Running Yarn Tension Monitoring With ML Fault Graph Diagnosis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for monitoring thread tension in thread treatment processes are inadequate for achieving uniform quality and quick identification of process disturbances, leading to inefficiencies and increased reject rates.
Innovation Solution
The implementation of a method that utilizes machine learning to analyze error graphs generated from thread tension measurements, allowing for early diagnosis and classification of process disruptions, and enabling automated process adjustments to address identified issues.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional thread tension monitoring with threshold comparison is used, then continuous quality monitoring is achieved, but quick identification and diagnosis of process disturbances is not possible
Solution Approach 1:
The system pre-generates reference error graphs representing known disturbance patterns and stores them in a database. When a disturbance occurs, the system compares the actual error graph against these pre-prepared references, enabling rapid identification without time-consuming analysis. This preliminary preparation of diagnostic tools resolves the contradiction between continuous monitoring and quick diagnosis.
2Ease of operation
If experienced operators manually analyze error graphs to identify process disturbances, then diagnostic capability is achieved, but automation and consistency are reduced
Solution Approach 1:
The system creates digital copies of expert knowledge by storing reference error graphs that represent disturbance patterns identified by experienced operators. These reference graphs serve as automated decision-making tools that replicate expert diagnostic capabilities. The system compares actual measurements against these copied expert patterns, achieving both ease of operation and high automation simultaneously.
3Productivity
If no classification system for error graphs is implemented, then all disturbances are treated equally, but targeted process changes and reject reduction are not achieved
Solution Approach 1:
The system segments the continuous stream of error graphs into distinct categories based on their characteristic patterns. Each category corresponds to specific types of disturbances (e.g., thread breaks, tension variations, equipment faults). This segmentation enables targeted responses for each disturbance type, reducing rejects by addressing root causes specifically rather than applying generic solutions to all errors.
Data Source
Figure 1
Figure 2
Figure 3.1~3.2
AI summary
The invention relates to a method and a device for monitoring a yarn tension of a running yarn in a yarn treatment process. To this end, the yarn tension of the yarn is continuously measured and the measurement signals for the yarn tension are compared with a threshold value of an admissible yarn tension. In the event of an inadmissible tolerance deviation of the measurement signals, a short-term signal path of the yarn tension is detected as a fault graph. In order to enable a fault diagnosis, the fault graph of the yarn tension is analysed using a machine learning program. The fault graph is then allocated to one of the existing fault categories or to a new fault category. The device according to the invention comprises for this purpose a diagnosis unit (18), which cooperates accordingly with the yarn tension evaluation unit (19).