Machine Wear State Detection from Sensor Time-Series Patterns
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Machine failures and quality deficits in industrial production often occur due to unfavorable operating points and unmonitored wear conditions of machine components, making it difficult to determine dynamic wear conditions and service life, especially in dynamic machining processes.
Innovation Solution
A computer-implemented method that records time series of physical variables, detects change points, extracts pattern sequence instances, and generates classes to determine machine properties like wear without direct measurement, using sensor data from existing sensors and unsupervised machine learning to assign operating states and predict maintenance needs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If direct measurement methods are used to determine machine properties, then measurement precision is improved, but device complexity and cost increase due to additional sensors
Solution Approach 1:
The patent uses readily available sensors (vibration, temperature, acoustic emission sensors) as intermediaries to indirectly determine wear states. Instead of directly measuring wear, the system measures intermediate physical quantities that correlate with wear conditions, thereby avoiding the need for complex direct wear measurement sensors while maintaining determination accuracy
Solution Approach 2:
The patent replaces direct mechanical measurement systems with signal processing and evaluation systems. By substituting physical direct measurement with computational analysis of sensor signals, the system achieves wear state determination without requiring complex mechanical measurement devices
2Device complexity
If statistical calculations are used to determine wear states, then device complexity is reduced, but reliability decreases due to inability to handle changing operating modes
Solution Approach 1:
The patent implements dynamic reference value adjustment that adapts to changing operating modes. The system continuously updates reference values based on current operating conditions, allowing reliable wear state determination even when machining parameters, tools, or workpieces change. This dynamic adaptation maintains reliability without requiring complex recalibration procedures
Solution Approach 2:
The system incorporates feedback mechanisms where determined wear states and operating conditions are used to continuously refine reference values. This feedback loop enables the system to automatically adapt to changing conditions, improving reliability while maintaining relatively simple calculation structures
3Reliability
If monitoring systems are implemented to track machine conditions, then reliability is improved, but loss of time increases due to complex data processing and analysis
Solution Approach 1:
The patent pre-calculates and stores reference values for different operating modes before actual machining operations. By preparing reference data in advance, the system eliminates the need for complex real-time calculations during machining, significantly reducing data processing time while maintaining reliable wear state monitoring
Solution Approach 2:
The system segments the monitoring process into distinct phases: pre-calculated reference value comparison during machining and periodic updates between machining operations. This segmentation allows rapid real-time monitoring using simple comparisons, with complex calculations performed only when necessary, thereby reducing overall processing time
Data Source
Figure 1
Figure 2~3
Figure 4~5
AI summary
The invention relates to a computer-implemented method for determining a property of a machine, in particular a machine tool, without metrologically capturing the property, comprising the following method steps: - capturing one or more first time series of one or more physical measurement variables of the machine; - detecting change points in the one or more first time series; - extracting pattern-sequence instances from the first time series on the basis of the detected change points; - producing a plurality of pattern-sequence classes in accordance with the extracted pattern-sequence instances; - identifying at least one characteristic of a plurality of pattern-sequence instances of the same pattern-sequence class and a time curve of the characteristic; - determining a property of a machine using the determined characteristic and/or using the time curve of the determined characteristic.