Design And Analysis Of Mechanical And Electronic System Providing Constant Force

With the continuous progress of space missions,the requirement of ground physical verification experiments becomes higher and higher.Thus,the verisimilitude of zero-gravity environment will be a great problem.The paper designs a mechanical and electronic system providing constant force for the traditional five-degrees-of-freedom satellite simulator to simulate the zero-gravity environment of the vertical direction.This system is able to provide high-precision constant force for its load and it has lost of advantages such as simple structure,small size,strong anti-interference ability,low cost and long working hours.The system consists of mechanical passive system which provides constant force and electromagnetic active force compensation system.The mechanical system is to undertake most gravity without any source,while the electromagnetic system can offset the error of the mechanical system.The paper firstly designs the total mechanical and electronic system.The mechanical passive system is based on the constant supporting spring hanger with cams.The paper designs its core components in detail and analyzes its error sources.The relation between the error of mechanical passive system and the state of motion of the load is also been stated.The electromagnetic system is divided into two parts.One is actuator and the other is gas bearing which plays a guiding role.Then a voice coil motor is chosen as the actuator and its theory as well as its characteristic and mathematical model is stated in paper.The paper also analyzes the theory and characteristic of the gas bearing and chooses its model.The paper builds an equivalent model for the mechanical passive system through the system identification based on neural network.This method is based on the experimental data.Through the physical experiment,the constant force output accuracy is got.At the same time,the sample for identification is collected.The samples are divided into two groups,i.e.,identification group and control group.The paper uses three different neural networks to identify the identification group.Then it compares each identification result with the data of control group.The paper finds that the performance of the fuzzy neural network is the best.Thus,the identification model of the fuzzy neural network is served as the equivalent model for the mechanical passive system.Moreover,the constant force output error of this model is the control aim of the electromagnetic system.Aiming at the problem that the control aim is non-linear,step and indeterminate,in this paper the neural network model reference adaptive controller is used to design the control system.This method is highly adaptive so that it can quicken the response speed of control system and improve the control quality.The force sensor gathers the output force of the system in real time,whose signal composes a feed close loop to realize the control of the voice coil motor.The paper builds the mathematical mode for each part of the control system and does the computer simulation for the whole system.By applying different external incentives to the system,the paper assesses how the constant force output accuracy of the system is influences.The influence factors include the difference between the real model and the reference model,the state of motion of the load and measuring error of the force sensor.The result shows three conclusions as follow.1)The influence form the difference between two models is slight.Its error rate is less than 10^-5.2)The faster the experimental load moves,the more the force output error of the system.But its error rate is still not big,which is less than 10^-4.3)The measuring error of the force sensor influences constant force output accuracy more.The experiment shows that the error rate can be 0.3639% if the experimental load is high-dynamic.After all,the mechanical and electronic system can provide constant force in high accuracy that better than the design index in paper.The system design and the control effect are validated.