Searching for optimal setting conditions in technological processes using parametric estimation models and neural...

Area of Science:

• Engineering Optimization
• Industrial Process Improvement
• Machine Learning Applications

Background:

• Engineering optimization is crucial for manufacturing and service industries.
• Traditional methods like Factorial Design (FD) and Central Composite Design (CCD) are used for process parameter optimization.
• There is a need for methods that visualize all optimal solutions for enhanced system performance.

Purpose of the Study:

• To compare traditional parametric estimation models with Auto Associative Neural Networks (ANN) for engineering optimization.
• To introduce the Feed Forward Bottle Neck Neural Network (FFBN NN) for visualizing optimal solutions.
• To evaluate the effectiveness of these methods in optimizing real-world processes.

Main Methods:

• Utilized traditional parametric estimation models: Factorial Design (FD) and Central Composite Design (CCD).
• Implemented a 2D mapping technique using Auto Associative Neural Networks (ANN), specifically the Feed Forward Bottle Neck Neural Network (FFBN NN).
• Applied both methods to optimize solder paste printing processes and cheese properties.

Main Results:

• The FFBN NN enables visualization of all optimal solutions by projecting input and output parameters onto a 2D map.
• This visualization facilitates a more efficient improvement of existing system performance.
• Comparative analysis demonstrated the application of both methods for double-checking results, increasing reliability.

Conclusions:

• The FFBN NN offers a more comprehensive approach to engineering optimization by visualizing all potential optimal solutions.
• Comparing traditional methods with FFBN NN enhances the reliability of identified optima and specification limits.
• This study highlights the potential of ANN-based methods for improving industrial process optimization and system performance.