Machine Diagnosis Model Selection for New Equipment Variations
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for diagnosing industrial machines struggle with accuracy when a new machine is introduced, as they require sufficient learning data and may not account for individual machine characteristics, leading to difficulties in diagnosing operations before a model is constructed.
Innovation Solution
A diagnosis apparatus that stores learning models for multiple machines, calculates characteristic differences, and selects the most similar model for diagnosis, allowing for accurate operation state assessment even before a specific model is constructed for the new machine.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a model is constructed for a newly introduced industrial machine using sufficient learning data, then diagnosis accuracy is improved, but the time required for model construction increases to one or two months or more
Solution Approach 1:
Learning models are pre-constructed for multiple reference industrial machines before the new machine is introduced. These reference models are stored in advance in the storage unit, so when a new machine needs diagnosis, the system can immediately compare and select from existing models without waiting for data collection from the new machine itself.
Solution Approach 2:
Instead of creating a new model from scratch for the newly introduced machine, the system copies the diagnosis approach by selecting the most similar existing reference model. The characteristic difference calculation unit compares the new machine's characteristics with reference models and selects the closest match, effectively copying the diagnostic capabilities of a similar machine.
2Productivity
If a model from another industrial machine of the same type is used for diagnosis, then diagnosis can be performed before sufficient learning data is acquired, but diagnosis accuracy deteriorates when characteristic differences between machines are large
Solution Approach 1:
The system calculates characteristic differences specifically for relevant parameters between the newly introduced machine and reference machines. By focusing on local characteristic comparisons rather than generic model matching, the system can identify the most suitable reference model even when machines have individual variations, thereby maintaining diagnosis accuracy while enabling immediate diagnosis capability.
Solution Approach 2:
The system changes the parameter of model selection from using a fixed or generic model to dynamically selecting models based on calculated characteristic differences. By introducing the characteristic difference calculation unit that compares specific parameters between machines, the system adapts the model selection process to match the actual characteristics of the new machine, improving accuracy while maintaining quick deployment.
Data Source
AI summary
A plurality of learning models generated by performing machine learning on physical quantities observed during respective operations of a plurality of machines are stored in advance. Then, using the stored learning models and a physical quantity observed during an operation of a machine which is an object to be diagnosed, characteristic differences between the machine which is the object to be diagnosed and the plurality of respective machines are calculated, a learning model used in a diagnosis of the operation of the machine which is the object to be diagnosed is selected based on the calculated characteristic differences, and the operation of the machine which is an object to be diagnosed using the selected learning model is diagnosed.