Deep Neural Network for Electrical Machine Design Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Designing electrical machines is a complex and time-consuming process due to limited understanding of interdependencies between components and parameters, often requiring numerous interim designs and computationally intensive finite element procedures.
Innovation Solution
A deep neural network system is used to generate and iteratively refine electrical machine designs based on goals and constraints, adjusting weights based on rewards from expert feedback to optimize design parameters and performance predictions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional mathematical equations and finite element procedures are used for designing electrical machines, then design accuracy is maintained, but design time and computational resources are significantly increased
Solution Approach 1:
The patent creates a neural network model that learns from and copies the design patterns and relationships embedded in historical design data and expert knowledge. Instead of performing computationally intensive finite element analyses for each design iteration, the system uses the trained neural network to generate and evaluate design proposals, dramatically reducing computational time while maintaining design quality through the patterns captured during training.
Solution Approach 2:
The system performs preliminary learning and pattern recognition during an offline training phase using historical design data and expert evaluations. This preliminary action embeds design knowledge into the neural network weights, enabling the system to make informed design decisions during the actual design process without requiring time-consuming iterative analyses for each specific design problem.
2Reliability
If numerous interim designs are tested and analyzed to obtain an acceptable final design, then design quality is improved, but the complexity and resource requirements of the design process increase
Solution Approach 1:
The patent implements a reinforcement learning framework where the neural network receives feedback in the form of reward signals based on how well design proposals meet specified goals and constraints. This feedback mechanism guides the iterative refinement of designs, allowing the system to learn from successful and unsuccessful design attempts and progressively improve design quality without requiring manual evaluation of numerous interim designs.
Solution Approach 2:
The system enables the neural network to autonomously generate, evaluate, and refine design proposals using the reward feedback mechanism. The network self-corrects and self-improves by adjusting its internal weights based on performance feedback, reducing the need for external expert intervention in evaluating each interim design while maintaining high design quality.
Data Source
AI summary
An example method of designing an electrical machine includes providing at least one goal and at least one design constraint for a desired electrical machine to a deep neural network that comprises a plurality of nodes representing a plurality of prior electrical machine designs, the plurality of nodes connected by weights, each weight representing a correlation strength between two nodes. A proposed design is generated from the deep neural network for an electrical machine based on the goal(s) and design constraint(s). A plurality of the weights are adjusted based on a reward that rates at least one aspect of the proposed design. The proposed design is modified using the deep neural network after the weight adjustment. The adjusting and modifying are iteratively repeated to generate subsequent iterations of the proposed design, each subsequent iteration based on the reward from a preceding iteration. A system for designing electrical machines is also disclosed.