Production Parameter Optimization Using ML Material Flow Simulation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing operator guidance systems in production processes rely on predefined rules that are not optimally adapted to dynamic production conditions, leading to deviations and inefficiencies.

Innovation Solution

A device utilizing a machine learning model trained on simulation data from a digital twin of the production process to optimize production parameters and sequences, determining a key performance indicator (KPI) and adjusting production steps dynamically.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If rule-based approaches are used in operator guidance systems, then production processes can be monitored and intervened upon, but the rules are not optimally adapted to dynamic production conditions leading to inefficiencies

Engineering Contradiction:
Improveadaptability to production conditionsVSAvoidproduction efficiency
Core Design Contradiction:
Adaptability or versatilityVSProductivity

Solution Approach 1:

The system changes the parameter of decision-making from static predefined rules to dynamic machine learning predictions. The ML model continuously learns from simulation data and operational data, adapting production parameters (worker assignments, task priorities, sequence optimization) to current production conditions, thereby resolving the contradiction between adaptability and productivity

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The system performs preliminary actions by conducting extensive material flow simulations beforehand to train the machine learning model. This pre-training with simulated data enables the system to make optimized decisions in real-time without requiring complex rule sets, improving both adaptability and productivity

Inventive Principle:
Principle #10Preliminary action

2Measurement precision

If extensive material flow simulations are conducted to train the machine learning model, then production optimization accuracy is improved, but computational time and resources increase

Engineering Contradiction:
Improveoptimization accuracyVSAvoidtraining time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The system performs the computationally intensive material flow simulations in advance during an offline training phase. This preliminary action creates a trained machine learning model that can then make rapid predictions during actual production operations, resolving the contradiction between achieving high optimization accuracy and minimizing time loss during production

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The process is segmented into two distinct phases: an offline training phase where extensive simulations are conducted to build the ML model, and an online execution phase where the trained model provides rapid predictions. This segmentation allows computationally expensive operations to be performed when production time is not constrained, resolving the time-accuracy trade-off

Inventive Principle:
Principle #1Segmentation

Data Source

PatentEP4657341A1Optimisation of production
Publication Date: 2025.12.03 SIEMENS AG
  • EP4657341A1 patent drawingFigure 1
  • EP4657341A1 patent drawingFigure 2
  • EP4657341A1 patent drawing

AI summary

The invention relates to a device (100) for optimizing the production of a product, in particular for optimizing worker deployment in production, comprising: • a first interface (101) configured to read in a production parameter set for production, • a second interface (102) configured to read in a trained machine learning model, wherein the machine learning model is trained using simulation data to output a key figure and a sequence of production steps depending on a given production parameter set, wherein the simulation data are acquired by a plurality of computer-aided material flow simulations based on a computer-aided material flow model of production, wherein modified production parameter sets are used for each computer-aided material flow simulation (description: various fluctuations, different processing rules, etc.)...) are applied, and a key figure and a sequence of production steps are determined based on each executed material flow simulation, • an optimization module (103) which is configured to determine a key figure for the input production parameter set using the trained machine learning model, and to modify the input production parameter set in such a way that the key figure is optimized, and to determine a corresponding sequence of production steps for the parameter set modified in this way using the trained machine learning model, and • an output module (104) which is configured to output the production parameter set optimized in this way and the corresponding sequence of production steps.