Neural Network Model Identification Research Articles

Neural network model identification control of dual-inertia system with a flexible load considering payload mass variation and nonlinear deformation

The dual-inertia system with a flexible load (DSFL) is a complex nonlinear model, which originates from the flexible load deformation and payload mass variation. In dynamic modeling, it is impossible to consider all the nonlinear factors, including high-order modes and inaccurate friction torque, which leads to the existence of uncertain components in the DSFL. Based on this, it is impossible to build a dynamic model of the DSFL that is completely consistent with the real physical model. This modeling errors increases the tracking errors of the DSFL. In this study, the neural networks are employed to recognize the lumped uncertain components of the DSFL dynamic model, so as to reduce the flexible load's tracking error. Firstly, based on the deformation accuracy comparison, a relatively high precision model simplification method is proposed, which neglects the second order mode. Next, the control law of the neural network identification control strategy is designed according to the stability theorem. Finally, through physical control experiments and simulation analysis, it is shown that this control strategy enhances the motion accuracy of the DSFL. The experimental results show that this control strategy predicted by the radial basis function (RBF) neural network can reduce the rotation angle error by 15.86%.

Application of Dynamic Process Neural Network Model Identification in Ethnic Dance Online Teaching System.

With the development of the times, education presents a new trend, but the teaching characteristics of dance classroom teaching cannot adapt to the current development trend. In this article, the author analyzes modern information technology, hoping to realize the teaching of folk dance on the Internet and provide a new model of online distance teaching for folk dance teaching. The author analyzes the current teaching problems in colleges and universities, and proposes to change the existing teaching situation based on dynamic process neural network model identification and artificial intelligence, and instead use online remote network ethnic dance teaching. Online distance education can enable flexible teaching of folk-dance courses, deeply dig into the theoretical basis of distance teaching, and use online distance network teaching to make teaching time more flexible, not only providing new teaching methods but also introducing new teaching concepts. Based on the traditional neural network model identification, a dynamic process neural network model identification is developed. This model is no longer subject to the input limitation of the traditional neural network model, the processing time is relaxed, and the advantages are more obvious. In this research, the author introduces dynamic process neural network model identification in time series data mining, and makes full use of artificial intelligence to deeply analyze the classification and prediction problems in the context of time series.

SWWS: A Smart Wildlife Warning Sign System

Every year in the US, millions of animals are run over by vehicles making wildlife vehicle collisions a real danger to both animals and human. In addition, road networks be-come abiotic barriers to wildlife migration between regions creating ripple effects on ecosystems. In this paper, a smart wildlife warning sign system (SWWS) is demonstrated, utilizing the technologies of Internet of Things, image recognition, data processing and visualization. This smart sign system is intended to prevent roadkill by warning drivers to slow down once sensors are triggered and simultaneously capture animal images via infrared cam-era. Data collection is conducted through local neural network model identification of wildlife images and saved along with metadata based on animal activity occurrence. Wildlife activity data can be exported wirelessly to cloud database to assist ecologists and government road agencies to investigate and analyze the wildlife activity and migration patterns over time.

UAV attitude Sensor Fault Detection Based On Fuzzy Logic and by Neural Network Model Identification

UAV attitude Sensor Fault Detection Based On Fuzzy Logic and by Neural Network Model Identification

Identification of Neural Network Model of Robot to Solve the Optimal Control Problem

To calculate the optimal control, a satisfactory mathematical model of the control object is required. Further, when implementing the calculated controls on a real object, the same model can be used in robot navigation to predict its position and correct sensor data, therefore, it is important that the model adequately reflects the dynamics of the object. Model derivation is often time-consuming and sometimes even impossible using traditional methods. In view of the increasing diversity and extremely complex nature of control objects, including the variety of modern robotic systems, the identification problem is becoming increasingly important, which allows you to build a mathematical model of the control object, having input and output data about the system. The identification of a nonlinear system is of particular interest, since most real systems have nonlinear dynamics. And if earlier the identification of the system model consisted in the selection of the optimal parameters for the selected structure, then the emergence of modern machine learning methods opens up broader prospects and allows you to automate the identification process itself. In this paper, a wheeled robot with a differential drive in the Gazebo simulation environment, which is currently the most popular software package for the development and simulation of robotic systems, is considered as a control object. The mathematical model of the robot is unknown in advance. The main problem is that the existing mathematical models do not correspond to the real dynamics of the robot in the simulator. The paper considers the solution to the problem of identifying a mathematical model of a control object using machine learning technique of the neural networks. A new mixed approach is proposed. It is based on the use of well-known simple models of the object and identification of unaccounted dynamic properties of the object using a neural network based on a training sample. To generate training data, a software package was written that automates the collection process using two ROS nodes. To train the neural network, the PyTorch framework was used and an open source software package was created. Further, the identified object model is used to calculate the optimal control. The results of the computational experiment demonstrate the adequacy and performance of the resulting model. The presented approach based on a combination of a well-known mathematical model and an additional identified neural network model allows using the advantages of the accumulated physical apparatus and increasing its efficiency and accuracy through the use of modern machine learning tools.

Low‐density polyethylene tubular reactor control using neural Wiener model predictive control

AbstractLow‐density polyethylene (LDPE) is a valued commodity plastic with versatile applications. However, controlling the LDPE polymerization reactor is a challenging task due to its nonlinear behavior and multivariable process and operates in a board operating region. This work aims to develop and explore the neural Wiener MPC (NWMPC) application in controlling the LDPE tubular reactor process. The motivation for NWMPC development originates from its advantage in terms of lower development effort, time, resource, and computational costs compared to nonlinear MPC (NMPC) using the first principle model (FPM). In order to develop the LDPE tubular reactor model, the Aspen Plus and Aspen Dynamic software are used to simulate the process in steady‐state and dynamic form. The neural Wiener (NW) model identification is performed using state space and neural network model identification using Matlab software. The identification result shows the ability of the NW model to identify the nonlinear LDPE tubular reactor process successfully. Furthermore, the NWMPC has outperformed the state space MPC (SSMPC) in controlling grade transition, feed pressure loss, feed impurity disturbance, and heat of polymerization change during the online closed‐loop performance tests. These results signify the ability of NWMPC to handle practical LDPE tubular reactor control scenarios.

Neural Network Model Identification and Advanced Control of a Membrane Biological Reactor

System identification with different input-output structures, for a membrane biological reactor (MBR), was performed using artificial neural networks (ANN) black-box modeling. The ANN models were able to capture the dynamic flux experimental literature data. Sensitivity analyses were applied on the ANN models to quantify the effects of variation in the process inputs (backwash pressure, vacuum pressure, backwash and vacuum time) on the process output (flux rate. Sensitivity analysis was applied on the developed NN in order to find the optimum backwash scheduling. The maximum flux was attained at around 165 (L/m2·day) that corresponded to an optimum vacuum-to-backwash time ratio of 10 minutes vacuum to 2 minutes backwash. Advanced control strategy using neuro-model predictive control (NN-MPC) methodology was applied to control the MBR system. The NN-MPC parameters were tuned to attain an optimum performance. The NN-MPC was efficient in tracking the flux set-point changes by adjusting vacuum-to-backwash time ratio within the operation constraints.

Global approach test improvement using a neural network model identification to characterise solar combisystem performances

Solar CombiSystems (SCSs) are very efficient systems for reducing conventional energy consumption of building but their thermal performances are strongly dependent on the environment where they are installed (type of climate and thermal quality of the building). Currently it is impossible to predict the energy savings generated by a SCS as there is no standard test to characterise SCS performances.Currently, the Short Cycle System Performance Test (SCSPT), based on a 12days test of the complete SCS on a semi-virtual test bench, is able to predict annual energy savings with a good accuracy, but the performance prediction is limited to only one environment (the building and the climate corresponding with the test).Based on the SCSPT procedure, this paper proposes an improvement of the method by identifying a global SCS model from the test data. Then, the identified model would be able to simulate the tested SCS in any environment and thus to characterise its performances.The proposed model to identify is a “grey box” model, mixing a “White Box” model composed of known physical equations and a “Black Box” model, which is an Artificial Neural Network (ANN). A complete process is developed to train and select a relevant global SCS model from such a test.This approach has been validated through numerical simulations of three detailed SCS models. Compared to those annual results, “Grey Box” SCS models trained from a twelve days sequence are able to predict energy consumption with a good accuracy for 27 different environments. An experimental application of this procedure has been used to characterise a real system.

Study of the Online Automatic Non-Destructive Detecting System of the Cracks in Vibrating Screen

The large-scale vibrating screen has been applied widely in coal industry and other industrial areas as a kind of important device. However, the lower crossbeam is the main carrying structure and is easily damaged, so it is regarded as a researched object in this thesis and it is tested under the load in the laboratory. Based on the test, acoustic emission wave signals can be gotten by modern acoustic emission testing technique. Then, the "wavelet packet - energy" from the characteristics of acquiring signals is used as neural network input vector. In Matlab6.5, neural network model identification is created and taking advantage of nonlinearity and the ability of learning and memory of Neural Network，this model is fixed by training structure of network with training samples. The work presented shows that acoustical emission signal processing and research on early fatigue fault diagnosis based on the wavelet and the neural network is viable.

Filtered gradient active fuzzy neural network noise control in an enclosure backed by a clamped plate

SUMMARYThe performance of conventional linear algorithms in active noise control applications deteriorates facing nonlinearities in the system mainly because of loudspeakers. On the other hand, fuzzy logic and neural networks are good candidates to overcome this drawback. In this paper, the acoustic attenuation of noise in a rectangular enclosure with a flexible panel and five rigid walls is presented both theoretically and experimentally using filtered gradient fuzzy neural network (FGFNN) error back propagation algorithm in which the secondary path effect is implemented in derivation of updating rules. Considering this effect in updating rules leads to faster convergence and stability of the active noise control system. On the other hand, the primary path in the investigated system comprises an identified nonlinear model of loudspeaker inside the aforementioned box, parameters of which vary with the input current. The loudspeaker is identified using series‐parallel neural network model identification method. As a comparison, the performance of filtered‐x least mean squares and FGFNN algorithms are compared. It is observed that FGFNN controller exhibits far better results in the presence of loudspeakers with nonlinear behavior in primary path.Copyright © 2012 John Wiley & Sons, Ltd.

Sequential Feature selection in a multi-objective optimization problem

AbstractThe design of air quality plans requires the development and the application of Decision Support System (DSS) to assess both the impact of emission reduction strategies on pollution indexes and the costs of such emission reductions. The problem can be formalized as a multi-objective mathematical program, integrating local pollutant-precursor models and the estimate of emission reduction costs. Both aspects present several complex elements. In particular the exposure index can be assessed by means of deterministic complex modelling systems with high computational costs that can't be used during the optimization procedure. The most valuable solution to this issue consists in the identification of neural network models linking precursor emissions and the air quality index starting from the results of a limited number of deterministic model simulations. One of the main points in the identification procedure is the selection of the input and consequently the neural network architecture. In this paper an interactive input selection technique is applied to identify the neural network linking the PM10 (particulate matter with size smaller than 10 μm) concentration to the emission of its precursor. The methodology has been applied to Northern Italy for PM10 exposure control. The results stress that high performance can be obtained with a very limited set of input data both in terms of neural network performances and optimization robustness.

A generalized procedure in designing recurrent neural network identification and control of time-varying-delayed nonlinear dynamic systems

In this study, a generalized procedure in identification and control of a class of time-varying-delayed nonlinear dynamic systems is developed. Under the framework, recurrent neural network is developed to accommodate the on-line identification, which the weights of the neural network are iteratively and adaptively updated through the model errors. Then indirect adaptive controller is designed based on the dichotomy principles and neural networks, which the controller output is designed as a neuron rather than an explicit input term against system states. It should be noticed that including implicit control variable in design is more challenging, but more generic in theory and practical in applications. To guarantee the correctness, rigorousness, generality of the developed results, Lyapunov stability theory is referred to prove the neural network model identification and the designed closed-loop control systems uniformly ultimately bounded stable. A number of bench mark tests are simulated to demonstrate the effectiveness and efficiency of the procedure and furthermore these could be the show cases for potential users to apply to their demanded tasks.

Model algorithmic control of grade change operations in paper mills

In this work, the Model Algorithmic Control (MAC) method is applied to control the grade change operations in paper mills. The neural network model for the grade change operations is identified first and the impulse model is extracted from the neural network model. Results of simulations for MAC control of grade change operations are compared with plant operation data. The major contribution of the present work is the application of MAC in the industrial plants based on the identification of neural network models. We can confirm that the proposed MAC method exhibits faster responses and less oscillatory behavior compared to the plant operation data in the grade change operations.

Adaptive Random Search Approach to Identification of Neural Network Model

Adaptive Random Search Approach to Identification of Neural Network Model