Machining Dimension Prediction Using Relevant Cutting Sections
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machining dimension prediction techniques, such as those described in Patent Literature 1, are inadequate for accurately predicting dimensions of workpieces machined with curving or partial cutting methods due to non-uniform cutting, resulting in large prediction errors.
Innovation Solution
A machining dimension prediction apparatus that acquires trend information of a machining tool's state during machining, calculates features for each section of the machining period, specifies the section with the greatest relevance to the measurement value, and uses this information to predict the dimension of a new workpiece based on the trend of the state in that specific section.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If features are extracted from information collected in a predetermined machining section, then the prediction model can be trained to predict machining dimension, but the prediction accuracy deteriorates for curving cutting or partial cutting due to non-uniform cutting
Solution Approach 1:
The machining period is divided into multiple sections, and features are calculated for each section separately. The section specifying means identifies the specific section most relevant to the measurement position, and the prediction model uses features from that specific section. This segmentation approach allows the system to adapt to different cutting types (uniform, curving, partial) by selecting the appropriate section for each prediction task, thereby improving both prediction accuracy and versatility.
2Loss of information
If the prediction model is trained using features from the entire machining period, then comprehensive information is utilized, but the prediction error increases for non-uniform cutting due to irrelevant sections diluting the relevant features
Solution Approach 1:
The section specifying means extracts and identifies the specific section within the machining period that has the greatest degree of relevance to the measurement position. Instead of using features from the entire machining period, the system selectively extracts features from the identified specific section. This extraction process eliminates irrelevant information from other sections that would dilute the predictive power, thereby improving prediction accuracy while maintaining efficient information utilization.
3Manufacturing precision
If inspection is performed on all workpieces after machining, then dimension verification is ensured, but the process becomes difficult and time-consuming
Solution Approach 1:
The system uses measurement values from inspected workpieces as feedback to train and improve the prediction model. The prediction model learns from the relationship between machining state features and actual measurement results, enabling it to predict dimensions of subsequent workpieces without physical inspection. This feedback mechanism maintains dimension verification accuracy while eliminating the need for time-consuming inspection of every workpiece, thereby improving productivity.
Data Source
AI summary
A machining dimension prediction apparatus includes a trend acquirer that acquires, for each workpiece, trend information indicating a trend of a state of a machining tool during a machining period of machining performed by the machining tool, a feature calculator that calculates, based on the trend information, a feature using the trend of the state in each of sections included in the machining period, a measurement value acquirer that acquires a measurement value of a dimension of each workpiece after being machined, a section specifier that specifies, as a specific section of the sections, a section including a calculated feature having a greatest degree of relevance to the measurement value, and a predictor that predicts, when a new-target workpiece is machined, a dimension of the new-target workpiece after being machined based on the feature calculated using the trend of the state in the specific section.