Production Machine Parameter Tuning Using Random Feedback Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Production machines often require complex parameter settings, which can be challenging for less experienced operators to manage effectively, especially during shifts when operator familiarity with the machine's state is limited.
Innovation Solution
A method for optimizing production machine parameters using sensors and computer-set parameters, involving random modification of parameter values, evaluation of output, fitting a linear function model, and estimating impacts to determine optimal modifications.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If human operators set parameters manually, then the machine can produce desired output, but the operation depends on operator know-how and experience
Solution Approach 1:
The production machine automatically optimizes its own parameters through the computer apparatus that randomly determines modification values, evaluates output, fits linear function models, and determines optimal parameter settings without continuous human intervention. This self-service capability resolves the contradiction by making the machine independent of operator know-how while maintaining reliable operation.
Solution Approach 2:
The system continuously evaluates the output of the production machine and uses this feedback to adjust parameters. The computer apparatus monitors machine operation, assesses output quality, and automatically modifies parameters based on the evaluation, creating a closed-loop feedback system that ensures consistent operation regardless of operator experience.
2Productivity
If parameters are set by experienced operators, then desired output is achieved, but less experienced operators cannot adequately operate the machine
Solution Approach 1:
The machine performs self-optimization through automated parameter adjustment based on output evaluation. The computer apparatus independently determines optimal parameter settings without requiring operator expertise, thereby maintaining high production quality while eliminating the need for specialized operator knowledge.
Solution Approach 2:
The patent replaces the mechanical knowledge-based system (operator expertise) with an automated computational system. The computer apparatus uses algorithms to determine parameter modifications, substituting human cognitive processes with automated computing, thereby decoupling production quality from operator skill level.
3Loss of information
If random modification values are used for parameter optimization, then the system can learn effectively about parameter impacts, but the operation becomes noisy
Solution Approach 1:
The system uses feedback from output evaluation to guide random parameter modifications. By continuously monitoring the results of random changes and using linear function models to estimate parameter impacts, the system learns effective modifications while maintaining operational stability through informed adjustment rather than pure randomness.
Solution Approach 2:
The patent systematically changes parameters through controlled random modifications within predefined step sizes. This approach allows the system to explore parameter space and learn impact relationships while maintaining stability through bounded changes and model-based estimation, balancing exploration with operational predictability.
Data Source
AI summary
Method for optimising a parameter of a machine, comprising: randomly determining a modification value for modifying a current value of the parameter, based on the respective current value of the parameter and on a step size; modifying, at the machine, the parameter to its modification value; evaluating an output of the machine, effected using the modified parameter; fitting a linear function model for the parameter, based on the evaluated output; estimating, using the linear function model, an impact on the output, if the parameter is modified by at least the step size; determining whether or not to modify the parameter, based on a desired output, taking into account the estimated impact; and, if so, modifying the parameter by at least its step size.