ML Training Data Generation via Simulation-Prototype Merging
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for training machine learning models to predict the behavior of technical systems are inefficient and costly due to the lack of sufficient data in the prototype phase, requiring large amounts of data to achieve the required quality and accuracy.
Innovation Solution
A method that combines simulation data with prototype data, using an active learning approach to generate training data by maximizing an information function, which correlates data from the sample phase with the prototype phase, allowing for reduced data requirements and flexible adaptation to model quality, with optional termination criteria based on model performance or resource constraints.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If simulation data are used to train ML models in the prototype phase, then data generation cost is reduced, but model accuracy and quality are insufficient
Solution Approach 1:
The patent combines simulation data from the sample phase with prototype data from the production phase to create a hybrid training data set. This merging allows the model to benefit from both the low cost of simulation data and the high accuracy of prototype data, resolving the contradiction between data generation cost and model accuracy.
Solution Approach 2:
The patent performs preliminary training of the ML model using simulation data from the sample phase before transitioning to the production phase. This preliminary action enables the model to be initialized with reasonable accuracy, and subsequent training with prototype data can then refine the model without requiring all data to be of high quality from the outset.
2Measurement precision
If more prototype data are collected to improve model quality, then model accuracy increases, but data collection cost and time increase
Solution Approach 1:
The patent merges simulation data with prototype data to create a hybrid training set, which reduces the quantity of prototype data needed while maintaining model quality. The simulation data compensates for the limited prototype data, allowing the model to achieve required accuracy without collecting extensive prototype data.
Solution Approach 2:
The patent uses simulation data as a copy or representation of prototype data. The simulation models replicate the behavior of the technical system, providing virtual copies of prototype data that can be used for training without requiring actual physical measurements from every prototype unit.
3Productivity
If active learning is used to select informative data, then data efficiency improves, but the complexity of the training process increases
Solution Approach 1:
The patent implements active learning with a feedback loop where the model continuously evaluates its own uncertainty and selects the most informative data points for training. This feedback mechanism improves data efficiency by focusing computational resources on the most valuable data, though it does increase the complexity of the training process through iterative selection and retraining cycles.
Data Source
AI summary
A method for providing training data for a machine learning (ML) model for predicting the behavior of a technical system. The method includes: generating a first data set containing simulation data for the technical system; generating a second data set containing prototype data of the technical system; generating a third data set by combining the first and second data sets; training the ML model based on the third data set; generating a fourth data set as first input data based on the third data set by maximizing an information function, to obtain a first feature combination as input data for the technical system; measuring the first feature combination for the prototype data of the technical system to obtain a fifth data set; and adding the fifth data set as output data and the generated fourth data set as first input data to the third data set as training data.