人工神经网络技术在CBN砂轮磨削表面粗糙度研究中的应用

针对CBN砂轮磨削 ,采用人工神经网络方法建立由磨削用量确定表面粗糙度的预测模型。计算结果证明 ,所建立的人工神经网络模型可很好地描述砂轮速度、砂轮进给速度、工件转速对磨削表面粗糙度的影响。预测结果具有良好的精度并得到了验证试验的检验。通过本模型 ,利用有限的试验数据可得出整个工作范围内表面粗糙度的预测值 ,可大量减少试验费用     

A smart force platform using artificial neural networks

The human body may interact with structures and these interactions are developed through the application of contact forces, for instance when walking. The aim of this paper is to propose a new methodology using Artificial Neural Network (ANN) for calibrating a force platform in order to reduce the uncertainties in the values of estimated vertical Ground Reaction Force and the positioning of the applied force in the human gait. Force platforms have been used to evaluate the pattern of human applied forces and to fit models for the interaction between pedestrians and structures. Linear relation assumptions between input and output are common in traditional Least Mean Square methods used in calibration. Some discrepancies due to nonlinearities in the experimental setup (looseness, wear, support settlements, electromagnetic noise, etc.) may harm the overall fitting. Literature has shown that nonlinear models, like ANN, can better handle this. During the calibration, the input data to the ANN were the reference voltages applied to the Wheatstone bridge, while the output data were the values of the standard weights applied in the force platform in defined sites. Supervised training based on k-fold cross validation was used to check the ANN generalization. The use of ANN shows significant improvements for the measured variables, leading to better results for predicted values with low uncertainty when compared to the results of a simple traditional calibration using Least Mean Squares.     

人工神经网络在智能机械设计中的应用

介绍了智能ＣＡＤ的概念和发展,分析了人工神经网络（ＡＮＮ）的特点,针对目前机械设计专家系统存在的问题,提出将ＡＮＮ应用到专家系统的设计中,是进行智能ＣＡＤ的一条有效途径。介绍了ＡＮＮ在概念设计、设计过程中形象思维的模拟、知识的获取和表示、回溯问题的模拟等方面的应用。     

Prediction of brake friction materials recovery performance using artificial neural networks

The brake friction materials in an automotive brake system are considered as one of the key components for overall braking performance of a vehicle. The sensitivity of friction material performance and accordingly brake performance, versus different operating regimes, has always been an important aspect of its functioning. In this paper, the influences not only on the brake operation conditions but also on the formulation and manufacturing conditions of friction materials have been investigated regarding friction materials recovery performance by means of artificial neural networks. A new neural network model of friction material recovery performance, trained by the Bayesian Regulation algorithm, has been developed.     

Application of artificial neural networks for prediction of coercivity of highly ordered cobalt nanowires synthesized by pulse electrodeposition

This study aims to predict the coercivity of cobalt nanowires fabricated by Alternating Current (AC) pulse. Coercivity is one of the most important properties of magnetic materials and its value shows the needed magnetic field in a way that magnetization of system is decreased to zero. There are many parameters such as pH of solution, oxidative and reductive times, oxidative and reductive voltages, interval between pulses (off-time), and concentration of deposition solution that have direct effect on materials magnetic properties of. Change of initial conditions to obtain the best results is very time consuming, therefore employing a method which can save both the time and cost is necessary. Hence, it this study Artificial Neural Network (ANN), which has numerous applications and has attracted many attentions in various fields, was applied. Through this study, an ANN was designed to present a template that is capable for predicting output data (coercivity) according to input data (pH, oxidative and reductive times, oxidative and reductive voltages, and off-time). Besides, in this research, the results for pH = 4 and 6 were investigated and the effect of off-time as well as the deposition time on coercivity were studied.     

Indirect model for roughness in rough honing processes based on artificial neural networks

In the present paper an indirect model based on neural networks is presented for modelling the rough honing process. It allows obtaining values to be set for different process variables (linear speed, tangential speed, pressure of abrasive stones, grain size of abrasive and density of abrasive) as a function of required average roughness Ra. A multilayer perceptron (feedforward) with a backpropagation (BP) training system was used for defining neural networks. Several configurations were tested with different number of layers, number of neurons and type of transfer function. Best configuration for the network was searched by means of two different methods, trial and error and Taguchi design of experiments (DOE). Once best configuration was found, a network was defined by means of trial and error method for roughness parameters related to Abbott–Firestone curve, Rk, Rpk and Rvk.     

Retrieval of machining information from feature patterns using artificial neural networks

The integration of design and manufacturing has been the subject of much debate and discussion over a long period of time. Recognition of feature patterns and the retrieval of necessary machining information from those patterns play vital roles in this process of integration, as they facilitate the selection of the necessary manufacturing parameters required to transform the designed product into a final physical entity. Although the problem of recognising features from a solid model has been exclusively studied, most existing product models are expressed as engineering drawings. Moreover, the solid model can only provide complete 3D topological and geometrical data and some of the essential machining information cannot be retrieved. In this paper, an approach for defining engineering features, like slots, steps and circular pockets is proposed using binary strings. Two artificial neural networks, one for slots and steps and the other for circular pockets, are designed and developed. These neural networks take the binary strings as inputs and give the relevant machining information as outputs. The networks are trained with non-interacting features and after training, those will become capable of providing the necessary machining information for both non-interacting and interacting features in the domains of slots, steps and circular pockets. This novel approach can further be extended to other features for retrieving relevant machining information and thus facilitating the effective integration of design and manufacturing.     

Improvement of the five-hole probe calibration using artificial neural networks

In the present study, the artificial neural networks (ANNs) technique was implemented to link non-dimensional pressure coefficients and flow characteristics to calibrate a five-hole probe. The experimental data of this work were obtained from a subsonic open-circuit wind tunnel at the velocity of 10 m/s. Here, the efficiency of ANNs was compared with two conventional data reduction methods, including linear interpolation technique and 5^th-order polynomial surface fit algorithm. Based on the statistical parameters of calibration data, it was concluded that the radial basis function (RBF) algorithm was more accurate and had more flexibility compared to the multi-layer perceptron (MLP) regression algorithm, the linear interpolation and 5^th-order polynomial methods. In the RBF method, the mean absolute errors of 0.11, 0.64, 0.02 and 0.03 were achieved for α, β, Cp_t and Cp_s , respectively. Furthermore, the effects of training data reduction and data selection on the performance of RBF were studied. The accuracy of the proposed RBF method was analyzed at different α angles and for random test data. Finally, the influence of increasing number of test data on the efficiency of calculated RBF method was evaluated.     

Process modelling and optimisation using artificial neural networks and gradient search method

Process modelling refers to the development of a process model that serves to provide the input-output relationship of a process, while process optimisation provides the optimum operating conditions of a process for a high-yield, low cost and robust operation. Normally, process modelling is a starting point of process optimisation. In this paper, a method of integrating artificial neural networks with a gradient search method for process modelling and optimisation is presented. Artificial neural networks are used to develop process models while a gradient search method is used in process optimisation. Application of the method to the modelling and optimisation of epoxy dispensing for microchip encapsulation is described. Results of the validation tests indicate that good quality of encapsulation can be obtained based on the proposed method.     

Parameter optimization of a multi-response process for lead frame manufacturing by employing artificial neural networks

This study proposes an effective means of applying a neural network approach to parameter optimization for a multi-response problem. No matter what type of experimental designs are being employed, the proposed approach can be directly applied. In addition, the design factors with level settings or with continuous values can be also solved with the proposed approach. Not only can parameter optimization be achieved, but the effects of the control factors reacting on a multi-response system can also be simultaneously determined. An illustrative example given courtesy of a lead frame manufacturer in Taiwan is employed to demonstrate the effectiveness of the proposed approach .     

预测15钢正挤压变形力的人工神经网络模型

针对现行挤压变形力计算的局限性 ,采用四层BP网络建立 15钢正挤压变形力预测的神经网络模型 ,其预测值与实验值吻合良好 ,为挤压力预测提供了新的途径。     

Predicting modulus of rupture (MOR) and modulus of elasticity (MOE) of heat treated woods by artificial neural networks

In this study, MOR and MOE of the heat-treated wood were predicted by artificial neural networks (ANNs). For this purpose, samples were prepared from beech wood (Fagus orientalis Lipsky.) and spruce wood (Picea orientalis (L.) Link.). The samples were exposed to heat treatment at varying temperatures (125, 150, 175 and 200 °C) for varying durations (3, 5, 7 and 9 h). According to the results, the mean absolute percentage errors (MAPE) were determined as 0.74%, 1.01% and 1.04% in prediction of MOR values, and 1.14%, 2.21% and 2.13%, in prediction of MOE values for training, validation and testing data sets, respectively. In the prediction of MOR and MOE, values of R² were obtained greater than 0.99 for all data sets with the proposed ANN models. The results show that ANN can be used successfully for predicting MOR and MOE of heat-treated wood.     

Condition monitoring of planetary gearbox by hardware implementation of artificial neural networks

The hazards of planetary gearboxes’ failures are the most crucial in the machinery which directly influence human safety like aircrafts. But also in an industry their damages can cause the large economic losses. Planetary gearboxes are used in wind turbines which operate in non-stationary conditions and are exposed to extreme events. Also bucket-wheel excavators are equipped with high-power gearboxes that are exposed to shocks. Continuous monitoring of their condition is crucial in view of early failures, and to ensure safety of exploitation. Artificial neural networks allow for a quick and effective association of the symptoms with the condition of the machine. Extensive research shows that neural networks can be successfully used to recognize gearboxes’ failures; they allow for detection of new failures which were not known at the time of training and can be applied for identification of failures in variable-speed applications. In a majority of the studies conducted so far neural networks were implemented in the software, but for dedicated engineering applications the hardware implementation is being used increasingly, due to high efficiency, flexibility and resistant to harsh environmental conditions. In this paper, a hardware implementation of an artificial neural network designed for condition monitoring of a planetary gearbox is presented. The implementation was done on a Field Programmable Gate Array (FPGA). It is characterized by much higher efficiency and stability than the software one. To assess condition of a gearbox working in non-stationary conditions and for chosen failure modes, a signal pre-processing algorithm based on filtration and estimation of statistics from the vibration signal was used. Additionally, the rewards-punishments training process was improved for a selected neural network, which is based on a Learning Vector Quantization (LVQ) algorithm. Presented classifier can be used as an independent diagnostic system or can be combined with traditional data acquisition systems using FPGAs.     

Identifying three-phase induction motor faults using artificial neural networks

This paper presents an artificial neural network (ANN) based technique to identify faults in a three-phase induction motor. The main types of faults considered are overload, single phasing, unbalanced supply voltage, locked rotor, ground fault, over-voltage and under-voltage. Three-phase currents and voltages from the induction motor are used in the proposed approach. A feedforward layered neural network structure is used. The network is trained using the backpropagation algorithm. The trained network is tested with simulated fault current and voltage data. Fault detection is attempted in the no fault to fault transition period. Off-line testing results on a 3 HP induction motor model show that the proposed ANN based method is effective in identifying various types of faults.     

Time domain modeling of cup anemometers using artificial neural networks

Under rapidly fluctuating wind speed, high inertia cup anemometers have a tendency to overspeed. The main objective of this article is to develop an inverse time domain model that could be used in real-time during cup anemometer operation to minimize the so-called “u-error”. A model proposed by Kristensen and an artificial neural network (ANN) direct model were first investigated to simulate the dynamic behavior of a heated cup anemometer with relatively high rotor inertia. Once the anemometer behavior was known, several virtual inputs were generated and the direct model was used to predict the instrument behavior. These models were built to emphasis the non-linear relationship between the free stream fluctuating wind and the wind speed measured by the anemometer. A semi-empirical inverse model derived from Kristensen's model was then studied and an ANN inverse model was suggested in order to minimize the so-called u-error. A methodology is proposed to gather the appropriate data to create both the direct and inverse model using an artificial neural network. The output of each model was compared with experimental data for validation and good agreement was found between the ANN models and the experimental data used for validation.     

Application of artificial neural networks instead of the orifice plate discharge coefficient

Differential pressure flowmeters are very often used in many industries. Therefore, the improvement of this method of flow measurement is an important task of flow measurement and instrumentation. One of the important characteristics of differential pressure flowmeters is the discharge coefficient of the flow transducers. A large number of studies and publications were devoted to modeling this coefficient. Therefore, in the framework of this research, this coefficient is simulated using artificial neural networks. The neural representation of this characteristic is made in the form of a multilayer perceptron. In this paper, we replace the traditional equation for the discharge coefficient with an artificial neural network. The advantages and disadvantages of such application of neural networks as discharge coefficients are discussed. The analysis of the results of gas flow measurement, where the neural network is used instead of the traditional equation, is presented. The estimation of flow rate measurement errors with such an approach is made; the error of calculation of the discharge coefficient is estimated.     

Analysis of hydraulic characteristics for hollow semi-circular weirs using artificial neural networks

Weirs are small overflow dams used to alter and raise water flow upstream and regulate or spill water downstream watercourses and rivers. This paper presents the application of artificial neural network (ANN) to determine the discharge coefficient (Cd) for a hollow semi-circular crested weirs. Eighty five experiments were performed in a horizontal rectangular channel of 10 m length, 0.3 m width and 0.45 m depth for a wide range of discharge. The results of examination for discharge coefficient were yielded by using multiple regression equation based on dimensional analysis. Then, the results obtained were also compared using ANN techniques. A multilayer perceptron MLP algorithm FFBP network was developed. The optimal configuration of ANN was [2,10,1] which gave mean square error (MSE) and correlation coefficient (R) of 0.0011 and 0.91, respectively. Performances of ANN model reveal that the Cd could be better estimated by the ANN technique in comparison with Cd obtained using statistical approach.     

Three-phase flow meters based on X-rays and artificial neural network to measure the flow compositions

The methodology presented in this study is based on a 149.5 keV X-ray beam and two planar germanium detectors for X-ray transmission and scattering measurements for prediction of volume fractions in a three-phase system. Fluid volume fractions have been modeled using the MCNP6 code for an annular flow regime. A mathematical algorithm based on an artificial neural network was used to correlate the energy spectra from both detectors with the fluids volume fractions. The pulse height distributions obtained by the detectors are used as input data of the network that outputs the volume fractions of gas and water. The mean relative error, using the procedure presented here, for all data, was below 2.5% for both phases investigated. These results show that the methodology based on an X-ray beam has the potential to be used with flow meters.     

New delay dependent stability criteria for recurrent neural networks with interval time-varying delay

This paper is concerned with the delay dependent stability criteria for a class of static recurrent neural networks with interval time-varying delay. By choosing an appropriate Lyapunov–Krasovskii functional and employing a delay partitioning method, the less conservative condition is obtained. Furthermore, the LMIs-based condition depend on the lower and upper bounds of time delay. Finally, a numerical example is also designated to verify the reduced conservatism of developed criteria.     

利用CHNN人工神经元网络进行排料优化计算

利用连续型Ｈｏｐｆｉｅｌｄ人工神经网络（ＣＨＮＮ）进行金属冲压剪切排料的优化求解计算。讨论了连续型Ｈｏｐｆｉｌｄ人工神经元网络的模型,以及排料问题向连续型Ｈｏｐｆｉｅｌｄ人工神经元网络进行排料优化计算可以得到较好的结果,并且求解质量稳定、速度快。