Welding Parameter Optimization Using ML-Based Quality Feedback

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current welding parameter optimization in welding controllers relies on manual trial and error, which is inefficient and fails to maintain optimal settings due to time variance in system behavior, particularly in welding aluminum and galvanized steel combinations with varying thicknesses, leading to unstable weld quality.

Innovation Solution

A computer-implemented method using a trained machine learning algorithm to automatically identify and program optimal welding parameters by receiving data sets from previous welds and target data, incorporating welding current, voltage, electrode force, and material-related parameters to approximate quality criteria, ensuring each weld is performed with tailored parameters.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If manual trial and error is used to optimize welding parameters, then the welding controller can be operated with simple preset parameters, but the optimization process is inefficient and cannot maintain optimal settings due to time variance in system behavior

Engineering Contradiction:
Improveoptimization efficiencyVSAvoidparameter optimization process
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The welding control system automatically optimizes welding parameters by evaluating actual weld quality outcomes and adjusting parameters without manual intervention. The system performs self-learning through iterative processes where weld quality assessments feed back into parameter optimization, eliminating the need for continuous manual trial and error while maintaining optimal settings despite system variations.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The system implements a closed-loop feedback mechanism where actual weld quality is measured and compared against target quality criteria. This feedback information is used to automatically adjust welding parameters, creating a continuous optimization cycle that adapts to time variance in system behavior and maintains optimal welding conditions without manual intervention.

Inventive Principle:
Principle #23Feedback

2Reliability

If automated optimization is implemented, then optimal welding parameters can be maintained over time, but the system complexity increases significantly

Engineering Contradiction:
Improveweld quality stabilityVSAvoidwelding control system
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The welding control system integrates multiple functions into a single unified platform: it performs weld execution, quality assessment, parameter evaluation, and automatic optimization. This multi-functional approach maintains reliability through comprehensive control while managing complexity by consolidating functions rather than adding separate systems.

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

Solution Approach 2:

The system autonomously maintains optimal welding parameters by continuously evaluating weld quality and self-adjusting parameters without external intervention. This self-service capability ensures consistent weld quality stability while avoiding the complexity of manual optimization processes.

Inventive Principle:
Principle #25Self-service

3Extent of automation

If machine learning algorithms are used to approximate quality criteria, then automated parameter optimization becomes possible, but computational requirements and system complexity increase

Engineering Contradiction:
Improveparameter optimizationVSAvoidcomputational system
Core Design Contradiction:
Extent of automationVSDevice complexity

Solution Approach 1:

The system replaces manual mechanical optimization processes with automated computational algorithms. Machine learning models approximate complex quality criteria and guide parameter optimization, substituting human expertise and trial-and-error mechanical adjustment with intelligent computational systems that automate the optimization process.

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

Solution Approach 2:

The system uses machine learning to dynamically change welding parameters based on approximated quality criteria. The algorithms evaluate how parameter changes affect weld quality and automatically adjust parameters to optimize outcomes, enabling high-level automation through computational parameter optimization rather than manual adjustment.

Inventive Principle:
Principle #35Parameter changes

4Manufacturing precision

If extensive data collection from previous welds is performed, then better parameter optimization can be achieved, but data processing time and computational load increase

Engineering Contradiction:
Improveparameter optimization accuracyVSAvoiddata processing time
Core Design Contradiction:
Manufacturing precisionVSLoss of time

Solution Approach 1:

The system performs preliminary actions by collecting and storing welding data during normal production operations. Rather than dedicating separate time for data collection, the system accumulates data continuously as welds are performed, preparing the dataset in advance for optimization processes. This approach improves optimization accuracy without adding significant processing time to individual weld operations.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP3915712B1Method of optimizing welding parameters for welding control, method of providing a trained algorithm for machine learning, welding control, computer program and computer-readable data medium
Publication Date: 2023.01.25 ROBERT BOSCH GMBH
  • EP3915712B1 patent drawingFigure 1
  • EP3915712B1 patent drawingFigure 2~3

AI summary

The invention relates to a computer-implemented method for optimizing welding parameters (SP1, SP2) for a welding control (1), comprising the steps of approximating (S2) at least one numerical value (W) of an objective function for the plurality of first welding parameters (SP1) for the welding control (1) using a trained machine learning algorithm (A1) applied to the plurality of first welding parameters (SP1), wherein the objective function represents at least one quality criterion of a weld in response to a weld produced by the plurality of first welding parameters (SP1), and optimizing (S4) the plurality of first welding parameters (SP1) for the welding control (1) by an optimization algorithm (A2).which, using the first data set (DS1) comprising the plurality of first welding parameters (SP1) for the welding control (1) and the approximate numerical value (W) of the objective function (F), calculates an optimized plurality of second welding parameters (SP2) for the welding control (1).