Adaptive Process Configuration Using ML and Real-Time Status Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for optimizing process configurations in complex industrial systems are often inefficient and rely on experimental phases, failing to account for dynamic changes such as temperature fluctuations and wear on systems.
Innovation Solution
A computer-implemented method for automatically determining an optimized process configuration using historical data, machine learning, and real-time status data to adapt process settings dynamically, thereby improving efficiency and quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If experimental phases are used to optimize process configurations, then sufficient process knowledge can be obtained, but the optimization process becomes time-consuming and inefficient
Solution Approach 1:
The patent applies preliminary action by using historical process data and machine learning models to pre-determine optimized process configurations before actual production. The system analyzes past operational data, identifies optimal parameter settings, and stores them as reference configurations, eliminating the need for time-consuming experimental phases during production changes.
Solution Approach 2:
The patent uses copying by creating virtual models and digital twins of physical production systems. These digital replicas allow optimization experiments to be conducted in silico rather than on actual production lines, transferring the optimized configurations back to the physical system without time loss from physical experimentation.
2Ease of operation
If static optimized initial configurations are used, then implementation is simple, but the system cannot adapt to dynamic changes during process execution
Solution Approach 1:
The patent implements dynamics by transitioning from static configuration files to dynamic, adaptive systems. Machine learning models continuously monitor current process conditions and automatically adjust process parameters in real-time based on detected changes in material properties, environmental conditions, or equipment status, maintaining optimality throughout the production process.
Solution Approach 2:
The patent applies feedback by establishing closed-loop control systems where process sensors continuously measure actual parameters, compare them against target values from the optimized configuration, and automatically trigger re-optimization when deviations exceed thresholds. This feedback mechanism enables the system to adapt to dynamic changes while maintaining operational simplicity through automated control.
3Reliability
If multiple alternative calculation methods are used for optimization, then optimization reliability increases, but computational complexity increases
Solution Approach 1:
The patent applies segmentation by dividing the optimization problem into multiple independent calculation tasks that can be executed in parallel. Different machine learning models (e.g., neural networks, genetic algorithms, linear programming) are trained on different subsets of historical data or different process parameters, then their results are aggregated to produce the final optimized configuration, reducing overall computational complexity.
Solution Approach 2:
The patent uses merging by combining results from multiple alternative calculation methods through ensemble techniques. Rather than selecting a single complex model, the system integrates predictions from multiple simpler models, weighting them according to their performance on validation data, thereby achieving high optimization reliability while keeping individual computational components manageable.
Data Source
AI summary
A method for automatically determining an optimized process configuration of a process for manufacturing or processing products that can be executed using a technical system and can be configured using a number of different process configuration parameters comprises: determining a process configuration of the process that is optimized with regard to a defined metric and is defined by respective target values of process configuration parameters using an optimization method that is adapted to the process and is at least partially based on machine learning, using input data that include production data and features that are given by historical process configuration data and status data of the system or process or are derived therefrom; and outputting target process configuration data representing the determined optimized process configuration by means of the target values of the process configuration parameters.