Building Material Board Quality Prediction Model Selection

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing mathematical models for predicting quality features of building material plates are often specific to a production system and material type, requiring laborious determination of process parameters and may deteriorate over time, necessitating frequent adjustments.

Innovation Solution

A procedure that divides trial data sets into training and test records, uses these to generate model candidates for mathematical models predicting quality features, and evaluates these models based on their prediction accuracy to determine if they meet a given quality criterion.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If mathematical models are created specifically for a production plant and panel type through trial and error, then prediction accuracy for quality characteristics is improved, but the complexity and time required for model development increases

Engineering Contradiction:
Improveprediction accuracyVSAvoidmodel development complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The system performs self-training by automatically selecting relevant process parameters and training the neural network model using historical process data and quality measurements from the production plant, eliminating the need for manual trial-and-error model development while achieving accurate predictions

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The system performs preliminary data collection and model training in advance using historical data, so that when the model is deployed for prediction, it is already prepared and configured, reducing the time and complexity of model development

Inventive Principle:
Principle #10Preliminary action

2Measurement precision

If mathematical models are created specifically for a production plant and panel type, then prediction accuracy is improved, but the adaptability to changing production conditions deteriorates

Engineering Contradiction:
Improveprediction accuracyVSAvoidadaptability to production changes
Core Design Contradiction:
Measurement precisionVSAdaptability or versatility

Solution Approach 1:

The neural network model is designed to be dynamically retrainable, allowing it to adapt to changing production conditions by incorporating new data over time, thus maintaining both accuracy and adaptability

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The system uses feedback from actual quality measurements to continuously improve the model through retraining, ensuring the model remains accurate and adaptable to changing production conditions

Inventive Principle:
Principle #23Feedback

3Reliability

If process parameters are determined through trial and error on a production plant, then the model fits the specific production conditions better, but the time and resources required for model development increase

Engineering Contradiction:
Improvemodel fit to production conditionsVSAvoidmodel development time
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The system automatically identifies and selects the most relevant process parameters from available production data using feature selection algorithms, eliminating the need for manual trial-and-error parameter determination while achieving reliable model fit

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The patent replaces the manual mechanical process of trial-and-error parameter tuning with automated computational algorithms for feature selection and model training, significantly reducing development time while maintaining reliability

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

Data Source

PatentEP4339715A9Method for generating a mathematical model for predicting at least one quality characteristic of a building material board
Publication Date: 2025.05.14 SIEMPELKAMP MASCHINEN UND ANLAGENBAU GMBH & CO KG
  • EP4339715A9 patent drawingFigure 1
  • EP4339715A9 patent drawingFigure 2
  • EP4339715A9 patent drawingFigure 3

AI summary

Disclosed is, among other things, a method comprising: dividing a plurality of sample datasets into at least one training dataset and at least one test dataset, wherein a sample dataset comprises quality parameter data representing at least one value for at least one corresponding quality characteristic of a building material panel sample, and process parameter data representing values ​​for a plurality of process parameters of a production process for manufacturing the building material panel sample; obtaining, based on the at least one training dataset for at least one process parameter of the at least one training dataset, at least one model candidate for a mathematical model for predicting at least one quality characteristic of a building material panel;Generating respective evaluation data that represent at least one evaluation parameter for the at least one model candidate, which represents for the at least one test data set a quality of prediction of the at least one quality characteristic by the at least one model candidate; keeping a model candidate available as a mathematical model for use in the manufacture of a building material panel, in a case where an evaluation parameter for the model candidate fulfills a predefined quality criterion.