Trained Back Propagation Neural Network Research Articles

Recognizing and Classifying Bone Ruptures of the Human Skeletal System Utilizing the Machine Learning Methodologies

Abstract: Rapidly expanding technologies appear every day in various disciplines, particularly medical ones. Nonetheless, there are still distinct outdated approaches that are still widely habituated, effective, and efficient. One of these techniques is the use of X-rays to pinpoint injured bones. However, occasionally the digit of fractures is minor and challenging to see. Procedures should be constructed that are efficient and intelligent. In this analysis, an artificial classification system that recognizes and classifies bone fractures is being designed. There are two major stages in the system that has been conceived. The images of the fractures are processed in the initial stage using eclectic image processing practices to determine their position and shapes. The classification phase tracks, in which a back propagation neural network training on the processed pictures before being put to the trial. The system has experimented with photographs of bone fractures, indicating a high rendition and classification speed.

Strike Velocity Prediction of Stick Blunt Instruments Based on Backpropagation Neural Network.

To analyze and predict the striking velocity range of stick blunt instruments in different populations, and to provide basic data for the biomechanical analysis of blunt force injuries in forensic identification. Based on the Photron FASTCAM SA3 high-speed camera, Photron FASTCAM Viewer 4.0 and SPSS 26.0 software, the tester's maximum striking velocity of stick blunt instruments and related factors were calculated and analyzed, and inputed to the backpropagation (BP) neural network for training. The trained and verified BP neural network was used as the prediction model. A total of 180 cases were tested and 470 pieces of data were measured. The maximum striking velocity range was 11.30-35.99 m/s. Among them, there were 122 female data, the maximum striking velocity range was 11.63-29.14 m/s; there were 348 male data, the maximum striking velocity range was 20.11-35.99 m/s. The maximum striking velocity of stick blunt instruments increased with the increase of weight and height, but there was no obvious increase trend in the male group; the maximum striking velocity decreased with age, but there was no obvious downward trend in the female group. The maximum striking velocity of stick blunt instruments has no significant correlation with the material and strike posture. The root mean square error (RMSE), the mean absolute error (MAE) and the coefficient of determination (R2) of the prediction results by using BP neural network were 2.16, 1.63 and 0.92, respectively. The prediction model of BP neural network can meet the demand of predicting the maximum striking velocity of different populations.

Combination of Feature Extractions for Classification of Coral Reef Fish Types Using Backpropagation Neural Network

Feature extraction is important to obtain information in digital images, where feature extraction results are used in the classification process. The success of a study to classify digital images is highly dependent on the selection of the feature extraction method used, from several studies providing a combination of feature extraction solutions to produce a more accurate classification. Classifying the types of marine fish is done by identifying fish based on special characteristics, and it can be through a description of the shape, fish body pattern, color, or other characteristics. This study aimed to classify coral reef fish species based on the characteristics contained in fish images using Backpropagation Neural Network (BPNN) method. Data used in this research was collected directly from Bunaken National Marine Park (BNMP) in Indonesia. The first stage was to extract shape features using the Geometric Invariant Moment (GIM) method, texture features using Gray Level Co-occurrence Matrix (GLCM) method, and color feature extraction using Hue Saturation Value (HSV) method. The third value of feature extraction was used as input for the next stage, namely the classification process using the BPNN method. The test results using 5-fold cross-validation found that the lowest test accuracy was 85%, the highest was 100%, and the average was 96%. This means that the intelligent model derived from the combination of the three feature extraction methods implemented in the BPNN training algorithm is very good for classifying coral reef fish.

Multiaxial fatigue life prediction method based on the back-propagation neural network

The study was devoted to developing a reliable multiaxial fatigue life prediction method with the effect of additional cyclic hardening considered. Based on the original assumptions of traditional critical plane methods, a new characteristic plane (subcritical plane) was defined to describe the particularity of additional cyclic hardening under non-proportional loading condition. On the new defined subcritical plane, a new multiaxial fatigue damage control parameter containing the effect of additional hardening was also built, by which the dynamic path of stress spindle, combining material property and loading environment, was fully analysed. In addition, a multiaxial fatigue life prediction back-propagation neural network (BPNN), as an alternative to the traditional physical model, was proposed to calculate the fatigue life under multiaxial loading. To calculate the multiaxial fatigue life of different materials, a more simplified combination parameter σb/E for different types of materials was chosen as input parameter in BPNN training. The availability of the proposed method was validated by reasonable correlations with experimental data of six alloy steel materials and two Non alloy steel materials under diverse loading paths.

Prediction of the Equivalent Steering Angle of a Front-Wheel, High-Clearance Paddy Field Management Machine

To solve the problem of poor steering consistency for each steering wheel of a four-wheel, independent-steering, high-clearance paddy field management machine, given that the true steering angle of the front wheel cannot be directly obtained through the left and right front wheels steering angle value, a BP (Back Propagation) neural network equivalent steering angle prediction method based on signal feature extraction is proposed in this paper, which can be used to obtain the equivalent steering angle of the front wheel. First, the kinematics model of the paddy field management machine was constructed with the prediction of the steering angle of the management machine as the object. The body was set up in two application environments of cement ground and paddy field, two moving speeds (25 cm/s, 50 cm/s) and three preset steering angles (0°, ±10°) to form six motion modes. The steering angle of the front wheel of the body, the three-axis acceleration of the body, and the angular velocity of the z-axis under twelve conditions were collected. Combining the collected data with the actual trajectory data of the differential BeiDou, the feasibility analysis of the equivalent steering angle prediction was carried out. With the aim of determining an appropriate combination of input factors for optimal equivalent steering angle prediction by the BP neural network, we used FFT (Fast Fourier Transform) and power spectrum conversion to extract and analyze the signal features of the airframe attitude data and obtain the frequency characteristics of the peak power point of the power spectrum, respectively. Through a factor combination test, the optimal BP neural network training factors were determined and, finally, we confirmed that the rotation angles of the front wheels and the z-axis angular velocity may be used as effective training factors. In our investigation on the test set, the mean square error of the equivalent steering angle was found to be less than 0.66°, demonstrating that our approach is effective for obtaining the true steering angle of the front wheel.

Classification and Detection of Bone Fracture Using Machine Learning

Abstract: Technologies that are rapidly growing are appearing every day in a variety of disciplines, particularly the medical one. However, there are still certain outdated methods that are still widely used, effective, and efficient. One of these methods is the use of X-rays to identify damaged bones . However, sometimes the number of fractures is insignificant and difficult to see. Systems should be created that are efficient and intelligent. In this study, an artificial classification system that can recognise and categorise bone fractures is being developed. There are two main steps in the system that has been designed. The photos of the fractures are processed in the first stage using various image processing techniques to identify their position and shapes. The classification phase follows, in which a back propagation neural network training on the processed images before being put to the test. The system was put to the test experimentally on various photographs of bone fractures, and the results indicate high performance as well as a classification rate.

Welding sequence optimization to reduce welding distortion based on coupled artificial neural network and swarm intelligence algorithm

This study aims to develop a welding sequence optimization (WSO) framework based on coupled artificial neural network (ANN) and swarm intelligence algorithm for minimizing welding distortion of thin-walled squared Al–Mg–Si alloy tube components. This framework is mainly composed of two critical computer programs. Firstly, a multilayer feedforward backpropagation neural network (BPNN) system was established to rapidly estimate residual distortion for an arbitrary welding sequence so that welding sequence can be optimized for achieving desired welding quality. For this purpose, a series of nonlinear thermo-elastic–plastic finite element (FE) simulations were conducted and verified with experiments to generate the input database of the neural network. Subsequently, a reliable BPNN model was successfully created and trained within an acceptable error. Secondly, a novel swarm intelligence algorithm, namely, bees algorithm (BA) was proposed to solve the complicated WSO problems. In this optimization process, the trained BPNN model was implanted into this proposed BA for computing the fitness value of arbitrary welding sequences. Moreover, welding experiments were also performed to confirm the performance of the proposed optimization method. Comparing the results from experimental measurements, FE simulations, and proposed WSO framework, it is demonstrated that this proposed BPNN-and-BA-based WSO framework can be successfully applied in practical engineering to obtain an optimal welding sequence for minimizing final welding distortion.

Inventory Management Optimization of Green Supply Chain Using IPSO-BPNN Algorithm under the Artificial Intelligence

This exploration is aimed at reducing the waste of resources in the supply chain inventory management and provide better services for green supply chain management. It mainly proposes a backpropagation neural network (BPNN) model based on improved particle swarm optimization (IPSO) (IPSO-BPNN) and applies it to inventory management prediction. First, the important technologies of green supply chain and intelligent supply chain are analyzed from the perspective of the ecological environment. Next, the particle swarm optimization (PSO) algorithm is optimized based on the adaptive improvement of the learning factor and the addition of the speed mutation operator. Then, it is applied to the learning and training of BPNN. Finally, the simulation experiment of the combination model is conducted. The application fields of the combination model are analyzed. The results show that a single BPNN model will produce large errors in the training process. The final error of BPNN using the traditional PSO algorithm is 0.0259, while the error of BPNN optimized by IPSO is 0.0163. The optimized combination model has higher accuracy, better performance, and the lowest error rate. The classification error rate of its training set and test set is 1.51 and 2.16, respectively. The mean square error of the training set is 0.0163 and that of the test set is 0.0229. Under 6 ~ 12 different hidden nodes, the daily measurement model error and monthly measurement model error are both low when the number of nodes is 11. Moreover, the training set is always better than the test set. Finally, the network structure of the combination model is determined as the structure of 6-11-1. This prediction module will provide purchase volume suggestions and inventory volume suggestions to provide a feasible direction for the green development of inventory management.

Research on Handwritten Chinese Character Recognition Based on BP Neural Network

The application of pattern recognition technology enables us to solve various human-computer interaction problems that were difficult to solve before. Handwritten Chinese character recognition, as a hot research object in image pattern recognition, has many applications in people’s daily life, and more and more scholars are beginning to study off-line handwritten Chinese character recognition. This paper mainly studies the recognition of handwritten Chinese characters by BP (Back Propagation) neural network. Establish a handwritten Chinese character recognition model based on BP neural network, and then verify the accuracy and feasibility of the neural network through GUI (Graphical User Interface) model established by Matlab. This paper mainly includes the following aspects: Firstly, the preprocessing process of handwritten Chinese character recognition in this paper is analyzed. Among them, image preprocessing mainly includes six processes: graying, binarization, smoothing and denoising, character segmentation, histogram equalization and normalization. Secondly, through the comparative selection of feature extraction methods for handwritten Chinese characters, and through the comparative analysis of the results of three different feature extraction methods, the most suitable feature extraction method for this paper is found. Finally, it is the application of BP neural network in handwritten Chinese character recognition. The establishment, training process and parameter selection of BP neural network are described in detail. The simulation software platform chosen in this paper is Matlab, and the sample images are used to train BP neural network to verify the feasibility of Chinese character recognition. Design the GUI interface of human-computer interaction based on Matlab, show the process and results of handwritten Chinese character recognition, and analyze the experimental results.

BPNN Based Indoor Fingerprinting Localization Algorithm Against Environmental Fluctuations

For mobile users, the ability to accurately acquire their own location is critical. By locating, mobile users can get information about their environment and access location-related services. RSS fingerprint-based indoor localization method collects a training database of measurement fingerprints and uses a machine learning classifier to determine a person’s location from a new fingerprint. However, as the environment changes over time due to furniture or other objects being moved, the new fingerprints diverge from those in the original database. Therefore, an RSS-difference based localization system is designed to deal with the above problem. This method combines back-propagation neural network (BPNN) and weighted K-Nearest Neighbor (WKNN) method to improve the fingerprint similarity based indoor location method (FSIL). We train BPNN in the off-line stage to obtain the optimal BPNN parameter settings. In the online stage, K nearest neighbor points are firstly selected based on the improved FSIL algorithm, and then the difference of signal strength values between the K nearest neighbor points and the target user is input into the BPNN network, to obtain the Euclidean distance between the K nearest neighbor points and the target user, and finally the WKNN algorithm is used to obtain the user’s ultimate location. Simulation experiments based on the LDPL model and the Wireless Insite software, as well as the test results based on the indoor localization dataset IPIN2016, show that the localization accuracy in complex indoor scenarios can be improved by at least 11% when using the method proposed in this paper.

A Novel Fault Diagnosis Method for Marine Blower with Vibration Signals

Abstract The vibration signals on marine blowers are non-linear and non-stationary. In addition, the equipment in marine engine room is numerous and affects each other, which makes it difficult to extract fault features of vibration signals in the time domain. This paper proposes a fault diagnosis method based on the combination of Ensemble Empirical Mode Decomposition (EEMD), an Autoregressive model (AR model) and the correlation coefficient method. Firstly, a series of Intrinsic Mode Function (IMF) components were obtained after the vibration signal was decomposed by EEMD. Secondly, effective IMF components were selected by the correlation coefficient method. AR models were established and the power spectrum was analysed. It was verified that blower failure can be accurately diagnosed. In addition, an intelligent diagnosis method was proposed based on the combination of EEMD energy and a Back Propagation Neural Network (BPNN), with a correlation coefficient method to get effective IMF components, and the energy components were calculated, normalised as a feature vector. Finally, the feature vector was sent to the BPNN for training and state recognition. The results indicated that the EEMD-BPNN intelligent fault diagnosis method is suitable for higly accurate fault diagnosis of marine blowers.

A Combined Prediction Model Composed of the GM (1,1) Model and the BP Neural Network for Major Road Traffic Accidents in China

This paper compared the expected accuracy of the gray GM (1, 1) model and the combined GMBP model using a data set for major road traffic accidents. A combined GMBP prediction model composed of the very first parameter gray model GM (1, 1) is able to make exact predictions for forecasting dreary type of processes, and BP (back-propagation) neural network for a major traffic accident is proposed to overcome the limitations of a single prediction model for a major traffic accident. The method first obtains predicted data using the gray GM (1, 1) model, then trains the BP neural network using the GM (1, 1) model’s predicted data and the original data as input and output data, respectively, and finally the trained BP neural network can be considered a combined GMBP prediction model. The predicted data of the digit of major traffic accidents, the digit of fatalities, and the digit of damages from 2008 to 2020 were obtained using the combined GMBP prediction model, and it is compared with the expected data of the single gray GM (1, 1) model. The grades showed that the exactitude of the combined GMBP prediction model was significantly higher than that of the single gray GM (1, 1) model. Finally, from 2021 to 2033, the combined GMBP prediction model is used to forecast the number of significant road traffic incidents, mortalities, and damages. The prediction results show that the number of accidents, faculties, and injuries is on the decline in the future years.

New Methods for Predicting Strain Demand of Arctic Gas Pipelines across Permafrost under Frost Heave Displacement

With increasing gas resource development in the Arctic region, gas pipeline installations in permafrost regions are becoming important. Frost heave of pipeline foundation soils may occur when a chilled gas pipeline passes through unfrozen areas with frost-susceptible soils. The stress and strain behaviors caused by the differential frost heave will directly affect the safety of the pipeline. A nonlinear finite element model (FEM) computing the mechanical responses of the buried gas pipeline subjected to frost heave load was established and successfully validated with the results of a large-scale indoor pipe-soil interaction experiment carried out in Caen in France. Utilizing C# language and object-oriented visual programming techniques, a new customized parametric strain calculation software was developed. The effects of pipe diameter, pipe wall thickness, pipe internal pressure, and peak soil resistance on the longitudinal strain of X60, X70, and X80 steel pipes have been investigated quantitatively. For the first time, a fitting semiempirical equation and trained backpropagation neural network (BPNN) for predicting pipeline strain demand subjected to frost heave load were proposed based on 2688 groups of FEM results. The comparison results have proved their high accuracy and lower running time cost. The proposed new methods can be applied in the strain-based pipeline design and safety evaluation of pipelines in service. It is in the hope of supplementing existing theory and identifying new approaches for arctic gas pipeline installations.

Application of Artificial Intelligence Elements and Multimedia Technology in the Optimization and Innovation of Teaching Mode of Animation Sound Production

Nowadays, with the rapid development of multimedia technology and computer information processing, the data of multimedia information presents explosive growth. At present, the method of using artificial recognition of sound materials is inefficient, and an automatic recognition and classification system of sound materials is needed. To improve the accuracy of sound recognition, two algorithm models are established to identify and compare the sound materials, which are the hidden Markov model (HMM) and back propagation neural network (BPNN) model. Firstly, HMM is established, and the sound material is randomly selected as the test sample. The comparison between the expected classification and the actual is tested, and the recognition rate of each classification is got. The final average recognition rate is 61%. The anti-interference characteristics of the training HMM are tested, and the identification rate of the training model is selected in 6 types of signal-to-noise ratio (SNR) environments. The recognition rate of the training model has an obvious downward trend with the decrease of the SNR. Secondly, the BPNN model is built, and 200 BPNN training experiments are conducted. The training model with the highest average recognition rate is selected as the experimental model. The average recognition rate of the final model is higher than 90%. The expression ability and stability of the trained model are simulated after the new sample is introduced, and the anti-interference performance of the model is tested in different environments of SNR. The results of performance test are good, and only the recognition rate of complex types of some sound sources decreased. Finally, the accuracy of the HMM in the experiment is not as high as that obtained by BPNN. Therefore, the training method of BPNN has a greater advantage in both recognition accuracy and recognition efficiency for the studied sound. It provides a reference for automatic recognition of sound materials.

Logging identification for diagenetic facies of tight sandstone reservoirs: A case study in the Lower Jurassic Ahe Formation, Kuqa Depression of Tarim Basin

The Ahe Formation is one of the significant hydrocarbon units in the Kuqa Depression of Tarim Basin, which is characterized by poor reservoir quality and strong heterogeneities due to the diagenetic effects. Thus, ‘diagenetic facies’ are introduced to explore high-quality reservoir layers. According to the reservoir spaces, main diagenetic minerals and diagenesis, four types of diagenetic facies are categorized: tightly compacted facies, dissolution facies, carbonated cemented facies and microfracture facies. Petrophysical characteristics of different diagenetic facies are summarized, which shows that the dissolution facies and microfracture facies have higher porosities, permeabilities and bigger pore radii. Logging characteristics indicate that logging values of four diagenetic facies are overlapped seriously, making it impossible to classify them by original logging crossplot. Principal component analysis (PCA), linear discriminant analysis (LDA) and back propagation neural networks (BPNN) are used to build diagenetic facies identification models, and the former two belong to the linear statistical algorithms, while the latter is a nonlinear algorithm. By comparing the identification accuracy of these three algorithms, principal components (PCs) crossplot cannot recognize diagenetic facies effectively. LDA can achieve automatic identification by establishing discriminant functions, but its identification accuracy (71.7%) is too low to satisfy the exploration requirements. However, the trained BPNN yield 80% accuracy, which is more advantageous than PCA and LDA. The blind test shows that the predicted results by trained BPNN have a very high coincidence with the thin section analysis and the logging interpretation results. Therefore, the BPNN as provided in this paper can be effectively used for diagenetic facies identification based on conventional well logs.

PREDICTING THE MENTAL STRESS LEVEL OF DRIVERS IN A BRAKING CAR PROCESS USING ARTIFICIAL INTELLIGENCE

Reducing the physical and mental weariness of drivers is significant in improving healthy and safe driving. This paper is aim to predict the stress level of drivers while braking in various conditions of the track. By discovering the drivers’ mental stress level, we are able to safely and comfortably adjust the distance in relation to the vehicle ahead. The initial step used was a study related to Artificial Intelligence (AI), Electroencephalogram (EEG), safe distance in braking, and the theory of mental stress. The data was collected by doing a direct measurement of drivers’stress levels using the EEG tool. The respondents were 5 parties around 30-50 years old who had experience in driving for> 5 years. The research asembled 400 pieces of data about braking including the data of the velocity before braking, track varieties (cityroad, rural road, residential road, and toll road), braking distance, stress level (EEG), and focus (EEG). The database constructed was used to input the machine learning (AI) – Back Propagation Neural Network (BPNN) in order to predict the drivers’ mental stress level. Referring to the data collection, each road type gave a different value of metal stress and focus. City road drivers used an average velocity of 23.24 Km/h with an average braking distance of 11.17 m which generated an average stress level of 53.44 and a focus value of 45.76.Under other conditions, city road drivers generated a 52.11 stress level, the rural road = 48.65, and 50.23 for the toll road. BPNN Training with 1 hidden layer, neuron = 17, ground transfer function, sigmoid linear, and optimation using Genetic Algorithm (GA) obtained the Mean Square Error (MSE) value = 0.00537. The road infrastructure, driving behavior, and emerging hazards in driving took part in increasing the stress level and concentration needs of the drivers. The conclusion may be drawn that the available data and the chosen BPNN structure were appropriate to be used in training and be utilized to predict drivers’ focus and mental stress level. This AI module is beneficial in inputting the data to the braking car safety system by considering those mental factors completing the existing technical factor considerations.

Piano Music Teaching under the Background of Artificial Intelligence

The work is aimed at solving the problems of easy trapping into local extremes and slow convergence speed of the traditional music teaching evaluation system on Backpropagation Neural Network (BPNN). The traditional note recognition methods are susceptible to high noise complexity. Firstly, the Levenberg Marquardt (LM) algorithm is used to optimize the BPNN; secondly, an improved endpoint detection algorithm is proposed by short-term energy difference, which can accurately identify the time value of each note in the piano playing audio. By the traditional frequency domain analysis method, a radical frequency extraction algorithm is proposed by the improved standard harmonic method, which can accurately identify the note’s pitch. Finally, a piano performance evaluation model by BPNN is implemented, and the model is implemented by the Musical Instrument Digital Interface (MIDI) system. This evaluation model can be used to correct the errors of students’ performances in the piano music teaching process and to perform overall evaluation, rhythm evaluation, and expressive evaluation. Teachers and students play minuet to collect experimental samples to train BPNN and test the performance of the evaluation model. The practical result shows that (1) after 3000 times of training, the neural network error is less than 0.01, and the network converges; (2) the evaluation results of the piano performance evaluation model designed are basically in line with the actual level of the performer and have specific feasibility; and (3) the optimized BPNN is used to correct errors during performances with an accuracy rate of 94.3%, which is 5.25% higher than the traditional method. The error correction accuracy rate for pitch is 92.9%, which is 5.21% higher than the traditional method. The optimized BPNN has significantly improved the error correction accuracy of the notes and pitches played by the player. The model can effectively help piano beginners correct errors and improve the accuracy and efficiency of the practice. The purpose of this study is to alleviate the scarcity of piano teachers, reduce the work intensity of piano teachers, realize automatic error correction and objective evaluation of playing, and provide necessary technical support for improving the efficiency of piano music teaching.

A Multisource Data Fusion Modeling Prediction Method for Operation Environment of High‐Speed Train

Providing accurate and reliable railway regional environmental data is a key consideration in operation control and dynamic dispatching of high‐speed train. However, there are problems of low reliability and high uncertainty in the single data processing of high‐speed train operating area environment. Therefore, this paper proposes a novel multisource sensor data fusion method based on a three‐level information fusion framework. Firstly, the feature of the same kind of sensor data is extracted by the Kalman Filter (KF) algorithm as the input of back propagation neural network (BPNN). Then input the sample site into the BPNN for training and recognition, the feature fusion of heterogeneous sensor data is carried out, the decision output of BPNN is obtained, the output results are normalized, and its output is used as the basic probability assignment of Dempster–Shafer (D‐S) evidence theory and synthesis rules. Finally, the decision fusion of multisource data is realized by D‐S evidence theory. The simulation results show that compared with the traditional single fusion algorithm, the algorithm improves the accuracy of the prediction of high‐speed train operation environment and reduces the MAPE from 13.82% to 7.455%, and the RMSE from 0.77 to 0.69, and meanwhile, increases the R2 from 0.87 to 0.97.

EVALUATING THE PERFORMANCE OF EMG BASED BACK PROPAGATION NEURAL NETWORK (BPNN) TIME DOMAIN FEATURES AND TIME ADVANCE FEATURES FOR KNEE JOINT ANGLE ESTIMATION

One of the challenges in lower limb rehabilitation robot design and control for assistance as needed robot is the interaction between the robot and the user. This can be overcome by quantifying the amount of assistance needed by the patient. Which attributed to the precision and accuracy of the estimated knee joint angle by the controller. In this research, the back propagation neural network (BPNN) was trained to estimate the amount of compensation needed by the patients who served as input to the robot. The Electromyogram (EMG) signals, time domain (TD) and time advance domain (TAD) were considered and root means square (RMS), mean absolute value (MAV), variance (VAR) and standard deviation (STD) features were extracted from each domain. Fourteen healthy subjects accepted to validate the experimental setup. However, the best eight subjects data were consider for this article. The features were used separately for training BPNN and the performance of each feature was evaluated. The EMG signals of the protagonist muscle (vastus lateralis) and antagonist muscles (semitendinosus) were used as the input signal to the BPNN after processing the signals. The electro-goniometer reading of the knee joint angle was used as the output to the network. The features were used separately for training BPNN and the performance of each feature was evaluated. The TAD RMSE values were significantly reduced compared to TD with 37.34% , 35.32% , 34.87% and 31.22% for RMS, MAV, VAR and STD features respectively. The result shows that the TAD features have better performance for estimating the knee joint angle.

Optimal Design of Regenerative Cooling Structure Based on Backpropagation Neural Network

Aiming at the problems of relying on traditional experience and high cost of schemes selection in the current research on regenerative cooling design, this paper establishes an optimal design model based on the backpropagation (BP) neural network method. In this paper, the cooling structure is parameterized to realize the segmented description. A BP neural network is constructed through the chamber performance simulation data sets to get the mathematical description between structural size parameters and optimization objectives. Taking the specified thrust chamber lifetime, safety margin of chamber wall, and upper limit of maximum wall temperature and pressure loss as constraints, the Globalsearch algorithm is applied to obtain the optimal design results of the regenerative cooling structural size parameters. The results show that the trained BP neural network model is able to predict the regenerative cooling performance parameters with the calculation accuracy similar to numerical simulation. The study of the space shuttle main engine–main combustion chamber indicates that under the same thrust chamber operating parameters and coolant inlet parameters, the optimized regenerative cooling channel reduces both the maximum wall temperature and the coolant pressure loss, and improves the overall cooling performance.