Artificial neural network based soft estimator for estimation of transducer static nonlinearity

Area of Science:

• * Artificial Intelligence
• * Signal Processing
• * Transducer Technology

Background:

• * Existing Artificial Neural Network (ANN) models for transducer static nonlinearity estimation, such as Functional Link Artificial Neural Networks (FLANN) and single-layer linear ANNs, suffer from slow convergence.
• * Accurate modeling of transducer static nonlinearity is crucial for precise measurement and control systems.

Purpose of the Study:

• * To develop a novel ANN-based soft estimator for accurately estimating transducer static nonlinearity.
• * To address the slow convergence issue associated with existing ANN models.
• * To introduce a Polynomial-ANN model trained with the Levenberg-Marquardt (LM) learning algorithm.

Main Methods:

• * Development of a Polynomial-ANN model based on a single-layer feed-forward back-propagation ANN architecture.
• * Training the Polynomial-ANN using the Levenberg-Marquardt (LM) learning algorithm.
• * Implementation and evaluation of the proposed model for transducer static nonlinearity estimation.

Main Results:

• * The proposed Polynomial-ANN model demonstrated extremely fast convergence speed.
• * The model achieved increased accuracy in estimating transducer static nonlinearity compared to previous methods.
• * The LM learning algorithm proved highly effective in the training process, yielding stimulating convergence results.

Conclusions:

• * The Polynomial-ANN model offers a superior approach for estimating transducer static nonlinearity.
• * The LM learning algorithm significantly enhances the convergence speed and accuracy of the proposed neural modeling technique.
• * This advancement provides a more efficient and accurate method for transducer characterization.