Batch Plant Regressor Architecture for Sparse-Data Quality Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for simulating and controlling batch processes in chemical or biological reactors are inefficient due to high computational requirements for whitebox models and inflexibility in blackbox models, which often lack sufficient training data and require retraining when process conditions change.
Innovation Solution
A computer-implemented regressor system using nested artificial neural networks (ANNs) that separates the modeling of educt quality parameters and process parameters, allowing for efficient training with sparse data sets and enabling the reuse of trained models across similar batch plants, thereby predicting product quality parameters reliably.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If whitebox models are used to simulate batch processes, then model accuracy is improved, but computational power and resources are excessively consumed
Solution Approach 1:
The patent segments the batch process into multiple phases (charging, reaction, discharging, cleaning) and trains separate neural network models for each phase. This segmentation allows the system to achieve accurate predictions without requiring a single complex whitebox model, thereby reducing computational power consumption while maintaining model accuracy.
2Use of energy by moving object
If blackbox models are used to predict product quality, then computational resources are reduced, but the models require extensive training data and retraining when conditions change
Solution Approach 1:
The patent implements a dynamic model selection mechanism that automatically selects and switches between different neural network models based on the current process phase and conditions. This dynamic approach allows the system to adapt to changing conditions without extensive retraining, as each phase-specific model is optimized for its particular operating conditions while the overall system remains flexible.
3Measurement precision
If conventional blackbox models are retrained when process conditions change, then prediction accuracy is maintained, but system flexibility and ease of operation deteriorate
Solution Approach 1:
The patent performs preliminary action by training separate neural network models for each anticipated process phase before actual operation. When process conditions change, the system can switch between pre-trained phase-specific models rather than requiring retraining of a single comprehensive model. This preliminary segmentation of training efforts maintains prediction accuracy while significantly improving system flexibility and ease of operation.
Data Source
AI summary
Disclosed herein is a computer-implemented regressor for simulating, monitoring and/or controlling a batch plant. The batch plant is implemented to receive one or more educts having associated educt quality parameters (x1, x2), to process the educt(s) where the process has associated process parameters (yj), and to output a product having associated product quality parameters (Q1, Q2). The regressor includes at least two regressor units based on machine-learning principles, each regressor unit having an input for receiving input data, and an output for outputting output data. A first regressor unit is implemented to receive the educt quality parameters (xi) and to output at least one educt impact parameter (R1). And a second regressor unit is implemented to receive the educt impact parameter (R1) and l process parameters (yj) and to output at least one product quality parameter (Qj).