Application of electrohydraulic servo system contr

  • Detail

Application of electro-hydraulic servo system controlled by neural network in TV Tower

structural active control technology is a new discipline developed in the field of civil engineering in recent 30 years. It uses modern control technology to realize real-time tracking and prediction of input earthquake or wind and structural response, and applies control force to the structure through actuator to change the system characteristics of the structure, so that the structure and system performance meet the optimization criteria of "e". In order to achieve the control method of reducing or suppressing the seismic response of the structure

Nanjing TV Tower is a large reinforced concrete high-rise structure (see Figure 1). The tower has an elevation of 310.10m and a total structural weight of 36852t. The main structure is composed of three hollow rectangular section stress legs. At 180m and 240m of the tower, there are large and small towers respectively. The large tower weighs 10000t. Under the seismic action of basic wind pressure 0.5kpa or fortification degree 8, the bearing capacity and horizontal displacement can meet the requirements. However, under the action of wind, the acceleration response value of the small tower is large. When the basic wind pressure is 35kpa (equivalent to grade 8 wind), the acceleration response value of the small tower is 0.2m2/s, which exceeds the limit value of 0.15m2/s required for human comfort. According to the research, the tower has the characteristics of obvious whip effect and vibration shape degradation. Considering the above factors, the mechanism actively controls amd (active mass damping system) for the small tower. The scheme is to install the AMD device in the fan room on the top of the TV Tower, continuously monitor the state vector of the structural response through the monitor installed on the structure, and calculate the time-varying state vector and feedback vector according to likati's closed-loop control theory, The control force signal is obtained, and then the control force is applied to the structure by using the electro-hydraulic servo device and 60t mass block

the system adopts electro-hydraulic servo to avoid potential safety hazards. Therefore, it has the characteristics of high power, low power consumption and small floor area, which is more suitable for the structural requirements inside the TV Tower. However, due to its inherent shortcomings, the electro-hydraulic servo system has overshoot, oscillation and dead zone phenomena in the operation process, especially in the starting stage of adjusting the mass block, which is easy to cause serious consequences in 2016

neural network control is an extremely active research direction in the international academic community in recent years. Neural network, with its highly nonlinear approximation mapping, unique associative memory, optimization and other functions, provides a new means to solve the control problems of complex systems

based on the analysis of the closed-loop characteristics of the hydraulic servo system that realizes TMD Control, this paper focuses on the analysis of using neural network control to improve the performance of the hydraulic servo system

1. Characteristic analysis of hydraulic servo system

the structure of servo system is shown in Figure 2

(1) actuator this system uses three 247.21 actuators of MTS company in the United States, that is, a single rod oil cylinder, which drives a 60t annular mass block with a rated tension of 50kN

(2) LVDT (linear variable differential trans-

former) is installed on the piston rod of three actuators, and bamboo powder is used to convert the position of the piston rod into a voltage signal feedback system

(3) electro hydraulic servo valve this system adopts MTS 265 series servo valve, the model is 265.18c, which is characterized by a closed-loop LVDT, which is directly fed back to the valve core controller, so as to improve the control accuracy. After theoretical analysis, its transfer function is realized as a second-order oscillation link. Due to its large damping ratio, it can be regarded as a proportional link with a pure delay of 0.5s

this system uses asymmetric 4-way zero opening valve to control asymmetric cylinder, which is derived from classical control theory δ H=0.09, natural frequency ω h-10.18Hz。 It can be seen that the electro-hydraulic servo system has a hydraulic mechanical comprehensive resonance with low frequency and relatively small damping coefficient, which seriously affects the stability of the system. The simplest and economic way to improve the stability is to try to improve the damping of the system

common methods include parallel damping holes, setting accumulators, etc., but the disadvantages of these methods are that parameters such as the size of damping holes or the volume of accumulators are not easy to verify, and once determined, the adjustability is poor and difficult to control, so neural network control is used in this paper to improve the performance of the system

2. Composition and principle of the new control system

the inverse system of the system can be obtained by using the inverse modeling of neural networks to realize self-learning tracking, BP algorithm is used to correct the weight wi (I-5). The system adjusts the weight of wi through learning algorithm to achieve the purpose of system optimal control. The learning process is the algorithm of error back propagation.

using the above neural network algorithm has the following advantages:

(1) the learning process is goal-oriented, and the system accepts the corresponding actual operation output as input during learning

(2) when the forward mapping of the system is not one-to-one, the inverse mapping with the required properties can be obtained

(3) after the learning algorithm completes the learning of the system, the network can serve as the correction network of the system Production and utilization platform of new industrialized building materials

(4) the algorithm system has only 5 nodes and 6 weights. Y is the desired output (step signal), which is equivalent to the inverse process of the system. Its weight system is updated on the time axis. Because of the nonlinearity of neural network, he can effectively overcome the nonlinear influence of the system. In addition, due to the small number of nodes, the learning algorithm is simple, and the algorithm steps are quite simple for a certain set trajectory, so there is no need to calculate the inverse with a large amount of storage space or distance matrix

3. System simulation

simulate the hydraulic servo system shown in Figure 3, and compare the graph trace diagram of adding nerve meridian and not adding nerve meridian. The transfer coefficient is:

under step load Δ Under the disturbance of t=100kn, the dynamic curve of cylinder displacement is shown in Figure 4

according to the characteristics of the hydraulic servo system of the TV tower structure, this paper proposes to use B

Copyright © 2011 JIN SHI