Cost-Driven Threshold Selection for Predictive Equipment Failure Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Predictive algorithms for equipment maintenance in industrial processes face challenges in setting threshold values, leading to either excessive service calls for functional equipment or inadequate maintenance for faulty equipment, due to the lack of consideration for economic costs associated with different prediction outcomes.
Innovation Solution
A cost-driven system and method that identifies and selects a threshold value for predictive equipment failure detection algorithms based on the economic costs of successful and unsuccessful predictions, using a Receiver Operating Characteristic (ROC) curve with a break-even line to maximize cost savings, thereby determining when maintenance is needed.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the threshold value is set low to increase detection sensitivity, then more equipment faults are detected, but the number of false alarms increases leading to excessive service calls for functional equipment
Solution Approach 1:
The patent applies parameter changes by adjusting the threshold value based on cost parameters. Instead of using a fixed or arbitrarily set threshold, the system dynamically determines the optimal threshold by incorporating cost data for false positives and false negatives. This allows the threshold to be tuned to achieve the economically optimal balance between detection sensitivity and false alarm rate, resolving the contradiction between reliable fault detection and minimizing false alarms.
2Object-generated harmful factors
If the threshold value is set high to reduce false alarms, then fewer false service calls are generated, but faulty equipment may not be detected leading to inadequate maintenance
Solution Approach 1:
The system dynamically adjusts the threshold parameter based on cost-benefit analysis. By incorporating cost data for different outcomes (false positives, false negatives, true positives, true negatives), the system determines the threshold that maximizes economic benefit. This resolves the contradiction by finding the optimal point where the cost of missed detections is balanced against the cost of false alarms, ensuring adequate maintenance without excessive service calls.
3Ease of manufacture
If traditional threshold selection methods are used without cost consideration, then the algorithm is simple to implement, but economic optimization of maintenance decisions cannot be achieved
Solution Approach 1:
The system performs preliminary action by pre-collecting and storing cost data for different prediction outcomes before running the predictive algorithm. This cost data (including costs of false positives, false negatives, true positives, and true negatives) is prepared in advance and used to determine the optimal threshold. This approach maintains algorithmic simplicity while achieving economic optimization, as the cost-based threshold determination is a straightforward post-processing step that does not complicate the core predictive modeling.
Data Source
AI summary
A method includes identifying costs associated with different outcomes of a failure prediction algorithm. The algorithm is configured to predict one or more faults with at least one piece of industrial equipment. The different outcomes include both successful and unsuccessful predictions by the algorithm. The method also includes identifying a threshold value for the algorithm using the costs, where the threshold value is used by the failure prediction algorithm to identify whether maintenance of the at least one piece of industrial equipment is needed. The method further includes providing the threshold value to the algorithm. The threshold value is selected such that a net positive economic benefit is obtained from use of the threshold value with the failure prediction algorithm. In addition, the method can include generating a signal indicating whether maintenance is needed based on a comparison of an indicator value calculated using the algorithm and the threshold value.