Manufacturing Abnormality Detection Using Multi-Method ML Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing abnormality determination methods for manufacturing facilities face challenges in accurately distinguishing between normal and abnormal states due to the reliance on threshold values and limited analysis methods, leading to inconsistent results and the need for extensive data collection for abnormal cases.
Innovation Solution
An abnormality determination support apparatus that employs a combination of primary and secondary determination units, utilizing machine learning with multiple analysis methods to derive numerical indices from time-series signals, and a machine learning device trained with teacher signals to provide accurate and consistent abnormality determination independent of threshold values.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple analysis methods are used to determine abnormality, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The patent combines multiple analysis methods (statistical analysis, FFT analysis, wavelet analysis, etc.) into a unified abnormality determination system. The determination unit integrates results from various analysis methods to comprehensively assess abnormality, achieving higher measurement precision while managing system complexity through structured integration.
Solution Approach 2:
The determination unit is designed with multi-functionality to handle multiple types of analysis methods simultaneously. It can perform statistical analysis, frequency analysis, time-frequency analysis, and other operations through a single unified component, reducing the need for separate dedicated systems for each analysis type.
2Ease of operation
If threshold values are used for abnormality determination, then ease of operation is improved, but measurement precision deteriorates
Solution Approach 1:
The patent transitions from static threshold-based determination to dynamic determination using machine learning models. The determination unit learns optimal decision boundaries from training data, enabling adaptive abnormality detection that maintains operational simplicity while significantly improving measurement precision through data-driven decision making.
Solution Approach 2:
The system changes the determination parameter from fixed threshold values to learned decision boundaries derived from training data. By transforming the determination criterion from a simple threshold comparison to a learned model output, the system achieves both ease of operation (automatic determination) and high measurement precision (accurate abnormality detection).
3Reliability
If more abnormal data is collected for training, then reliability is improved, but loss of time increases
Solution Approach 1:
The patent performs preliminary action by collecting and storing normal operation data in advance during regular manufacturing operations. This pre-collected data serves as the foundation for training the machine learning model, reducing the need for extensive abnormal data collection later and minimizing time loss while maintaining model reliability.
Solution Approach 2:
The system creates synthetic abnormal data by introducing artificial faults or anomalies into collected normal operation data. This copying approach generates additional training samples without requiring actual abnormal events, thereby improving model reliability while avoiding the time-consuming process of waiting for real abnormal occurrences.
Data Source
AI summary
An abnormality determination support apparatus includes an analysis object data preparation unit, a primary determination unit, and a secondary determination unit. The analysis object data preparation unit acquires a time-series signal representing at least one of a state of the manufacturing facility and a product quality from a data collection apparatus of the manufacturing facility, and extracts analysis object data from the time-series signal. The primary determination unit derives a plurality of primary determination results from common analysis object data by applying a plurality of different types of analysis methods to the analysis object data extracted by the analysis object data preparation unit. The secondary determination unit includes a machine learning device.