Rotor Diagnosis Using Adaptive Parameter Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The challenge in predictive rotor diagnosis is selecting relevant parameters for accurate diagnosis without relying on individual knowledge or skills, as the precision of diagnosis greatly depends on the selection of sensor values and their processing, which can vary with the rotor's environment and age.
Innovation Solution
A rotor diagnostic apparatus that calculates evaluation values to determine parameter contribution and selects parameters for abnormality diagnosis based on these values, using a genetic algorithm to identify suitable parameters in a multidimensional space, allowing for objective selection independent of technician knowledge.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If parameters are selected relying on the knowledge of technicians or using typical textbook parameters uniformly, then the parameter selection process is simple and fast, but the diagnostic precision decreases because the selected parameters may not actually contribute to the diagnosis results
Solution Approach 1:
The system performs self-diagnosis by automatically evaluating which parameters contribute to diagnostic results using genetic algorithms and evaluation functions, eliminating the need for external expert knowledge while improving diagnostic precision through objective, data-driven parameter selection
Solution Approach 2:
The system pre-evaluates parameter contributions by calculating evaluation values for each parameter before actual diagnosis, using genetic algorithms to identify which parameters will be most useful for detecting rotor abnormalities, thereby preparing the optimal parameter set in advance
2Measurement precision
If more parameters are used for diagnosis to improve detection accuracy, then the diagnostic precision improves, but the device complexity and computational load increase
Solution Approach 1:
The system extracts only the essential parameters that contribute to abnormality detection by evaluating each parameter's contribution value and selecting only those above a threshold, removing redundant parameters to maintain simplicity while preserving detection accuracy
Solution Approach 2:
The system dynamically changes the set of parameters used for diagnosis based on the specific rotor and operating conditions, selecting different parameter combinations for different rotors rather than using a fixed large set, thereby adapting to specific needs without unnecessary complexity
3Adaptability or versatility
If typical textbook parameters are selected uniformly for all rotors, then the parameter selection process is simple and consistent, but the adaptability to specific rotor characteristics decreases
Solution Approach 1:
The system applies local quality by selecting parameters specifically tailored to each rotor's characteristics rather than using uniform parameters for all rotors, evaluating which parameters are most relevant for each specific rotor type and operating condition to maximize adaptability
Solution Approach 2:
The system makes the parameter selection dynamic and adaptive, allowing the parameter set to change based on the specific rotor being diagnosed, using genetic algorithms to optimize parameters for each individual case rather than maintaining a static uniform parameter set
Data Source
AI summary
A rotor diagnostic apparatus includes a parameter selection part that calculates evaluation values indicating how much sensor values of a rotor at a given time point deviate from a reference value at which the rotor is known to be in a normal state, the given time point being after the rotor is known to be in the normal state and before the rotor is to be diagnosed, and selects a plurality of parameters to be used for abnormality diagnosis of the rotor from the sensor values based on the calculated evaluation values. A diagnosis part outputs an abnormality value which is indicative of how abnormal the rotor is and which is obtained based on a distance, in a multidimensional space having the selected parameters as coordinate axes, between a figure represented by sensor values in the normal state and a figure represented by sensor values targeted for diagnosis.