College】 【weak dynamic characteristics of electronic expansion valve identification】..

<br> Abstract: The refrigeration system evaporator superheat with EDM type electronic expansion valve opening changes in the relational model, the electronic expansion valve subjected to electrical pulses drive the stepper motor, evaporator superheat to obtain the dynamic response of the operation as. law ., and apply system identification methods 8 to identify the refrigeration system. After identification of various identification methods comparing the results given in a variety of identification methods in dealing with this type of problem identification accuracy and processing in specific. characteristics, are given for this type .of model identification methods. It also illustrates the link described by a third-order link is closer to the actual situation, after the refrigeration system for the simulation of this simulation provide a better basis.. <BR> Keywords: Electronic expansion valve; .system identification; evaporator superheat <BR> present a variety of refrigeration appliances, because of its complex internal systems and a strong mutual coupling, it is difficult using the usual method to establish the exact mechanism of the system. model, using the conventional proportional .integral derivative (PID) controller is difficult to achieve good control performance, such as operating and working conditions in the initial changes, will need to re-adjust the PID parameters, and sometimes can not even meet the basic requirements. Theoretical analysis and application. .Experience has shown that this type of structure as part of the refrigeration system is known, but changes slowly with unknown parameters to control objects, especially suitable for adaptive control [1]. the most important issue facing the process control is process modeling [2] ., modeling methods are usually. Organic hair cut, and system identification. System identification is through input and output data to construct a mathematical model, which is the basis of self-tuning control system. System identification methods are many, the most commonly used method .is least squares method, in addition to an auxiliary variable., gradient correction method and the maximum likelihood method. but not every problem of identification methods are applicable to all. literature [3,4] gives a random search method, practice shows that this .method is very suitable. in the cooling system of this type of model identification [5]. In this paper, a few of the experimental data identification methods to identify, the results were compared, while analysis of these methods of online real-time requirements .to be suitable for this type of model. identification method. These documents are many ways a more detailed derivation is given on the identification method in the selection of some of the parameters have some way to explain [6,7]. But according to the .parameters of the literature, this paper, the recognition result is often biased estimate. Therefore, the specific model, this paper set the values of these parameters. a test device, such as supermarket display cases for this type of refrigeration system, by adjusting the .expansion valve to regulate the flow of refrigerant system, is on display. cabinet cooling capacity and power to control a simple and effective method. for the realization of the evaporator superheat to control the export target, expansion valve opening to be self-tuning real- .time control, the need for evaporator superheat with the expansion of exports. The relationship between valve opening degree of real-time identification, to determine the link structure. In this paper, [5] of the test subjects - DNS-106 1.1 kW- .type supermarket refrigerated cabinets, stepper motor driven EDM application of electronic expansion valve. by the four-phase stepper motor driver. Pt1000 platinum resistance with two attached to the evaporator, respectively, import and export of wall to feel the evaporator inlet and outlet temperature. .evaporator for the three rows of staggered pipes, pipe length of 1 410 mm,. test temperature was 28 ℃. test device shown in Figure 1.. Figure 1 Schematic experimental device boot time, stability of the system, a step change in the electronic .expansion valve opening, to the expansion valve plus 200 electronic pulse, the valve off small, with 10 s sampling period for the dynamic acquisition evaporator. inlet and outlet temperature, evaporator superheat to obtain the signal. In order to achieve better test results, the .testing process must pay attention to changes in expansion valve opening not too much, nor too small [5]. 2 of the identification method. compared using the least squares (LS) (recursive least squares (RLS)), random search method (LJ .), generalized least squares (GLS), recursive extended least squares (RELS), recursive instrumental variable method. (RIV) (two methods), recursive maximum likelihood method (RML) and random Newton method (SNA) 8 kinds of methods, the .test device shown in Figure 1, collected test data identification, specific process is as follows. (.1) LS method. This recognition system using the step function model, that will be encountered during the matrix operation singular matrix inversion problem, when u read .here in the variance of adding some small random noise. RLS method can also be used for online. identification, initial value: P = 106I, θ = 0. the results are very close identification between the two, so the text is given only the .LS, the recognition result. The result compared with the experimental data shown in Figure 2. Figure 2 shows. the results of recognition have a certain error ratio test data over a small adjustment. This shows that there is noise in the case LS and RLS .can not give high recognition accuracy.. Figure 2 Comparison of various identification methods (2) LJ method. It is assumed that the system is second order. Figure 2 shows that the recognition accuracy is high, but this method is not suitable for online identification .. (3) GLS method. After a few. attempt a first order noise model chosen, the number of iterations of the error bound of 0.1. the number of iterations to obtain error bounds should not be too small, too small, will eventually lead .to iterative divergence, the reason may be due to the increase with the number of iterations., the computer account for rounding errors caused by the weight increase. This third order model and the first noise model, iteration 2 to achieve accuracy. The identification method .and accurate results than the LS, however, little improvement (4).. RELS method. The initial algorithm with RLS, take the first order noise model, the results can be seen from Figure 2, the recognition accuracy is better (because of RELS ., RIV-1, RIV-2, RML, SNA and other identification method results are very close, so. Figure 2 shows RIV-2 is only the recognition results). (5) RIV method. This selection has nothing to do with the .noise of the auxiliary variables: hπ (k) = [-x (k-1), ...,- x (k-. na), u (k-1), ..., u (k-nb) .] for x (k), [7], several algorithms are given as follows: (1) (2) α = 0.01 ~ 0.1;. d = 0 ~ 10 this method were used to identify the two terms, the results were recorded .as RIV-1 and RIV-2. in the process of trying to find the middle of the calculation of the initial value of the variable P strongly influenced the identification accuracy. not even cause divergence values. For the first approach to take P = 106I .; on the second approach taken P = 400I. At the same time, the recommended value of α = 0.01 ~ 0.1, the recognition result is biased. this recommendation to take. α = 0.9 ~ 1, taking d = 1. can be seen .from Figure 2, recognition results are good. (6) RML method. In the calculation that, according to the literature [7] will be initialized to the unit matrix P, the recognition result is. biased. This paper selects P = 16I, .Figure 2 shows the results of recognition is better. (7) SNA method. In this paper the initial value of R obtained for the unit matrix, the convergence factor ρ (k) = 0.9 / (k + .0.3), Figure 2 shows .the recognition results are good. 3 evaporator superheat with the electronic expansion valve opening change models to determine the stepper motor-driven EDM electronic expansion valve, evaporator superheat as applied to the electronic expansion valve. stepper motor pulses the dynamic relationship, in addition to LJ .identification method, the other was to increase the accuracy of the order can not increase the number to determine the model order. according to the literature [5], remove the delay, LJ, LS, RLS will. The second part of the identity of .the other methods identified as the link order. To determine the order of the links will recognize the results of LJ pulse transfer function is derived, which can be seen, the smaller part of the gain, so the dynamic. process of change is more dependent .on the initial state, the initial value, which is shown in Figure 3, simulation results can be drawn. For comparison, here are identified with the RIV-1 method to export the results of the pulse transfer function: from here. very clear that .the order is better than the second order, which can assume the link is the third order. Figure 3 Comparison of simulation results by adjusting the expansion valve 4 Conclusion cooling system to regulate the flow of refrigerant, thereby cooling capacity and. power control, which .is a simple and effective method. In this paper, system identification methods of the 8 session to identify, by comparison, can be considered a part of this, RELS, RIV-1, RIV-2, RML. and SNA identification methods are good ., and also suitable for online identification. LJ identification method of data processing competence, identification accuracy is also high, but because this way the model structure to be determined in advance, which not only adds to the complexity. nature, but also because the model .allowed the pre-determined limits its application. Also, LJ method is slow, it is not suitable for online identification. for the refrigeration system, evaporator superheat with the import and export of electronic expansion valve opening changes. pulse transfer function, this comparison by .simulation that the link order should be part of this refrigeration system simulation for the future lay a certain foundation. Foundation item: National Natural Science Foundation of China (59576044) Author: Zhong Hua (1971 ~.), male, PhD. Author: ( .Shanghai Jiaotong University School of Power and Energy Engineering, Shanghai 200030) Reference: [1] Chen Zhi-long, Zhu Ruiqi, Jing-Yi Wu. refrigeration equipment automation [M]. Beijing: Mechanical Industry Press, 1997 .. <BR> .; [2] Hui Jin. Development and Prospect of process control [J]. Control Theory and Applications, 1997,14 (2): 145 ~ 151. <BR> [3] Stark PA, Ralston D L.. Comparative assessment .of two recent on-line process identification techniques [A]. American Control Conference [C]. Arlington, VA, June 1982. <BR> [4] Luus R, Taakola TH I. Optimization by direct search and systematic <BR ..> reduction of the size of search region [J]. AICHE Journal, 1973,19 (4): 760 ~ 766. <BR> [5] Sun, Zhonghua, Chen Zhi-long. refrigeration system off-line adjustment .process identification [A.] .98 'National Seminar on Mechatronics General Machinery Proceedings [C]. Huangshan, 1998. <BR> [6] Ren Jintang. System identification [M]. Shanghai: Shanghai Jiaotong University Press, 1989. < .; BR.> [7] Fang Chongzhi. the process of identification of [M]. Beijing: Tsinghua University Press, 1988 ...