Method for determining at least one target process parameter and control unit
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for determining target process parameters in complex manufacturing processes, such as cooking, cooling, storage, and transport, face challenges in precision and simplicity due to the influence of numerous factors and the distribution of processes across multiple devices, making it difficult to predict and control the process result effectively.
Innovation Solution
A method utilizing an artificial neural network that considers past and future values of process parameters, including environmental, control, and product parameters, to accurately determine target process parameters, with the network including convolutional and recurrent layers for efficient prediction, allowing for precise control of processes across various devices.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional methods are used to determine target process parameters in complex manufacturing processes, then the system is simpler to implement, but the precision and accuracy of process control deteriorates due to numerous influencing factors and distributed processing devices
Solution Approach 1:
The patent replaces traditional mechanical/mathematical control systems with an artificial neural network system. The neural network learns complex non-linear relationships between process parameters and outcomes through training data, substituting conventional control algorithms and enabling accurate prediction of target process parameters without requiring explicit mathematical models of the complex manufacturing process
Solution Approach 2:
The artificial neural network acts as an intermediary between process measurements and target parameter determination. It processes historical and real-time data from multiple sensors and processing devices, transforming raw process data into accurate predictions of target parameters through its learned representations, thereby simplifying the overall control architecture while maintaining high precision
2Measurement precision
If process parameters are determined without considering future values, then the computational load is reduced, but the accuracy of process prediction deteriorates
Solution Approach 1:
The patent applies preliminary action by using recurrent neural network layers that process historical process data in sequence, preparing and organizing temporal patterns before final prediction. The network pre-processes and encodes historical values and their temporal relationships, making the subsequent prediction of future process outcomes more accurate while managing computational complexity through efficient sequence processing
Solution Approach 2:
The patent transforms the prediction problem from considering only current state to incorporating temporal dimension by processing historical sequences. The recurrent neural network adds the time dimension to the analysis, allowing the system to leverage patterns from past and future values (in training context) to improve prediction accuracy without proportionally increasing computational burden through efficient temporal feature extraction
3Productivity
If multiple processing devices are used for product processing, then the productivity increases, but the difficulty of controlling quality and predicting process results increases
Solution Approach 1:
The patent implements a universal artificial neural network system that can handle multiple processing devices and various process parameters simultaneously. The neural network is trained on comprehensive data from different devices and processing stages, creating a unified model that predicts quality outcomes across the entire distributed processing system, thereby maintaining quality control accuracy while enabling high productivity through multiple devices
Solution Approach 2:
The patent merges data from multiple processing devices and various process parameters into a unified neural network model. By combining information from different sources (environmental parameters, control parameters, product parameters) and multiple devices into a single predictive system, the patent achieves coordinated quality control across distributed processing operations, enabling accurate prediction despite the complexity of multiple devices working in parallel
Data Source
Figure 1
Figure 2a~2b
Figure 3a~3b
AI summary
A method for determining at least one target process parameter (Z1, Z2; ZK1 - ZKn, ZZ1 - ZZn) in which a product is processed in at least one processing device (10) comprises the following steps: a) providing past values of at least one first and/or second process parameter (P1 - P6; PK1 - PKn, PZ1 - PZn) of the processing process, b) providing future values of the at least one first process parameter (P1 - P4; PK1 - PKn, PZ1 - PZn) of the processing process, and c) determining the values of the at least one target process parameter (Z1, Z2; ZK1 - ZKn, ZZ1 - ZZn) for the present and/or for a future time by a prediction module (20, 22) comprising at least one artificial neural network (24, 26, 28). Furthermore, a control unit, a computer program, and a computer-readable data carrier are disclosed.