The reciprocating cutter is a crop stem cutting device commonly used in modern agricultural harvesting machinery, and is widely used in various windrowers and combine harvesters. The working performance is mainly evaluated by the area of ​​the primary cutting zone, the heavy cutting zone and the missing cutting zone in the cutting diagram. When selecting the working parameters, attention should be paid to controlling the area of ​​the heavy cutting zone and the missing cutting zone to reduce the heavy cutting. Power wastage and the phenomenon of cutting or missing due to missing cuts. The traditional evaluation method of cutter performance is mainly carried out by using the cutting diagram drawn by the drawing method. It can only analyze the change trend of the 3 regions in the cutting diagram when the cutter operating parameters change. This is just a qualitative evaluation method. MATAB is a set of high-performance numerical calculation software, which can realize the accurate drawing of the cutting diagram and the quantitative calculation of the cutting area, so as to realize the quantitative evaluation of the working performance of the cutter. However, for the working performance of the cutter under certain working conditions, it must be completed according to the mathematical model of establishing the cutter movement, drawing the cutting diagram, calculating the coordinates of the key points in the cutting diagram and calculating the area of ​​the 3 areas in the cutting diagram. More complex 121. Using artificial neural network technology to establish a model of cutter performance evaluation, the cutter performance can be evaluated directly, effectively and quantitatively, avoiding the inaccuracy of traditional qualitative evaluation and the complicated process of using software drawing calculation. This will form a new quantitative evaluation method, which has important theoretical significance, and has certain guidance for the reasonable design of the working parameters of the cutter.
In order to concisely and effectively evaluate the workability fund project of the reciprocating cutter: Ningxia University Natural Science Fund Project (ZR200813) can use the powerful MATAB software to obtain the cutting diagram under different working conditions. The area value of the 3 regions in the cut graph provides a certain number of sample sets for the training of the artificial neural network. In the model establishment of the neural network, due to the layer (S-type linear) structure, if the S-type layer has enough neurons, it can train the characteristics of the rational function relationship between any input and output, which can be realized by BP network. Input-output training for samples. Using the trained network, it is possible to predict and simulate the working performance of reciprocating cutters with the same structural parameters under different working conditions. 4.1 Obtaining training samples and detecting samples 1.1 Sample acquisition methods and steps Deriving reciprocating cutters The equation of motion of the moving blade.
Use the drawing commands of P0 and hldn of MATAB software to program the cutting diagram of the reciprocating cutter under certain working conditions (certain machine forward speed and crank speed).
The MATAB software solves the univariate function zero command fZe to calculate the intersection coordinates of the curve in the cut graph.
Calculate the area value of the 3 areas in the cut chart 1671. 1.2 Obtain the sample data According to the above steps, the area values ​​of the 3 areas in the cut chart of the reciprocating cutter under different working conditions can be obtained, as shown in Table 1.
2 Establish an artificial neural network model to evaluate the performance of the reciprocating cutter. The input of the network is the forward speed (mm/s) of the machine and the crank speed of the cutter drive mechanism (rmn). The output of the network is the area of ​​one cutting area and the area of ​​heavy cutting. Area and area of ​​the missing area (mm). Taking the data numbers 118 and 2536 in Table 1 as training samples and 1924 as the test samples, the M file arxnnX m of the 3-area area model is established as follows: 8-9: Input sample output sample = prenrmxp) % sample normalization) % Establish corresponding BP network net hw=5Q% Training display interval setting netmc=Q%% Momentum factor setting net攸11=200000% Maximum training times Set network% to simulate BP network%Renormalization processing f2=, test Sample processing an = ln (netf2n)% test sample simulation value åº = psmrmx (a2nmntmxt)% test sample simulation value inverse normalization 2 =, output target % test sample output relative error Table 1 under different working conditions in the cutting diagram 3 Area area value Tb1Thedtofhredisitsraaathediffernt working parameters working performance parameter serial number machine advance crank one cutting area heavy cutting area missing cutting area speed speed area 32 test sample target output value matrix t continued Table 1 working parameters working performance parameter serial number machine forward crank one cut Zone heavy cutting zone missed zone speed speed area 33 test sample actual output absolute error (2 - empty) value of 34 test sample actual output relative error (100 The value is the matrix Xwuha network training weight and threshold matrix 3 network training and network simulation to achieve network performance effectiveness, through different network structure training network, and finally determine the network output error minimum network structure.
Run the M file in the MATAB command window, and adjust the number of neurons in the hidden layer of the network. It is found that the output error is the smallest when the number of neurons is n=6. Therefore, the BP network model of the 2-6-3 structure established by 6 is determined. Show.
The parameters and performance indicators after network training are as follows.
The actual output value of the 31 test samples is matrix 2 W1. The error curve of the network training is as shown.
4 Conclusions 1) An artificial neural network model for evaluating the performance of the reciprocating cutter was established. For reciprocating cutters under different working conditions, it is only necessary to input the machine forward speed and crank speed in the model, and the network can output the area values ​​of the primary cutting zone, the heavy cutting zone and the missing cutting zone which reflect the working performance of the cutter. A fast, quantitative and effective evaluation mode is achieved for the performance of the cutter.
After 725 trainings, the expected error target of 7.5e-005 was achieved. The absolute and relative errors of the output of the test sample are small, indicating that it is feasible to apply this network to predict the performance evaluation index of the reciprocating cutter with certain accuracy.
4) Artificial neural network technology can be used to predict complex systems. Under the condition that the relationship between the target and its influencing factors is not established, the output target of the system can be predicted, which provides a convenient way for people to understand and study complex systems.
Water Based Rheological Additives
Rheological Additives,Rheology Modifiers Thickeners,Organoclay Rheological Additive,Water Based Rheological Additives
CHANGXING GUANGDA NEW MATERIALS CO.,LTD. , https://www.gdbentonite.com