Neural Network Remaining-Life Learning With End-Cycle Feedback
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for predicting the remaining life of devices are inefficient in accurately determining the maintenance timing due to variations in end-of-maintenance cycle data, leading to improper learning and maintenance scheduling.
Innovation Solution
A neural network system comprising a first model for predicting remaining life at an arbitrary time and a second model for predicting remaining life at the final time of the maintenance cycle, with weight parameter updates using outputs from both models to improve prediction accuracy based on end-of-maintenance cycle data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a single model is used to predict remaining life at arbitrary times, then the model is simple and easy to implement, but the prediction accuracy deteriorates due to variations in end-of-maintenance cycle data
Solution Approach 1:
The patent divides the prediction task into two separate models: a first model that predicts remaining life at arbitrary times during the maintenance cycle, and a second model that predicts remaining life at the final time of the maintenance cycle. This segmentation allows each model to be optimized for its specific prediction task, improving overall prediction accuracy while maintaining reasonable complexity in each individual model.
2Ease of operation
If maintenance timing is determined without accounting for variations in end-of-maintenance cycle data, then the maintenance process is simple, but the maintenance scheduling becomes improper and inefficient
Solution Approach 1:
The patent implements a feedback mechanism where the second model's prediction of remaining life at the final time is used to update and refine the first model's predictions. This feedback loop allows the system to account for variations in end-of-maintenance cycle data, enabling more accurate and efficient maintenance scheduling without significantly increasing operational complexity.
Data Source
AI summary
Provided is a method of learning a neural network that predicts a remaining life of a target device that is a maintenance target. The neural network includes: (i) a first model for predicting a remaining life at an arbitrary time of maintenance cycle data, as a value based on an arbitrary reference value; and (ii) a second model for predicting a remaining life at a final time of the maintenance cycle data, as a value based on the reference value. The method comprises updating a weight parameter of the first model so as to predict a remaining life based on an end of the maintenance cycle data, by using an output of the first model and an output of the second model that are obtained from learning data including a plurality of maintenance cycle data.