The change of these parameters is mainly affected by the compressor suction and discharge pressure ratio pd/ps. Therefore, if the compound variable on the left side of the medium (4)(6) is used as the output parameter of the neural network, the suction-exhaust pressure ratio of the compressor must be used as the input parameter of the neural network, and the neural network structure can be large. For simplification.
Considering that the value range of the input and output parameters of the neural network is generally normalized to the interval, the final neural network input and output parameters are input ps/pd, output Vcom/Vth, (Vthps)/Pcom and Ts/Td. Simple analysis can be It is known that the above input and output parameter values ​​can naturally fall within the <0, 1> interval, and there is no need to perform normalization and denormalization operations.
In the simulation calculation, the actual calculation value of the compressor thermal performance parameter can be obtained by simply outputting the parameters through the neural network computer and then performing a simple variable replacement. In summary, a schematic diagram of a neural network model for thermal calculation of a small refrigeration compressor as shown can be obtained. For the sake of brevity, the threshold inputs of the intermediate hidden layer and output layer neurons are omitted from the figure.
The second step is to predict the test values ​​using the trained forward neural network. The calculation results are shown in the table. In the table, max and av represent the maximum and average values ​​of the error (both absolute values). It can be seen from the table that the method of the present invention is a general method and its calculation accuracy is satisfactory.
Conclusion The neural network method <4> for thermal calculation of refrigeration compressor proposed by the author has achieved good results in practical applications. As a versatile method, the work of this paper is further improved on the basis of the traditional model, which makes the method more simple and effective, and realizes the effective fusion of the traditional model and the neural network, and the actual effect is satisfactory. The ideas in this paper are also universal and can be used as reference for researchers in different engineering fields.
Friction And Wear Testing Machine
Friction And Wear Testing Machine,Wear Testing Equipment,Four Ball Friction Testing Machine,Friction And Wear Tester
Jinan Chenda Testing Machine Manufacturing Co., Ltd. , http://www.jncdtester.com