Machine State Classification for Unknown Operation Damage Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine state detection technologies fail to effectively identify unknown operations that may not cause damage and do not provide methods for detecting damage before an anomaly occurs, as they rely on preset operation recognition techniques and position/state recognition methods that cannot handle unforeseen states.
Innovation Solution
A machine state detecting device that collects and processes chronological state information using a classification process to identify feature vectors, detect damage, and tie damage intervals to classification IDs, enabling the detection of unknown operations and damage states.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If preset operation recognition techniques are used, then known operations can be identified, but unknown operations cannot be detected
Solution Approach 1:
The system performs preliminary classification of operational states by dividing state information into temporal segments and extracting feature vectors that represent different operational conditions. These pre-classified states are stored as reference data, enabling the system to recognize both known and unknown operations by comparing current state features against the pre-established classification framework.
Solution Approach 2:
The state information is segmented into multiple temporal divisions, with feature vectors extracted from each segment. This segmentation allows the system to analyze different phases of operation independently and identify unknown operations by detecting patterns that differ from pre-classified states without requiring complete prior knowledge of all possible operations.
2Reliability
If damage detection is performed only after anomaly occurrence, then anomaly causes can be identified, but preventive detection is not possible
Solution Approach 1:
The system performs preliminary classification of operational states and continuously monitors state information against these classifications. By detecting deviations from normal operational patterns before they escalate into anomalies, the system enables preventive damage detection and early warning, allowing maintenance to be scheduled before actual damage occurs.
Solution Approach 2:
The system continuously compares current state information with pre-classified operational states and provides feedback when deviations are detected. This real-time feedback mechanism enables the system to identify emerging damage conditions early and alert operators before anomalies occur, bridging the gap between preventive detection and reliable anomaly identification.
3Ease of operation
If position/state recognition means with preset states are used, then preset states can be identified, but processes for unknown states are not defined
Solution Approach 1:
The system pre-classifies operational states by dividing state information temporally and extracting characteristic feature vectors from each classification. These pre-established classifications serve as a reference framework that enables the system to identify both known states and detect unknown states by recognizing patterns that fall outside the pre-classified categories.
Solution Approach 2:
The system dynamically adapts to unknown states by continuously analyzing temporal patterns in state information and comparing them against the pre-classified framework. When unknown states are detected, the system can dynamically adjust its monitoring and classification approach, enabling flexible handling of unforeseen operational conditions while maintaining ease of identification for known states.
Data Source
AI summary
Provided is a machine state detecting device capable of easily detecting the state of a machine, regardless of whether an unknown operation that is not expected to cause damage is performed. A state detecting device 100 includes a collection device 1a that collects, at a plurality of times, state information if indicating the state of the machine that changes chronologically, and an arithmetic processing device 1b that processes a plurality of pieces of the state information if collected at the plurality of times to detect the state of the machine. The arithmetic processing device 1b includes: a classification process section 1i that classifies a feature amount vector having the type of the state information if as a feature amount into any of a plurality of clusters, and provides a classification ID; a detection process section 1j that detects a damage the machine received, based on state information 1g indicating the state of the machine that changes chronologically; and a tying process section 1l that ties an interval in which the damage was detected by the detection process section 1j to the classification ID provided in the interval by the classification process section 1i.