A Sequential Inspection Procedure for Fault Detection in Multistage Manufacturing Processes.

Area of Science:

• Industrial Engineering
• Manufacturing Systems
• Statistical Process Control

Background:

• Fault diagnosis in multistage manufacturing processes (MMPs) is complex.
• Existing methods often rely on predefined inspection schemes.
• These predefined schemes may not be optimal for identifying variation sources.

Purpose of the Study:

• To propose a novel sequential inspection procedure for MMP fault detection.
• To develop a minimum monitoring system coupled with a sequential testing algorithm.
• To sequentially select inspection features based on expected information gain.

Main Methods:

• Implementing a sequential testing algorithm for process monitoring.
• Utilizing a minimum monitoring system to detect statistical control deviations.
• Sequentially defining inspection features (in-process or end-of-line measurements) based on information gain.
• Comparing the proposed method with predefined and randomized sequential inspection schemes.

Main Results:

• The proposed sequential inspection procedure demonstrates significant benefits.
• Effectiveness is shown in a case study analyzing fault diagnosis.
• Performance is evaluated considering historical fault probabilities.

Conclusions:

• The sequential inspection approach offers an efficient alternative to fixed schemes.
• Optimizing inspection based on information gain enhances fault diagnosis.
• This method improves diagnostic accuracy in complex manufacturing environments.