Volume 09,Issue 04

Enhanced Recognition of Manufacturing Process Anomalies: A Tri-Level Approach Using Shape and Statistical Features with an Optimized Fuzzy Logic Classifier

Authors

Milad Khormali, Jonathan Chen

Abstract
The imperative role of fault detection in manufacturing processes cannot be overstated, as it is essential for ensuring the utmost quality, efficiency, and safety standards. This study introduces a sophisticated anomaly detection method for manufacturing processes, capable of recognizing nine distinct control chart patterns (CCPs). This technique is founded on the intelligent integration of shape descriptors and statistical indicators, further enhanced by an optimized fuzzy classification system. The methodology unfolds across three stratified stages, where each tier employs a meticulously chosen array of shape and statistical features that feed into the classifier to identify subsets of patterns. The adaptive neuro-fuzzy inference system (ANFIS), known for its prowess in pattern recognition challenges, serves as the classifier within each layer, honed by the Harris Hawks Optimization (HHO) algorithm. This research’s core contributions are the strategic extraction of novel features, the augmentation of ANFIS’s robustness, and the comprehensive inclusion of nine CCPs in the detection framework. Empirical simulations underscore the superior performance of the proposed approach, achieving a remarkable 99.6% accuracy in pattern classification, thus outstripping comparable methodologies in efficacy. The industrial applicability of this system is its capacity to adapt to diverse manufacturing settings, significantly reducing the time and resources typically required for fault detection.

Keyword: Anomaly detection, ANFIS, HHO, Manufacturing process, Quality signatures.

PDF [ 779.31 Kb ] | Endnote File | XML