Deep Convolutional Neural Network with Deconvolution and a Deep Autoencoder for Fault Detection and Diagnosis

Area of Science:

• Chemical Engineering
• Artificial Intelligence
• Process Control

Background:

• Accurate fault detection in industrial facilities is critical for preventing accidents.
• Existing methods may not fully capture complex process dynamics and intervariable nonlinearities.

Purpose of the Study:

• To develop an advanced deep learning model for rapid and accurate root cause analysis of process faults.
• To enhance the performance of fault detection and diagnosis in industrial settings.

Main Methods:

• A novel deep convolutional neural network with deconvolution and a deep autoencoder (DDD) was developed.
• The DDD model assesses process dynamics and intervariable nonlinearity.
• Fault detection is performed using reconstruction error, and root cause analysis employs gradient-weighted class activation mapping.

Main Results:

• The DDD model demonstrated effectiveness in fault detection and diagnosis on the Tennessee Eastman process dataset.
• The proposed method achieved improved performance over conventional fault detection and diagnosis techniques.

Conclusions:

• The DDD model offers a robust solution for real-time fault detection and root cause diagnosis in industrial processes.
• Deep learning approaches, like DDD, show significant potential for improving industrial safety and operational efficiency.