Machine Learning Screen Prediction for CNC Controllers
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In numerical controllers for machine tools, the increasing number of operation screen functions leads to larger data sizes, prolonging start-up times and causing memory shortages, as well as frequent screen switching that reduces productivity, due to the need to load all screen data into memory.
Innovation Solution
A machine learning device predicts the next operation screen based on a series of user operations and machine states, using a state vector sequence to generate a learned model that determines the type of screen to be displayed next, allowing only the necessary screen data to be loaded, thereby reducing the volume of data required.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If all screen data about operation screens are loaded into memory at start-up, then screen switching speed is improved, but start-up time is extended and memory shortage occurs
Solution Approach 1:
The patent applies preliminary action by loading screen data into memory in advance before it is actually needed. The prediction unit predicts which screens will be needed based on the current screen and historical usage patterns, then the screen data loading unit loads those predicted screens into memory beforehand. This allows screen switching to be fast when needed while avoiding the overhead of loading all possible screens at system start-up.
Solution Approach 2:
The patent implements dynamics by making the screen loading strategy adaptive rather than static. The system continuously learns from user behavior patterns and dynamically adjusts which screens are loaded into memory based on predicted future usage. This dynamic approach replaces the traditional static method of loading all screens at start-up, optimizing both start-up time and memory usage while maintaining fast screen switching performance.
2Speed
If all screen data about operation screens are loaded into memory at start-up, then screen switching speed is improved, but memory allocation is insufficient
Solution Approach 1:
The patent applies partial action by loading only the necessary subset of screen data into memory rather than all available screen data. The prediction unit identifies which screens are likely to be needed next, and only those specific screens are loaded into the limited memory resource. This partial loading strategy ensures sufficient memory allocation while maintaining fast screen switching for the predicted screens.
Solution Approach 2:
The system dynamically adjusts memory allocation based on real-time predictions of screen usage patterns. Instead of a fixed memory allocation for all screens, the system adaptively loads only the predicted necessary screens into memory, optimizing the balance between memory consumption and screen switching performance.
3Quantity of substance
If screen data is loaded in advance for multi-window screens, then data volume is reduced, but memory management complexity increases
Solution Approach 1:
The patent implements feedback through the learning unit that continuously monitors actual screen usage patterns and compares them with predictions. This feedback loop allows the system to refine its prediction accuracy over time and adjust its data loading strategy accordingly, managing the complexity through adaptive learning rather than rigid rules.
Solution Approach 2:
The system performs self-service by automatically learning and adapting to user behavior patterns without requiring manual configuration or complex memory management rules. The learning unit autonomously identifies usage patterns and the prediction unit autonomously determines which screens to load, simplifying memory management while reducing data volume.
4Quantity of substance
If machine learning is used to predict operation screens, then data loading volume is minimized, but computational overhead is introduced
Solution Approach 1:
The patent applies preliminary action by performing the computationally intensive machine learning training process in advance during periods when the system is not in active use. The learning unit trains prediction models beforehand based on accumulated usage data, and then the trained models are deployed for rapid prediction during actual operation. This separates the heavy computational work from the real-time operational phase, minimizing computational overhead during active use while still achieving data volume reduction through accurate predictions.
Data Source
AI summary
A machine learning device includes: a state observation unit that acquires a state vector sequence as a group of state vectors corresponding to a transition of operation screens responsive to a series of operations or change in a machine state; a training data acquisition unit that acquires training data containing input data and label data; and a learning unit that performs supervised machine learning using the training data to generate a learned model for predicting screen type data to be used next or on an occasion after the next by a user.