Thermal Displacement Compensation Using Ambient Temperature Model Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing thermal displacement compensation technologies for manufacturing machines are inefficient in adapting to changes in ambient temperature environments, requiring frequent relearning and additional training when machines are moved to different locations, which is time-consuming and inefficient.
Innovation Solution
A thermal displacement compensation device that performs initial machine learning in an isothermal environment and logs temperature changes, allowing for the selection of a suitable learned model from stored data when the machine is moved to a new location with similar temperature patterns, reducing the need for extensive additional learning.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If machine learning is performed from scratch in a new ambient temperature environment, then the thermal displacement compensation accuracy is improved, but the time required for relearning increases significantly
Solution Approach 1:
The system performs preliminary machine learning in an isothermal ambient temperature environment before actual use, storing the learned model in advance. When the manufacturing machine is deployed, this pre-learned model serves as a foundation, requiring only additional learning rather than complete relearning, thus reducing relearning time while maintaining compensation accuracy
Solution Approach 2:
The system dynamically adapts the learned model based on the actual ambient temperature environment. It logs temperature changes, calculates statistical values (average and variance), and performs additional learning to adjust the pre-learned model to match the new environment, optimizing both accuracy and efficiency
2Loss of time
If a learned model is used without additional learning, then the relearning time is reduced, but the compensation accuracy deteriorates in different ambient temperature environments
Solution Approach 1:
Instead of performing complete machine learning from scratch or using the pre-learned model without any adjustment, the system performs partial additional learning. This additional learning is targeted and limited, focusing only on adapting to the specific ambient temperature environment, thus achieving a balance between time consumption and accuracy improvement
Solution Approach 2:
The system changes the learning parameters by performing additional learning with specific focus on the new ambient temperature environment. It uses temperature change logs and statistical calculations to adjust the pre-learned model parameters, enabling accurate compensation in different environments without requiring complete relearning
3Measurement precision
If multiple temperature sensors are introduced and complex selection algorithms are used, then the thermal displacement prediction accuracy is improved, but the device complexity increases
Solution Approach 1:
The system uses the manufacturing machine's own temperature sensors and operational data to perform self-learning and self-compensation. The machine learning device learns from the machine's actual temperature changes during operation, eliminating the need for external complex sensor systems or manual calibration, thus achieving accurate prediction without increasing device complexity
Data Source
AI summary
A machine learning device includes a model data selection unit to select, under a change in ambient temperature of a manufacturing machine, a learned model for additional learning of a thermal displacement compensation amount in each axis included in the manufacturing machine with respect to an operation state of the manufacturing machine, and a learned model storage unit to associate and store a pattern of an ambient temperature change curve indicating a transition of a change in the ambient temperature of the manufacturing machine and the learned model that is learned under the change in the ambient temperature. Based on the ambient temperature change curve stored in the learned model storage unit, the model data selection unit selects a learned model suitable for the additional learning of the thermal displacement compensation amount in each axis included in the manufacturing machine with respect to the operation state of the manufacturing machine.