Peradeniya, Peradeniya 20400, Sri Lanka. Royal Institute of Technology, Stockholm, Sweden.

Transcription

1 REMOTE MONITORING AND DISTRIBUTED REAL-TIME CONTROL VIA ETHERNET SANATH ALAHAKOON 1, LILANTHA SAMARANAYAKE 2, THILAKASIRI VIJAYANANDA 1, MATS LEKSELL 2 1 Dept. of Electrical and Electronic Engineering, University of Peradeniya, Peradeniya 20400, Sri Lana. sanath@ee.pdn.ac.l 2 Division of Electrical Machines and Power Electronics, Royal Institute of Technology, Stocholm, Sweden. ABSTRACT Ethernet today is the most widely used information carrier and the service provider for many important applications in our day-to-day life such as , voice-image data and web based information. It is also emerging strongly in the area of industrial communication. This paper presents research and development being carried out to enhance the possibilities of using standard TCP/IP Ethernet for real-time condition monitoring and distributed real-time control. 1. INTRODUCTION Condition monitoring and closed-loop control are essential and well-nown techniques in any industrial environment. A controller or an observer receives information about the system (or process) to be controlled or observed from the sensors and in case of a controller it sends out driving signals to the actuator. Condition monitoring done from a location remote to the place at which the particular industrial process is located is more common nowadays (e.g. Control room of a factory). Control loops that are closed over a communication networ also get more and more common as the hardware devices for networ and networ nodes become cheaper due to advanced cost effective silicon technology. In such a system, measurement and control signals are transmitted between process and controller/observer modules as pacets of data. A control system communicating with sensors and actuators over a communication networ is called a distributed control system. Networ nodes that are of specific interest here are sensor nodes, actuator nodes, and controller nodes. These types of industrial applications demand fast, reliable and robust data communication at a reasonable cost. Employing a suitable field-bus full-fills some of them. Profibus, ControlNet, DeviceNet, Ethernet, Suconet, and Interbus etc are among

2 SENSOR 1 PROCESS OUTPUT SENSOR RJ45 Ethernet Rx/ Tx TCP/IP SENSOR 2 Ethernet Ethernet PROCESS OUTPUT SENSOR Microcontroller Monitoring Computer Control Computer Additional memory A/D Converter D/A Converter (a) Condition monitoring (b) Distributed Control External system Figure 1: Ethernet for industrial communication Figure 2: Ethernet module for ADC and DAC the commonly used field-buses. Whatever the networ used for the communication, one major requirement of such a system is acquiring the ability to easily connect any physical sensor or actuator to it. In other words it is the interfacing of the sensor/actuator node to the communication networ. Being a commonly available hardware and software solution for communication networing, Ethernet has received a lot of attention from industry as the future industrial communication medium [1]. One objective of this research is to address this problem of interfacing of the sensor/actuator node to the communication networ, when the communication is done via Ethernet. The second objective is to investigate the possibility of using standard TCP/IP Ethernet for distributed control [2]. This would enable real-time condition monitoring and distributed control system topologies based on Ethernet shown in Figure ETHERNET-READY SENSORS AND ACTUATORS Objective of this part of the research is to design and implement sensor and actuator nodes that can easily be connected to Ethernet. Most of the outputs of physical sensors can be obtained in the form of an analog voltage (a taco generator) or a digital word (an incremental encoder). Similarly, inputs of many actuators can be provided in the form of an analog voltage (linear power amplifier) or a digital word (PWM inverter). Thus, the whole design problem is reduced to design and implementing the hardware needed to interface (a) An Analog to Digital Converter (ADC), (b) Digital to Analog Converter (DAC) and (c) A digital I/O to the Ethernet. The complete design problem is divided into two phases. Namely, Phase I - Interfacing of an ADC and a DAC to the Ethernet and Phase II - Interfacing of a digital I/O to the Ethernet.

3 2.1 Interfacing of an ADC and a DAC to the Ethernet A hardware module that interfaces an ADC and a DAC to the Ethernet through a hardware TCP/IP stac as shown in Figure 2 has been designed and implemented. Associated software programming is being done at present. The unit will be capable of first receiving the information such as sampling time etc. from the central control or monitoring computer and operating the ADC and DAC accordingly. The prototype module has been designed to meet the performance demands of hard real-time applications (closed loop control) as they are more demanding than real-time condition monitoring. The particular application considered here is the closed loop speed control of motor drives through Ethernet. The prototype module will be capable of completing one AD conversion and a DA conversion needed for a single actuation cycle within 1 ms. The expected timing diagram is shown in Figure 3. Communication lin diagram obtained from the protocol analyzer while the prototype is maintaining a TCP/IP connection with the server is given in the Figure 4. The hardware module designed and fabricated is shown in Figure DISTRIBUTED REAL TIME CONTROL VIA ETHERNET Investigations on adapting Ethernet for distributed control systems is an interesting topic in the motion control industry, due to its commercially off the shelf hardware availability and compatibility at a comparatively lower price. Conversely, communication through shared media in standard Ethernet is non-deterministic due to its media access control method CSMA/CD (Carrier Sense Multiple Access with Collision Detection). This results in non-deterministic (stochastic) delays of data transfer. The main goal of this part of the research is to develop controller strategies, which can successfully overcome the problems associated with these non-deterministic time delays when a real-time control loop is closed via Ethernet. In other words, no changes will be suggested to the standard TCP/IP. It is believed that this approach will finally lead to a unified industrial communication standard based on Ethernet TCP/IP. The approach is also justified by the recent developments in the TCP/IP (the newer version of the IP protocol IPv6, which has Data conversion Data conversion PC Received the Data Transmission delay Data storing, processing, sampling & holding DAC Send to actuator ADC Transmission delay Control signal computation & Data conversion for the Ethernet Tc<1ms Figure 3: Timing diagram showing a single actuation cycle

4 Figure 4: Lin between the prototype module (IP address ) and the server (IP address ) Figure 5: Ethernet connectable hardware Figure 6: BLDC motor drive system with module speed control loop closed via Ethernet hard real-time capabilities) [1]. As the test rig for verifying controller strategies developed to handle the non-deterministic time delays in the Ethernet, a Brush Less DC (BLDC) motor drive system of which the speed sensor and the power electronic actuator

5 (inverter) are interfaced to the Ethernet by means of a micro-controller is used [3]. The complete system is shown in Figure Speed Controller Strategies It is intended to implement different speed controller strategies that can cater the nondeterministic time delays of the standard Ethernet in order to mae a good comparative study between them. The controllers that have been implemented so far are (a) PI with special tuning for time delays [4,5,6], (b) Smith Predictor modified for time delayed systems [7,8] and (c) State feedbac controller with online delay compensation [2]. Each of the above controller strategies will be summarized here. It must be noted here that in the Proportional and Integral (PI) controller design and in the Smith Predictor design, the control delay τ is taen as constant and equivalent to its mean τ mean, which is evaluated off-line prior to activate the controller. In this way the need to compute controller parameters during each sampling period can be avoided. (a) PI Controller This method is applicable only for the case where 0<τ<h (sampling time). In other words, it guarantees that all samples arrive at the controller node in chronological order. Then the harm caused by the delay is limited to the non-periodic arrival of the control signals at the actuator node. The tuning rules are first derived in [6] as 1 1 K p τ K i 2 mean mean 3τ sτ for a first order, type 1, and time-delayed system. e. Later some modifications were mean s added by trial and error in [4] and [5] as a a 1 K K p i aτ mean aτ mean for the first order, type 0, time delayed system. Stability enhancement together with the integrator wind-up is discussed in [9] and [10] and its implementation is discussed in [2]. (b) Smith Predictor This is an effective dead-time compensator for stable processes with time delays. The controller is applicable for τ>h as well. The closed loop transfer function for the complete distributed system in Figure 7 can be given by τs ω() s Gc () s G() s e () 1 ()[ () () () ], (1) τs τ m s ωref s Gc s Gm s G s e Gm s e where G(s)} - system to be controlled, G m (s) - estimated model system of G(s), G c (s) controller, τ mean - estimated control delay. Stability of the Smith predictor is affected by the accuracy with which the model i.e., G m (s) and τ m represent G(s) and τ respectively. Based on the assumption that model matches actual system perfectly, the transfer function reduces to

6 (s) ω ref - - G c (s) G (s) Motor e τs Networ ω(s) G m (s) Motor model e τ ms Networ model - ω ω () s () s ref τ mean s () s G( s) e G () s G (). Gc 1 s c m Figure 7: Smith predictor The parameters of the primary controller are determined using a model of the delay free part of the system. G c (s) can be PI, PD or PID depending on the order of G(s) [7]. To derive G c (s) the Coefficient Diagram Method (CDM), whose mathematical derivations are found in [8] is used. (c) State feedbac controller A general disturbance free continuous time control system with negligible input time delay and a discrete time state feedbac controller can be described as dx Ax( t) Bu( t), (3) dt y ( t) Cx( t), (4) u ( h) Kx( h), 0,1,2.. (5) n m p where x R, u R and y R and K s dimensions are compatible with A, B and C. In the following analysis, the control delay corresponding to the th sample will be taen as τ in place of τ in the previous discussions. i For 0< τ <h With this constraint, at most two system inputs u(h-h) and u(h) are available during the th sampling period. Then the system equations in continuous time are dx() t Ax() t Bu() t, t [ h τ, h h τ 1] dt y t Cx t, (6) () () ( t ) Kx( t τ ), t [ h τ, 0,1,2... ] u where u(t ) is piecewise continuous and only changes the value at time version of the system is (2) h τ.the discrete

7 ( h h) Φ( h) x( h) Γ ( h, τ ) u( h) Γ ( h, τ ) u( h h) 1 ( h) Cx( h) x y 0 (7) h Ah As As where Φ( h) e, Γ1 ( h, τ ) e ds. B, Γ0 ( h, τ ) e ds. B. h τ 0 The closed loop system becomes, ζ ( h h) Ψ( h) ζ ( h) ζ ( h) [ x( h), u( h h) ] T and Φ( h) Γ ( h, τ ) K Γ ( h, τ ) 0 1 Ψ( h). h τ, where K 0 ii For τ >h With 0< τ <dh, where d>1 there can be zero, one or more than one (upto d) sample(s) in a single sampling period. In the case where (d-1)h< τ <dh, for all >d, only one sample is received in every sampling period. Following the same analysis as ' ' Φ( h) Γ ( h, τ ) Γ (, ) h τ I... 0 Ψ( h) I K Hence in the time-invariant systems, the instantaneous τ can be compensated by using it on the measured plant state, in estimating the plant state by the time the control signal reaches the actuator. An estimator is used to determine the actual speed state of the motor using the delayed speed measurement, delay magnitude of the previous control signal and the dynamics of the system, i.e. the motor. This demands the real-time delay data, which is achieved through time stamped speed measurements and the characteristics of the networ used. A time delay profile is illustrated in Figure 8. The process state represents the actuator communication delay τ ca, where their summation is termed as control delay τ. τ sc is the only parameter nown a priori and τ ca can be predicted based on the delay patterns of the networ. Hence τ c is nown off-line and τ can be evaluated in advance and used in the estimator as h τ Aτ A( h τ m ) x( h τ ) e x( h) e Bu( m) dm (8) h where A and B are the system and input matrices respectively of the plant. 3.3 Experimental Results The experiments conducted so far reveals that the state feedbac controller with online delay compensation, which is a new contribution of this wor has better step response performance. This is verified from Figure 9.

8 Process State x(h) Estimator x h τ ) ( sc τ sc Controller u τ c ( h τ ) τ ca Actuator ( 1)h h ( 1) h τ h: sampling period : th sample Figure 8: Time delay profile. Figure 9: Step response comparison 4. ACKNOWLEDGEMENTS The authors would lie to pay their gratitude to SIDA Sweden for the financial support. 5. REFERENCES [1] N.D. Aavaag, N.A. Nordbotten, Implications of the Next-Generation Internet Protocol for ABB IPv6 the new Internet Protocol, ABB Review, Jan 2003, pp [2] L. Samaranayae, Distributed Control of Electric Drives via Ethernet, Tech. Lic. Thesis, Department of Electrical Engineering, Royal Institute of Technology, Sweden, [3] C-Programmable Single-Board Computer with Ethernet and Operator, User Manual, Z-World Inc, California, USA. [4] A. O Dwyer, Performance and robustness issues in the compensation of FOLPD processes with PI and PID controllers, Irish Signals and Systems Conference, Dublin, Ireland, June [5] J. Syder, T. Heeg, A. O Dwyer, Dead-time compensators: performance and robustness issues, Dublin Institute of Technology, [6] J. G. Ziegler and N. B. Nichols, Optimum settings for automatic controllers, ASME Trans.1942, Vol. 64, pp [7] S. E. Hamamchi, I. Kaya, D. P. Atherton, Smith Predictor Design by CDM, European Control Conference [8] S. Manabe, Application of Coefficient Diagram Method to MIMO design in aerospace, IFAC 15th Triennial World Congress, Barcelona, Spain, [9] K. S. Walgama, On the Control Systems with Input Saturation or Periodic Disturbances, Luleå University of Technology, Luleå, Sweden, [10] S. Alahaoon, Digital Motion Control Techniques for Electrical Drives, Royal Institute of Technology, Stocholm, Sweden, 2000.