Machine Learning Parameter Adjustment for Multi-Element Machining
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional machine learning methods require a large number of parameters and significant computational resources when learning multiple characteristic machining elements simultaneously, leading to increased calculation load and prolonged learning times.
Innovation Solution
An adjusting device and method that switches machine learning parts to perform learning based on specific machining elements, allowing for simultaneous learning of multiple elements using separate models during a single learning operation, optimizing learning for each element.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a single machine learning model is used to learn multiple characteristic machining elements simultaneously, then the system structure is simple, but the number of parameters increases and computational load becomes large
Solution Approach 1:
The patent divides the machine learning system into multiple independent learning models, with each model dedicated to learning a specific characteristic machining element. This segmentation reduces the number of parameters per model and decreases computational load compared to a single comprehensive model, while maintaining the ability to learn multiple elements through parallel model execution.
2Device complexity
If a single machine learning model is used to learn multiple characteristic machining elements, then the model structure is simple, but the learning period of time increases
Solution Approach 1:
By segmenting the learning task into multiple specialized models, each model can converge faster on its specific characteristic element. The overall learning time is reduced because each model learns its dedicated element more efficiently rather than a single model attempting to learn all elements simultaneously with a larger parameter space.
3Device complexity
If a single machine learning model is used to learn multiple characteristic machining elements, then the system is simple, but memory requirements increase
Solution Approach 1:
The patent segments the memory requirements by allocating separate memory spaces for each independent learning model. This approach reduces the peak memory usage compared to a single large model, as each smaller model only requires memory for its own parameters and computation, rather than all parameters being loaded simultaneously into one large model.
4Productivity
If multiple machine learning parts are used to learn different machining elements respectively, then learning efficiency is improved, but the system becomes more complex
Solution Approach 1:
The patent implements segmentation by creating multiple independent learning models, each optimized for a specific characteristic machining element. This segmentation improves learning efficiency by allowing parallel processing and specialized learning for each element type, while the modular structure manages complexity through clear separation of functions.
Data Source
AI summary
To provide an adjusting device and an adjusting method for appropriately controlling the machine learning reduced in cost with respect to calculation load and learning period of time in the case where an evaluation program for machine learning is used separately from a machining program and the like. The present invention includes a feedback information acquiring part configured to acquire, from a control device, feedback information obtained when an evaluation program including various types of learning elements is executed in the control device, a determination part configured to determine which learning element the acquired feedback information corresponds to among the various types of learning elements, a feedback information transmitting part configured to transmit the acquired feedback information to a machine learning part corresponding to the learning element, a parameter setting information acquiring part configured to acquire control parameter setting information obtained through machine learning by use of the feedback information, and a parameter setting information transmitting part configured to transmit the acquired control parameter setting information to the control device.