Training Backpropagation Networks Research Articles

Modified Nonlinear Hysteresis Approach for a Tactile Sensor.

Soft tactile sensors based on piezoresistive materials have large-area sensing applications. However, their accuracy is often affected by hysteresis which poses a significant challenge during operation. This paper introduces a novel approach that employs a backpropagation (BP) neural network to address the hysteresis nonlinearity in conductive fiber-based tactile sensors. To assess the effectiveness of the proposed method, four sensor units were designed. These sensor units underwent force sequences to collect corresponding output resistance. A backpropagation network was trained using these sequences, thereby correcting the resistance values. The training process exhibited excellent convergence, effectively adjusting the network's parameters to minimize the error between predicted and actual resistance values. As a result, the trained BP network accurately predicted the output resistances. Several validation experiments were conducted to highlight the primary contribution of this research. The proposed method reduced the maximum hysteresis error from 24.2% of the sensor's full-scale output to 13.5%. This improvement established the approach as a promising solution for enhancing the accuracy of soft tactile sensors based on piezoresistive materials. By effectively mitigating hysteresis nonlinearity, the capabilities of soft tactile sensors in various applications can be enhanced. These sensors become more reliable and more efficient tools for the measurement and control of force, particularly in the fields of soft robotics and wearable technology. Consequently, their widespread applications extend to robotics, medical devices, consumer electronics, and gaming. Though the complete elimination of hysteresis in tactile sensors may not be feasible, the proposed method effectively modifies the hysteresis nonlinearity, leading to improved sensor output accuracy.

Research on electronic nose for compound malodor recognition combined with artificial neural network and linear discriminant analysis

The foul odor of foul gas has many harmful effects on the environment and human health. In order to accurately assess this impact, it is necessary to identify specific malodorous components and levels. In order to meet the qualitative and quantitative identification of the components of malodorous gas, an electronic nose system is developed in this paper. Both principal component analysis (PCA) and linear discriminant analysis (LDA) were used to reduce the dimensionality of the collected data. The reduced-dimensional data are combined with a support vector machine (SVM) and backpropagation (BP) neural network for classification and recognition to compare the recognition results. Regarding qualitative recognition, this paper selects the method of LDA combined with the BP neural network after comparison. Experiments show that the qualitative recognition rate of this method in this study can reach 100%, and the amount of data after LDA dimensionality reduction is small, which speeds up the pattern speed of recognition. Regarding quantitative identification, this paper proposes a prediction experiment through Partial least squares (PLS) and BP neural networks. The experiment shows that the average relative error of the trained BP network is within 6%. Finally, the experiment of quantitative analysis of malodorous compound gas by this system shows that the maximum relative error of this method is only 4.238%. This system has higher accuracy and faster recognition speed than traditional methods.

Evaluating the Vocal Music Teaching Using Backpropagation Neural Network

The vocal music teaching for evaluating performers is affected by multiple factors. Evaluators are greatly influenced by subjective factors in scoring outputs. The backpropagation (BP) neural network provides a novel technology that can theoretically simulate any nonlinear continuous function within a certain accuracy range. The backpropagation neural network is composed of adaptive feedforward learning network that is widely used in artificial intelligence (AI). In addition, the backpropagation neural network can simulate the nonlinear mapping composed of various factors. The novelty of the neural network is that it can model the nonlinear process without knowing the cause of the data, which can overcome the human subjective arbitrariness and make the evaluation outcomes. Furthermore, accurate and effective scoring systems can be designed using neural networks. In this paper, we establish a vocal music evaluation research system in order to objectivize each vocal music teaching evaluation index. To do so, we use the score vector as the input and obtain a reasonable and objective output score through the backpropagation neural network. Moreover, according to the characteristics of the backpropagation neural network, the factors of vocal music teaching evaluation are analyzed, and a backpropagation neural network model for vocal music teaching evaluation and evaluation is constructed. The experimental outcomes demonstrate that the trained backpropagation network can simulate a stable vocal music teaching evaluation research system. Furthermore, we observed that the backpropagation neural network can be well utilized for vocal music teaching evaluation research.

Adjustment of laser scattering focus with machine learning

The adjustment of the intensity of laser scattering focus with machine learning based on the back propagation (BP) neural network is proposed in this paper. Focus based on ‘laser scattering’ is more flexible than that based on a lens and imaging based on ‘laser scattering focus’ can break through the diffraction limit of lens imagingin order to obtain higher resolution. Therefore, knowing how to adjust the scattering focus is very important. In this paper, we used known data to train a BP network, and then used the trained BP network to realize the rapid and arbitrary adjustment of the intensity of the scattering focus. This method has potential application value in the field of ‘laser scattering focus’. In addition, with the rapid development of artificial intelligence, machine learning is increasingly being used in various fields, including physics. We hope that machine learning can also be applied to ‘laser scattering focus’. Therefore, in this paper, we conducted some preliminary investigations, which could provide reference for further research.

A Data-Driven Learning-Based Continuous-Time Estimation and Simulation Method for Energy Efficiency and Coulombic Efficiency of Lithium Ion Batteries

Lithium ion (Li-ion) batteries work as the basic energy storage components in modern railway systems, hence estimating and improving battery efficiency is a critical issue in optimizing the energy usage strategy. However, it is difficult to estimate the efficiency of lithium ion batteries accurately since it varies continuously under working conditions and is unmeasurable via experiments. This paper offers a learning-based simulation method that employs experimental data to estimate the continuous-time energy efficiency and coulombic efficiency of lithium ion batteries, taking lithium titanate batteries as an example. The state of charge (SOC) regions and discharge current rates are considered as the main variables that may affect the efficiencies. Over eight million empirical datasets are collected during a series of experiments performed to investigate the efficiency variation. A back propagation (BP) neural network efficiency estimation and simulation model is proposed to estimate the continuous-time energy efficiency and coulombic efficiency. The empirical data collected in the experiments are used to train the BP network model, which reveals a test error of 10−4. With the input of continuous SOC regions and discharge currents, continuous-time efficiency can be estimated by the trained BP network model. The estimated and simulated result is proven to be consistent with the experimental results.

A novel optimization method for maintaining aerodynamic performances in noise reduction design

Noise reduction has now become an indispensable part of civil aircraft design, however, it is difficult to reduce noise and maintain aerodynamic performances simultaneously. In this case, a novel method based on artificial neural network is introduced. An established database lays foundation for and serves as the training sample of a back propagation (BP) artificial neural network, which uses confidence coefficient reasoning method for optimization later on. Then the trained BP network is employed to select the most satisfactory configuration for validating computations. In order to verify the design method, a process of slat cove filler (SCF) design for EET HLD is presented. Unsteady simulations based on DES method are carried out to validate aerodynamic performances of both the baseline and design configurations, while noise reduction effect is verified by a hybrid method which combines unsteady DES method with acoustic analogy theory. The numerical results indicate a significant noise reduction at the given observation point while the aerodynamic performances are still maintained.

Application of artificial intelligence for the prediction of delamination in drilling GFRP composites

Glass fibre reinforced composite materials are used in different engineering fields due to their excellent properties. Precision hole making drilling operation is essentially needed for this composite to perform assembly. Delamination is one of the important problems associated with composite drilling process. Delamination after drilling leads to rejection and imposes heavy loss in production. In this paper, a neural network based on back-propagation (BP) algorithm with two hidden layers are used for the modelling of delamination factor in drilling glass fibre reinforced plastic (GFRP) composites. The input-output data sets required for training are obtained from drilling experimentation. Fifty-four sets of data were used for training and 18 sets of data were used for testing. Residuals were used for checking the performance. The result shows that the well trained BP network model can precisely predict the delamination in drilling of GFRP composites.

Function Approximation and Documentation of Sampling Data Using Artificial Neural Networks

Biodiversity studies in ecology often begin with the fitting and documentation of sampling data. This study is conducted to make function approximation on sampling data and to document the sampling information using artificial neural network algorithms, based on the invertebrate data sampled in the irrigated rice field. Three types of sampling data, i.e., the curve species richness vs. the sample size, the curve rarefaction, and the curve mean abundance of newly sampled species vs.the sample size, are fitted and documented using BP (Backpropagation) network and RBF (Radial Basis Function) network. As the comparisons, The Arrhenius model, and rarefaction model, and power function are tested for their ability to fit these data. The results show that the BP network and RBF network fit the data better than these models with smaller errors. BP network and RBF network can fit non-linear functions (sampling data) with specified accuracy and don't require mathematical assumptions. In addition to the interpolation, BP network is used to extrapolate the functions and the asymptote of the sampling data can be drawn. BP network cost a longer time to train the network and the results are always less stable compared to the RBF network. RBF network require more neurons to fit functions and generally it may not be used to extrapolate the functions. The mathematical function for sampling data can be exactly fitted using artificial neural network algorithms by adjusting the desired accuracy and maximum iterations. The total numbers of functional species of invertebrates in the tropical irrigated rice field are extrapolated as 140 to 149 using trained BP network, which are similar to the observed richness.

Measurement system using ultraviolet and multiband infrared technology for identifying fire behavior

Multisensor-based fire alarm systems have been developed as an automatic fire detection technology. However, only two or three fire signatures do not suffice to recognize a fire. Conventional fire detectors must touch sufficient targets (heat or smoke) so that they respond slowly to the alarm. This article proposes several approaches that use the radiation-sensing technologies to detect fire behaviors. Four features, temperature, hot CO2 gas, flicker, and the UV energy of the flame are adopted in the system to analyze the fire traits. Four kinds of sensors and one IR transducer are used successfully to evaluate these features. Experimental results show that the use of two pyroelectric sensors can accurately “see” the appearance of combustion by detecting the CO2 emission instead of the particle density of smoke. The combination of a blue and an UV sensor is used effectively to pickup the UV component of the flame. The information is presented on a personal computer using LABVIEW software and a data-acquisition interface. Although this article focuses on techniques for measuring IR and UV energies, in off-line work, a trained back-propagation network verifies that the measuring system is effective in collecting the information on fire behaviors. Results of this study provide a valuable reference for designers in developing a novel fire detector.

Modeling parison formation in extrusion blow molding by neural networks

AbstractA series of experiments were carried out on the parison formation stage in extrusion blow molding of high‐density polyethylene (HDPE) under different die temperature, extrusion flow rate, and parison length. The drop time of parison when it reached a given length and its swells, including the diameter, thickness, and area swells, were determined by analyzing its video images. Two back‐propagation (BP) artificial neural network models, one for predicting the length evolution of parison with its drop time, the other predicting the swells along the parison, were constructed based on the experimental data. Some modifications to the original BP algorithm were carried out to speed it up. The comparison of the predicted parison swells using the trained BP network models with the experimentally determined ones showed quite a good agreement between the two. The sum of squared error for the predictions is within 0.001. The prediction of the parison diameter and thickness distributions can be made online at any parison length or any parison drop time within a given range using the trained models. The predicted parison swells were analyzed. © 2005 Wiley Periodicals, Inc. J Appl Polym Sci 96: 2230–2239, 2005

A modified learning strategy for neural networks to support spectroscopic ellipsometric data evaluation

It has been shown that worst-case learning, a slightly modified strategy in backpropagation network (BPN) training, results in constrained maximal error at the expense of slightly increased root mean squared error (RMSE) using BPN in spectroscopic ellipsometry (SE). Traditionally the evaluation of SE needs an anticipated multifilm optical model possessing initial parameters close enough to the targeted ones, because a non-linear gradient descent algorithm is performed on the basis of the selected initial guesses. If the initial guesses are not close enough, erroneous results may appear when the algorithm inadvertently falls into the trap of a local minimum. Backpropagation-trained feedforward neural networks prove to be able to give initial estimations that are sufficient for avoiding problems in most cases. Nevertheless, such a network trained with a traditional backpropagation algorithm cannot guarantee error free operation at high confidence levels as it aims for minimised RMSE. The suggested worst-case method gives high confidence levels of convergence from the neural network performed pre-evaluation results to the actual physical parameters. We present two types of problems as demonstrations. Firstly, Separation by Implantation of Oxygen (SIMOX) structural models were used. The maximal errors of structural prediction have fallen from 37% to 7.8% of the entire training region when using a worst-case strategy instead of a traditional backpropagation. Meanwhile RMSE values rose from 3.2 to 4.61%. Secondly, porous silicon layer (PSL) investigations show the effectiveness of using trained BPNs successively. Using a second BPN trained in a smaller region after the first one, we get a more precise approximation for the parameters. Further tips are given for exploiting the advantages of the worst-case learning method, to facilitate creating a more effective and safe apparatus for real-time material parameter determination.

A comparative study of optimization techniques for backpropagation

A large variety of optimization techniques has been used to train backpropagation networks. The most important distinction between these methods rests on whether they employ batch-mode or sequential optimization. It is shown that batch-mode methods are currently more advanced; however, problems with very large training sets or in non-stationary environments will require the development of successful sequential approaches.