Neural Network Training Data Generation for Flow Reaction Conditions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In flow reaction processes, finding optimal reaction conditions is challenging due to numerous condition parameters, requiring extensive trials and time, especially in new reaction systems, and adjusting one parameter complicates determining changes in others, while neural networks struggle to be accurately trained without sufficient learning data, which is limited by material, facility, and time constraints.
Innovation Solution
A data generation device and method that sets a range of process conditions for a neural network, generating learning data by simulating various conditions, and a learning device that iteratively refines connection weights to achieve accurate predictions, ensuring appropriate condition setting.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a large amount of learning data is prepared to improve neural network training accuracy, then the learning accuracy is improved, but the time required and resource consumption increase significantly
Solution Approach 1:
The system performs preliminary action by proactively identifying and collecting relevant learning data before neural network training is needed. The data collection unit gathers process conditions and results in advance, storing them in a database so that when training is required, the data is already prepared and available, eliminating the time-consuming data collection phase.
Solution Approach 2:
The system implements self-service by automatically collecting, storing, and managing learning data without requiring manual intervention. The data collection unit automatically accumulates process conditions and results from the production process, and the storage unit maintains this data ready for training, making the data preparation process autonomous and efficient.
2Manufacturing precision
If the number of condition parameters is increased to improve the comprehensiveness of process control, then the control accuracy is improved, but the complexity of determining optimal conditions increases
Solution Approach 1:
The system applies feedback by using the neural network to predict optimal process conditions based on stored learning data, then implementing these conditions in the production process. The results are fed back into the system to update and refine the neural network's understanding, creating a continuous improvement loop that handles multiple parameters systematically rather than through complex manual optimization.
Solution Approach 2:
The system replaces the mechanical/manual approach to optimizing multiple condition parameters with a computational neural network system. Instead of manually analyzing and adjusting numerous parameters, the neural network automatically processes the relationships between multiple condition parameters and results, substituting complex manual optimization with automated computational analysis.
Data Source
AI summary
A data generation device generates a data set consisting of a plurality of pieces of learning data for training a neural network in which a plurality of layers are connected by a plurality of connection weights, the neural network outputting a production result corresponding to a process condition in a case where the process condition is input in a process for producing a product. At this time, assuming that a total number of the connection weights of the neural network is M0, a plurality of the process conditions of 2×M0 or more are set. In addition, a production result corresponding to each of the plurality of process conditions is acquired, which is derived by producing the product under each of the plurality of process conditions. The plurality of pieces of learning data consisting of the plurality of process conditions and the production result are generated as the data set.