ML-Based Part Determination for CAD Assembly Design

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Modern CAD and PDM systems face challenges in efficiently managing immense part libraries, where current search strategies are cumbersome and ineffective due to the overwhelming number of potential parts, disparate data storage, and the need for constant data maintenance to ensure accuracy, especially when considering various factors like physical characteristics, cost, availability, and industry standards.

Innovation Solution

Implementing machine learning-based part determination features that leverage hierarchical relationships within CAD assembly structures and integrate data from disparate sources through a common ontology, using ML models like naive Bayesian classifiers to recommend parts efficiently and accurately.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If traditional search strategies are used to manage part libraries, then data accuracy can be maintained through constant maintenance, but user time and operational efficiency are significantly reduced

Engineering Contradiction:
Improvepart determination accuracyVSAvoiduser time for part search
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent replaces traditional mechanical search strategies (manual filtering and browsing) with a machine learning-based system that automatically determines part candidacy. The ML model analyzes assembly structures and part data to automatically identify candidate parts, eliminating the need for users to manually search through immense part libraries while maintaining accurate part determinations.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

2Reliability

If traditional search strategies are used, then data integrity can be maintained, but device complexity and computational requirements increase

Engineering Contradiction:
Improvedata integrityVSAvoidsearch system complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent replaces complex traditional search systems with a streamlined machine learning model. Instead of implementing complex filtering and search algorithms, the system uses trained ML models (such as naive Bayesian classifiers) that automatically process assembly structures and part data, reducing system complexity while maintaining or improving data integrity through automated, consistent evaluation.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The ML-based part determination system is self-service in that it automatically determines candidate parts without requiring user intervention or complex search configurations. The system self-manages the part determination process by automatically analyzing assembly structures, evaluating part data, and identifying candidate parts, thereby simplifying the overall system architecture.

Inventive Principle:
Principle #25Self-service

3Measurement precision

If comprehensive part data from disparate sources is integrated, then part determination accuracy improves, but data maintenance burden increases

Engineering Contradiction:
Improvepart determination accuracyVSAvoiddata maintenance ease
Core Design Contradiction:
Measurement precisionVSEase of manufacture

Solution Approach 1:

The patent creates a universal ML-based part determination system that can handle data from disparate sources through a common ontology. The ML model is designed to process multiple data types and sources uniformly, integrating physical characteristics, cost, availability, and industry standards data through a single automated framework that maintains accuracy while reducing maintenance burden through standardized processing.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Solution Approach 2:

The patent replaces manual data maintenance processes with automated ML-based evaluation. Instead of requiring constant manual updates and verification of part data from disparate sources, the system uses ML models to automatically process and evaluate data, reducing the maintenance burden while maintaining or improving determination accuracy through consistent automated analysis.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

4Productivity

If ML-based part determination is implemented, then productivity and resource efficiency improve, but initial system complexity increases

Engineering Contradiction:
Improvepart determination efficiencyVSAvoidML system complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The patent applies preliminary action by training ML models in advance with comprehensive part data and assembly structures. The models are pre-trained to recognize patterns and relationships, so that during actual part determination tasks, they can quickly and efficiently identify candidate parts without requiring complex real-time processing. This preliminary training phase consolidates the complexity, making the operational system simpler and more productive.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP3951633B1Machine learning-based part determinations for computer-aided design (CAD) assemblies
Publication Date: 2025.11.19 SIEMENS INDUSTRY SOFTWARE INC
  • EP3951633B1 patent drawingFigure 1
  • EP3951633B1 patent drawingFigure 2
  • EP3951633B1 patent drawingFigure 3

AI summary

A computing system (100) may include an assembly access engine (108) configured to access a computer-aided design, CAD, assembly (120) that digitally represents a product component that includes multiple parts. The computing system (100) may also include a part determination engine (110) configured to determine a recommended part (150) for the CAD assembly (120), including by providing the CAD assembly (120) as an input to a machine-learning (ML) model (130) trained with assembly structure data of CAD assemblies of a common product type as the CAD assembly, generating a candidate part set (140) through the ML model (130), filtering the candidate part set (140) based on physical and cost characteristics of the different candidate parts of the candidate part set, and identifying the recommended part (150) from the filtered candidate part set (140). The part determination engine (110) may also be configured to insert the recommended part (150) into the CAD assembly (120) and provide the CAD assembly (120) in support of physical manufacture.