An Anomaly Detection and Diagnosis Method Based on Real-Time Health Monitoring for Progressive Stamping Processes

Abstract

Progressive stamping processes have been applied to fabricate an extended range of products from centimeter-scale parts to meter-scale parts. The quality of stamped products may vary and be out of specification due to various anomalies during manufacturing process. Therefore, an effective online health monitoring and fault diagnosis technique is of great practical significance. This paper develops a two-stage systematic approach to enhance the fault detection and fault identification capability for the progressive stamping process with aggregated system-level tonnage signals. The first stage uses a combined Haar transform and power spectrum analysis to map features extracted from aggregated signals to individual operations. The second stage develops a two-step control chart strategy for anomaly detection and identification. The proposed method can improve the monitoring effectiveness and the quality assessment of individual operations based on an aggregated tonnage signal especially when single working range of different operations in the multi-station system are highly overlapped. The results show the method efficacy of quick and accurate anomaly detection and identification in real time.

A Data-Based Tool Failure Prevention Approach in Progressive Die Stamping

“…Several examples of algorithms for anomaly detection can be found. As an example, Shiu et al [13] developed a twostage approach for progressive stamping using tonnage signals. The first stage uses a combined Haar transform and power spectrum analysis to map features extracted from aggregated signals to individual operations.…”

A Comprehensive Survey of Sensor Deployment Strategies to Diagnose Discrete-Part Manufacturing System

“…Firstly, Key components, which are critical to system's reliability, safety and repair cost, etc., are identified based on the actual maintenance records; Secondly, sensors are selected and located to pick up signal signatures to faults, and finally, transitional data read by sensors is processed to identify the fault root causes. Therefore, fault diagnosis is widely used in manufacturing industry, such as Tool condition monitoring(TCM) [6,7], rotary machines [8,9], assembly processes [10,11], and stamping process [12], etc. Sensors and sensing technologies constitute the fundamental basis for fault diagnosis.…”