Technical System Calibration Under Random Operating Conditions

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional calibration methods for technical systems are not robust against real-world operational influences, such as environmental variables and random factors, leading to potential deviations in output variables during actual operation.

Innovation Solution

A method that involves executing a load cycle multiple times under the influence of random variables, defining a cost function with a risk measure, and optimizing control variables to account for stochastic influences, ensuring the technical system is calibrated to perform optimally across various operating conditions.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If conventional calibration methods are used with predetermined operating points, then calibration can be completed with a single load cycle, but the calibration is not robust against real-world operational influences and environmental variables

Engineering Contradiction:
Improverobustness of calibrationVSAvoidcomplexity of calibration method
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent applies preliminary action by generating multiple synthetic load cycles with predetermined random influencing variables before the actual calibration process. These synthetic cycles pre-incorporate environmental variations, allowing the calibration to be performed on a representative dataset that already accounts for real-world uncertainties, thus improving robustness without requiring complex real-time adjustments during calibration.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent creates synthetic copies of real-world operating conditions through simulated load cycles with random influencing variables. Instead of calibrating on a single actual load cycle, multiple synthetic copies are generated that replicate the statistical characteristics of real operational variability, allowing the calibration to learn from diverse scenarios without requiring physically diverse test setups.

Inventive Principle:
Principle #26Copying

2Reliability

If multiple load cycles are executed with random influencing variables, then robustness against operational variability is improved, but the computational effort and calibration time increase

Engineering Contradiction:
Improverobustness of calibrationVSAvoidcalibration time
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The patent replaces physical execution of multiple load cycles with computational generation of synthetic load cycles. Instead of physically running the technical system through multiple test cycles with varying environmental conditions, the method uses computer-generated random variables to simulate these conditions mathematically, dramatically reducing calibration time while maintaining the benefits of multiple-scenario calibration.

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

Solution Approach 2:

The patent changes the state of the calibration process by transforming physical load cycle execution into parameter-based synthetic generation. By varying random influencing variables as parameters in a computational model rather than physically adjusting environmental conditions, the method achieves diverse calibration scenarios efficiently without the time cost of physical reconfiguration.

Inventive Principle:
Principle #35Parameter changes

3Productivity

If calibration is performed on a single load cycle, then the process is simple and fast, but deviations in output variables occur during actual operation due to unaccounted random influences

Engineering Contradiction:
Improvecalibration efficiencyVSAvoidaccuracy of output variables
Core Design Contradiction:
ProductivityVSMeasurement precision

Solution Approach 1:

The patent applies preliminary action by pre-generating multiple synthetic load cycles that incorporate random influencing variables before calibration. This preliminary preparation creates a comprehensive training dataset that accounts for operational variability, allowing the calibration to achieve higher precision on actual operations without sacrificing the efficiency of a single-pass calibration process.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent creates synthetic copies of diverse operating conditions through computational simulation. These copies provide the calibration process with exposure to multiple scenarios including rare edge cases, improving the accuracy and generalization of the calibrated model without requiring physically diverse test cycles that would consume excessive time.

Inventive Principle:
Principle #26Copying

Data Source

PatentEP3871052B1Method for calibrating a technical system
Publication Date: 2024.02.07 AVL LIST GMBH
  • EP3871052B1 patent drawingFigure 1~2
  • EP3871052B1 patent drawingFigure 3~4
  • EP3871052B1 patent drawingFigure 5a~5e

AI summary

In order to calibrate a technical system (1) against stochastic influences during real operation of the technical system (1), according to the invention a duty cycle is carried out several times under the influence of at least one random influencing variable (Z), wherein a random sequence (Tm) of the number i of operating points (xi) is produced during each implementation of the duty cycle, a cost function c(θ,Z) is defined which contains a target function y having a model f for an output variable of the technical system (1), wherein the model f is dependent on the number of control variables (θ) of the technical system (1) and on the number i of random operating points (xi) of the technical system, so that the value of the cost function c(θ,Z) for each implementation of the duty cycle is itself a random variable which has a probability distribution P(c(θ,Z)), a risk measure ρ of the probability distribution P(c(θ,Z)) is defined using which the probability distribution P(c(θ,Z)) is mapped onto a scalar variable, and the risk measure ρ is optimised by varying the number of control variables (0) in order to obtain the optimal control variables (θopt) for calibration.