Bayesian Optimization for Injection Molding Condition Derivation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing injection molding methods require skilled workers to derive proper molding conditions, and they often necessitate a large number of training data and expensive high-resolution measurement devices, making it difficult to easily obtain a proper molding condition.
Innovation Solution
An injection molding method using a Bayesian optimization method with a regression model to infer a predictive distribution of quality values, allowing for the derivation of molding conditions that satisfy required quality without relying on skilled workers or extensive data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a neural network is used to optimize molding product quality, then the prediction accuracy of quality values is improved, but a large number of training data (several hundred to several ten thousand data) are required
Solution Approach 1:
The patent changes the algorithmic approach from neural network to Bayesian optimization with regression model, which fundamentally alters the data requirements. This parameter change in the optimization method enables achieving comparable prediction accuracy with significantly fewer training data points, directly resolving the contradiction between accuracy and data quantity requirements
Solution Approach 2:
The patent substitutes the neural network mechanism with a Bayesian optimization mechanism. This replacement changes the underlying computational approach from one requiring extensive training data to one that can effectively optimize with limited data, thereby resolving the contradiction without sacrificing prediction accuracy
2Measurement precision
If quality values requiring high-resolution measurement devices are used (sink mark, flow mark), then the measurement precision is improved, but the device cost increases and additional cutting work is required
Solution Approach 1:
The patent uses sensor data from the molding process as a copy or proxy for direct quality measurement. Instead of measuring actual quality defects with expensive equipment, the system uses correlated sensor measurements (pressure, temperature, time) that can indirectly indicate quality outcomes, eliminating the need for costly measurement devices and sample preparation
Solution Approach 2:
The patent introduces sensor data as an intermediary between the molding process and quality assessment. These sensors measure process parameters that correlate with quality outcomes, serving as a mediator that provides quality information without requiring direct measurement of defects using expensive equipment
3Ease of operation
If a worker with less knowledge and experience derives molding condition, then the operational flexibility is improved, but the time required for trial and error increases significantly
Solution Approach 1:
The patent implements a feedback loop where sensor data from each molding cycle is fed back into the Bayesian optimization model. This feedback mechanism allows the system to learn from previous results and automatically adjust molding conditions, enabling workers with less experience to achieve optimal results without extensive trial and error time
Solution Approach 2:
The system performs self-optimization by automatically analyzing sensor data and deriving improved molding conditions without requiring external expert intervention. The Bayesian optimization algorithm serves itself by using its own outputs to generate better inputs, allowing less experienced operators to achieve expert-level results efficiently
Data Source
AI summary
This injection molding method includes the steps of: constructing a prediction model on the basis of an input parameter including a molding condition for a molding product and an objective variable value including a quality value that quantifies a required quality of the molding product with respect to the input parameter; inferring a predictive distribution of the objective variable value with respect to the input parameter, using the prediction model; and deriving such a molding condition that satisfies the required quality of the molding product, by a Bayesian optimization method utilizing a regression model for obtaining the input parameter that yields a quality value highest in evaluation of the objective variable value as compared to an initial quality value, on the basis of the predictive distribution.