Sensor Data Evaluation Using Neural Network and Least Squares Regression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for evaluating sensor data, such as least squares regression and AI-based approaches, face challenges in determining initial parameter estimates and ensuring trustworthy results, especially in complex scenarios where iterative methods fail and AI behavior is unpredictable.
Innovation Solution
Combining artificial intelligence of a neural network with least squares regression, where the neural network is trained to predict parameter vectors for sensor data evaluation, providing initial estimates for least squares regression and verifying results with a model-based approach to ensure accuracy and robustness.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If iterative methods (Levenberg-Marquardt) are used for model inversion in least squares regression, then measurement precision can be improved, but the method requires relatively good initial parameter estimates and fails in complicated problems
Solution Approach 1:
The patent applies preliminary action by using a neural network to generate initial parameter estimates before performing least squares regression. The neural network is trained offline to predict plausible starting values based on the measurement vector, which then feed into the iterative optimization process. This preliminary estimation step eliminates the need for manual or heuristic initial value selection, allowing the least squares method to converge reliably even in complicated problems where traditional initial estimation methods fail.
2Productivity
If AI-based data evaluation (artificial neural networks) is used for sensor data evaluation, then productivity can be improved, but trustworthiness cannot be ensured and incorrect analysis cannot be performed easily
Solution Approach 1:
The patent merges two different evaluation approaches: a neural network for rapid initial parameter estimation and least squares regression for precise, verified parameter optimization. The neural network provides fast predictions that serve as starting points, while the subsequent least squares regression refines these estimates with mathematical rigor and convergence guarantees. This combination allows the system to maintain the speed benefits of AI while ensuring the reliability and trustworthiness of the final results through the systematic optimization process.
Solution Approach 2:
The least squares regression acts as an intermediary between the neural network's rapid but potentially unreliable estimates and the final trusted results. The regression process takes the neural network's predictions, validates them through iterative optimization with convergence criteria, and produces verified parameter estimates. This intermediary step bridges the gap between fast AI-based evaluation and reliable, trustworthy results, allowing both speed and accuracy to coexist.
3Ease of operation
If heuristic methods are used to determine starting values in least squares regression, then ease of operation is improved, but these methods fail in complicated problems
Solution Approach 1:
The patent replaces the mechanical/heuristic method of determining starting values with a data-driven neural network model. Instead of using rule-based or experience-based heuristic approaches that work for simple cases, the neural network learns complex patterns from training data and generalizes to complicated problems. This substitution maintains ease of operation (automatic starting value determination) while significantly improving reliability in complicated scenarios where heuristics fail.
Data Source
AI summary
The invention specifies an apparatus for evaluating sensor measured values (1.1), having: —a sensor (1), wherein a model function that is suitable for a least squares regression and definable by a parameter vector is provided for evaluating the sensor measured values (1.1) of the sensor (1), wherein at least one parameter of the parameter vector forms a sensor output signal (3), and —a computing and evaluation unit (2) that has a neural network (2.1), which estimates the parameter vector on the basis of actually ascertained sensor measured values (1.1), and a least squares regression module (2.2), wherein the neural network (2.1) is trained with parameter vectors and the associated sensor measured values, and that is set up: ∘—to use the trained neural network (2.1) to ascertain at least one parameter estimate vector for sensor measured values (1.1) measured using the sensor (1) as an input variable for the least squares regression module (2.2), ∘—if a convergence criterion is satisfied for the performance of the least squares regression, to terminate the least squares regression and ∘—to output the at least one parameter of the most recently ascertained parameter vector as sensor output signal (3). An associated automated method for evaluating sensor measured values and a use of the apparatus are likewise specified.