary industrial data. Additionally, their lack of interpretability and poor generalization ability pose problems in practical applications. Thus, this research introduces an intelligent Decision Support System (DSS) using Elephant Clan Optimized-Convolutional Recurrent Neural Network (ECO-CRNN). The proposed approach is experimented with by applying an industrial sensor dataset, which involves vibration, temperature, and current signals under two classes, namely normal and faulty states of machines. Firstly, Isolation Forest (IF) is used as the data preprocessing method to eliminate outlier data. Feature extraction is executed using the Discrete Wavelet Transform (DWT), which extracts time-frequency features from non-stationary sensor data for efficient fault representation. The findings from experiments reveal that the technique has achieved MAE 0.050, MSE 0.010, and R² 0.960 when compared with the conventional Deep Learning (DL) and Machine Learning (ML) algorithms, and the system uses an implementation model based on Python. In conclusion, this research proposes an effective combination of DL and data engineering techniques for providing intelligent decision-making capabilities for fault detection and PdM in Industrial environments.