Learning Model Maintenance Data Weighting for Part Identification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems face challenges in accurately identifying the necessary parts for maintenance, as they often amplify noise by treating incorrect parts as correct answer data during re-learning, leading to difficulties in specifying the correct parts among multiple replacements.
Innovation Solution
A system that includes processors to acquire and re-train a learning model based on new maintenance information, weighting the maintenance data to suppress the amplification of incorrect answer noise by assigning lower weights to frequently presented or incorrectly identified parts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If all executed maintenance information is uniformly re-learned as correct answer data, then the learning model is continuously updated with new data, but incorrect maintenance information is amplified as noise leading to reduced accuracy
Solution Approach 1:
The patent applies local quality by differentiating the treatment of maintenance information based on its correctness. Instead of uniformly processing all maintenance data, the system identifies and weights correct answer data differently from incorrect data. This is achieved by comparing presented parts with actually replaced parts, identifying correct answers, and assigning appropriate weights to prevent noise amplification while maintaining continuous learning capabilities.
Solution Approach 2:
The patent changes the parameter of data weighting in the learning process. By introducing weight values that differentiate between correct and incorrect maintenance information, the system modifies how data is processed during re-learning. Correct answer data receives appropriate weighting to prevent noise amplification, while maintaining the continuous update capability of the learning model through parameter adjustment rather than uniform data treatment.
2Ease of operation
If multiple parts are replaced during maintenance, then comprehensive maintenance is performed, but it becomes difficult to specify which parts are really necessary among the replaced parts
Solution Approach 1:
The patent extracts the essential information from multiple replaced parts by identifying which parts are actually necessary. The system compares the parts presented by the learning model with the parts actually replaced during maintenance, identifies the correct answer data among them, and extracts only the necessary part information for continuous learning. This extraction process filters out unnecessary information while preserving the essential maintenance knowledge.
Solution Approach 2:
The patent implements feedback by using actual maintenance outcomes to correct and improve future maintenance recommendations. The system feeds back the comparison results between presented parts and actually replaced parts to the learning model, adjusting weights and improving accuracy. This feedback mechanism ensures that only truly necessary parts are identified and learned, preventing information loss despite multiple part replacements.
3Adaptability or versatility
If the learning model is re-trained frequently with new data, then the system adapts to new troubles, but noise from incorrect answers is continuously amplified
Solution Approach 1:
The patent applies preliminary action by identifying and weighting correct answer data before the re-learning process. Instead of allowing noise to accumulate during frequent re-training, the system pre-processes the maintenance information, identifies correct answers in advance, and assigns appropriate weights. This preliminary weighting prevents noise amplification even when re-training occurs frequently, maintaining both adaptability and reliability.
Solution Approach 2:
The patent converts the potentially harmful effect of frequent re-training (noise amplification) into a benefit by using the re-learning process to systematically identify and weight correct answer data. The frequent updates, which could amplify noise, are instead used to reinforce correct maintenance information through proper weighting, transforming the potential harm into improved reliability and adaptability.
Data Source
AI summary
A system includes: one or plural processors configured to: acquire information related to a trouble and information on maintenance executed for the trouble; generate a learning model to which the information related to the trouble is input and from which the information on the maintenance is output; re-train the learning model based on information related to a new trouble and information on the maintenance output for the new trouble; and perform weighting on the information on the maintenance in a case where the learning model is re-trained.