CERTAIN INVESTIGATION OF REAL TIME NETWORKED CONTROL DC MOTOR USING SMITH PREDICTOR

Article History:Received:11 november 2020; Accepted: 27 December 2020; Published online: 05 April 2021 ABSTRACT : In this paper, DC motor is controlled by direct structure Networked Control System (NCS) with smith predictor control strategy. Mathematical model of DC motor is identified by empirical model building and estimated Round Trip Time Delay (RTT) is considered as time delay which is induced by network. Delay time which is induced in Networked DC motor is compensated with the smith predictor controller then the designed controller algorithm is comparatively analyzed with conventional PI controller. The simulation results are formulated for identified motor model with real time network. In addition to that the hardware fabrication of real time Networked DC motor has been developed and the results of both the simulation and hardware implementation are compared.


INTRODUCTION
Nowadays many automation industries are replacing their conventional control into networked based control because of its extensive functionality, rapid development of low cost microprocessors and communication The system which is shown above in Fig.1 shows that the data transfer from the host certain network delays can occur among controller and the distant system processing delay in the controller since they are working under a network. The delay between the controller and sensor is τ sc and the delay across actuator and control τ ca are two types of network delays in NCS, depending on the path of data transmission. The total network delay has been calculated as τ = τ ca + τ sc + is greater than 1.
Transfer function of the NCS's is as follows: Y(s) R(s) = G c (s)e −τ ca s G p (s) 1 + e −τ sc s G c (s)e −τ ca s G p (s) (1) From (1), a time delay term in equation, which limits controller gain and as a consequence, the delay dominant system's control action is limited. When the system contains time delay, the closed loop performance will be degrade and even destabilize the stable loop due to the constraints imposed by time delays. Figure 2 portrays the design of a direct structure networked control system. The UDP/IP protocol was used to establish contact between the server and the client. The software and hardware configuration for UDP/IP protocol is done with the help of real time windows target tool box in MATLAB. The server and client node applications are implemented as simulink models and it has been run on real time operating system in external mode connection. The components used in building the direct structure NCS experiment are given as follows.

NCS SYSTEM ARCHITECTURE
Client Node: It acts as closed loop control controller, in which the controller receiving speed measurement data via the network and it produces control signals after comparing it to the reference signal. These signals are compressed in packets and guided it to the server node through communication.

Fig. 2 Architecture of networked DC motor plant
Server Node: It acts as a gateway through which the measurement data has been collected from plant and frame the data as pocket then send it to the slave node (controller). In the same way Server Node collects the control data packet from the slave node and sends the control signal to the plant. UDP/IP protocol: UDP/IP protocol is used for transmitting the information between client and server node across the network. The Internet protocol (IP) has the collection of network protocols that are used to connect on the internet, in which the User Datagram Protocol is the core member. UDP has no handshaking signals hence there is no possibility of retransmission. Because of these characteristics it is mostly used in real time applications. PCI 6251 DAQ card: DAQ PCI 6251 is the PCI based on-board data acquisition hardware. It serves as a data acquisition device which contains ADC, DAC, digital I/O, internal clock and a timer. PCI 6251 DAQ card connected to the PCI slot on the computer mother board. PCI 6251 cannot be interfaced directly to the plant so NI ELVIS-II is used between the plant and the DAQ card. All data acquired from these NI ELVIS-II are then send to PCI 6251 DAQ card via data cable. The PCI-6251 DAQ card consist of ADC and DAC both having 16 bit resolution and the voltage ranges are -5 to +5v and -10 to +10v. DC motor plant: DC motor is an electromechanical device in which electrical energy is converted into mechanical energy like motion and angular speed. The control signal which has been generated by the DAQ card to the DC motor is received through the PWM based actuator. The DC motor speed is measured using the encoder setup in which the pulse is generated based on the speed. The frequency of the pulse is converted into voltage with the help of frequency to voltage circuitry. The voltage range is (0-5) v and the corresponding DC motor Speed is (0-1500) rpm.

DC MOTOR MODEL IDENTIFICATION
This section briefly describes the DC motor model identification for design of controllers (PI Controller and smith predictor controller).The DC motor model is required for controller design and implementation so the identification and modeling of DC motor is obtained from the experimental setup as shown in Fig The validation of the model and the real system are shown in Fig. 4. It shows, that the identified DC motor model approximately fits with the real model.

CONTROLLER DESIGN
This section describes the design procedure of Strategy using a PI controller and a Smith predictor controller. PI Controller design: Because of its structural simplicity, the proportional-integral controller is most generally proposed controller configuration in many industrial systems. The PI controller is a conventional feedback controller that uses a weighted sum of errors and the integral of the value to control the plant. The PI controller is mathematically denoted as: Kp and Ki are the controller parameters where these two parameters are called controller design or tuning parameters, It is obtained by the IMC based tuning method formula [9]. The designed PI controller gain values are Kp =0.02913 & Ki =0.04163 for closed loop time constant τ c = 0.2. Smith predictor controller design: Figure 5 shows the smith predictor structure. It consists of model Gm(s), actual process GP(s) to be monitored, and the actual control loop's expected RTT delay. It provides a simulated environment in which the loop delay is compensated for using the plant model and assumed delay. The NCS's actual feedback loop is the outer loop, which includes a regular PI controller Gc (s). The inner loop is a virtual loop consisting of process models Gm(s), which was obtained using the device identification method and has an approximate RTT delay of . The loop outputs are deducted in order to withdraw delay effect in control loop.
The results of network-induced delay were included, as seen in (4). Its effect can degrade system performance and few information, contribute to instability of the system. As a result, it is important to reduce negative outcomes via achieving condition (5).
(e −τ ca s G p (s) − G c (s)G m (s)e −τ est s ) ≈ 0(5) Equation (5) As a result, compensator of predictor eliminates the delay from the equation, permitting for an rise in gain as shown in the equation (6).

RTT delay estimation:
A clock signal is sent through the network to calculate the delay in the NCS loop. In each time sample across the network, the is considered as the modification among transmitted and received clock signals. The average estimated delay after the delay estimation procedure is 0.33sec, which is much longer than the sample period of 0.01sec.

SIMULATION AND EXPERIMENTAL RESULTS OF NCS
Various simulation and trials are piloted to investigate NCS control performance for designed controllers.
Simulation results: With the support of MATLAB simulink and the real-time windows target tool box. The sampling time has been set to 0.01 seconds. The device response obtained without the use of a network environment using the PI controller is shown in Fig.6. With a settling time of 1.4 seconds and a peak overshoot of 2.77 percent, this response meets the controller's design requirements.  The Fig.8shows the system response obtained using smith predictor controller using controller with network environment. In this case instability of PI controller is stabilized by using smith predictor controller and performance is improved as settling time of 2.77 sec and peak overshoot as 23.53%. The estimated RTT delay between two PC is 0.33 sec. The experimental results is obtained by using without a network, a PI controller environment is depicted in Fig.10. This response shows the settling time of 2.8 sec and peak overshoot as 17.64%.

Fig.10 DC motor response using a PI controller
The Fig.11 shows the experimental results obtained using PI controller with network environment. This experimental result indicates the instability of the PI controller used in NCS.

Fig.11 PI controller response to a networked DC motor
The Fig.12shows the experimental results using smith predictor controller with network environment. In this case instability of PI controller is stabilized by using smith predictor controller and performance is improved with the settling time of 4.1 sec and peak overshoot as 24.65%. According to the simulation and experimental findings, that the smith predictor controller ensures better control performance in the NCS. The performance index of the experiment is much larger than simulation because uncertainty of identified DC motor system as shown in Table.1.

Table.1 Performance Index Comparison CONCLUSION
The smith predictor controller is used to solve the problem of the NCS with constant delay is discussed, and tested on a simultaneously using networked DC motor drive. This design, however, necessitates a precise mathematical model of the process as well as an accurate estimate of the total network delay. Results of simulations and experiments are compared, in which smith predictor controller performed better than the PI controller. In the future, a method for balancing simulation execution on the master and slave nodes will be developed. This research can also be used to build a filtered adaptive smith predictor for varying delay NCS and to minimize noise effect on the output response.