Process Fault Detection on Tennessee Eastman Using SPC and Machine Learning

Can statistical process control and machine learning tell when a chemical plant goes wrong, and why? This project monitors the Tennessee Eastman benchmark process (52 measured and manipulated variables, 21 programmed faults) with Shewhart, EWMA and CUSUM charts on every variable, PCA Hotelling T-squared and Q charts, dynamic PCA and an autoencoder, names the variables behind an alarm with contribution plots, and diagnoses the fault with Random Forest and 1-D CNN classifiers. All detectors are compared fairly at the same false-alarm rate. A web application monitors new process data with control charts and alarm reports. R qcc and GNU Octave cross-check the results, and a complete MATLAB / Simulink version is included.

Category: Industrial Engineering

Industrial Engineering Project IE02 from ProTechSC Academy, with graphical abstract, video demonstration and project guidance.