Genetic Algorithm Turbomachinery Failure Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for predicting turbomachinery failure events are hindered by noise in sensor data and the limitations of standard z-scores when dealing with small datasets, making it difficult to identify anomalies and predict failures accurately.
Innovation Solution
A method using a genetic algorithm to analyze operational data from a machine and peer machines, generating clauses with alleles that include time periods, performance metrics, comparison operators, and threshold values to characterize data and identify anomalies, with a fitness function to select and evolve clauses for predicting past, present, or future events.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If standard z-scores are used to evaluate anomaly scores, then anomaly detection is effective for well-populated groups, but the method loses effectiveness when used on small datasets
Solution Approach 1:
The patent transforms the anomaly detection problem by changing the parameter space from individual z-score evaluations to multi-dimensional pattern spaces. The genetic algorithm evolves clauses that combine multiple performance metrics, time periods, and thresholds, creating composite anomaly patterns that are robust even with limited data. This parameter transformation allows the system to detect anomalies effectively regardless of dataset size.
Solution Approach 2:
The invention adds multiple dimensions to anomaly detection by combining performance metrics across different time periods and machines. Instead of evaluating single data points, the system creates multi-dimensional clauses that consider temporal patterns, peer machine comparisons, and multiple performance attributes simultaneously. This dimensional expansion enables reliable anomaly detection even with small datasets by leveraging patterns across multiple dimensions.
2Measurement precision
If many different corrections and controlling factors are applied to remove noise from sensor data, then data quality improves, but the complexity of monitoring and diagnosing anomalies increases significantly
Solution Approach 1:
The patent merges multiple corrections and controlling factors into unified anomaly patterns through genetic algorithm evolution. Instead of applying numerous separate corrections to sensor data, the system evolves clauses that inherently account for noise, peer machine variations, and temporal patterns in a single integrated framework. This consolidation maintains data quality while dramatically reducing monitoring complexity.
Solution Approach 2:
The system employs self-service through automatic pattern evolution and selection. The genetic algorithm automatically evolves and selects the most effective anomaly patterns from the data without requiring manual configuration of correction factors or controlling parameters. This automation reduces the complexity burden on operators while maintaining high data quality through adaptive, data-driven pattern discovery.
3Ease of operation
If subjective assessment is used to determine whether a tag is anomalously high or low, then the process is simple, but the reliability and objectivity of anomaly identification decreases
Solution Approach 1:
The patent implements feedback through the fitness function that evaluates evolved clauses against known failure and non-failure cases. The genetic algorithm receives feedback on pattern performance and iteratively improves anomaly detection reliability by selecting clauses with higher fitness scores. This automated feedback loop maintains simplicity while dramatically improving objectivity and reliability compared to subjective assessment.
Solution Approach 2:
The invention replaces the mechanical process of subjective human assessment with an automated computational system. The genetic algorithm and fitness evaluation mechanism substitute for human judgment, providing objective, consistent anomaly identification. This substitution maintains ease of operation through automation while eliminating the subjectivity inherent in manual assessment.
Data Source
AI summary
A method for predicting or detecting an event in turbomachinery includes the steps of obtaining operational data from at least one machine and at least one peer machine. The operational data comprises a plurality of performance metrics. A genetic algorithm (GA) analyzes the operational data, and generates a plurality of clauses, which are used to characterize the operational data. The clauses are evaluated as being either “true” or “false”. A fitness function identifies a fitness value for each of the clauses. A perturbation is applied to selected clauses to create additional clauses, which are then added to the clauses group. The steps of applying a fitness function, selecting a plurality of clauses, and applying a perturbation can be repeated until a predetermined fitness value is reached. The selected clauses are then applied to the operational data from the machine to detect or predict a past, present or future event.