P1: KUF/OVY JWCL373-05-II P2: OSO/OVY QC: SCF/OVY T1: SCF JWCL373-Brannan-v1 October 21, 2010 Projects 389 Project 3 The Watt Governor, Feedback Control, and Stability In the latter part of the 18th century, James Watt designed and built a steam engine with a rotary output motion (see Figure 5.P.3). It was highly desirable to maintain a uniform rotational speed for powering various types of machinery, but fluctuations in steam pressure and work load on the engine caused the rotational speed to vary. At first, the speed was controlled manually by using a throttle valve to vary the flow of steam to the engine inlet. Then, using principles observed in a device for controlling the speed of the grinding stone in a wind-driven flour mill, Watt designed a flyball or centrifugal governor, based on the motion of a pair of centrifugal pendulums, to regulate the angular velocity of the steam engine’s flywheel. A sketch of the essential components of the governor and its mechanical linkage to the throttle valve is shown in Figure 5.P.4. FIGURE 5.P.3 The Watt steam engine, 1781. To understand how the mechanism automatically controls the angular velocity of the steam engine’s flywheel and drive shaft assembly, assume that the engine is operating under a constant load at a desired equilibrium speed or operating point. The engine, via a belt and pulley assembly, also rotates a straight vertical shaft to which a pair of flyballs are connected. If the load on the engine is decreased or the steam pressure driving the engine increases, the engine speed increases. A corresponding increase in the rotational speed of the flyball shaft simultaneously causes the flyballs to swing outward due to an increase in centrifugal force. This motion, in turn, causes a sliding collar on the vertical shaft to move downward. A lever arm connected to the sliding collar on one end and a throttle control rod on the other end then partially closes the throttle valve, the steam flow to the engine is reduced, and the engine returns to its operating point. Adjustable elements in the linkage allow a desirable engine speed to be set during the startup phase of operation. A Problem with Instability. Although some governors on some steam engines were successful in maintaining constant rotational speed, others would exhibit an anomalous CONFIRMING PAGES 15:52 P1: KUF/OVY JWCL373-05-II 390 P2: OSO/OVY QC: SCF/OVY T1: SCF JWCL373-Brannan-v1 Chapter 5 October 21, 2010 The Laplace Transform Sliding collar Steam flow to engine Throttle valve φ θ0 FIGURE 5.P.4 The centrifugal governor. rhythmic oscillation in which the system appeared to “hunt” unsuccessfully about its equilibrium value for a constant speed. The amplitude of oscillations would increase until limited by mechanical constraints on the motion of the flyballs or the throttle valve. The purpose of this project is to develop a relatively simple mathematical model of the feedback control system and to gain an understanding of the underlying cause of the unstable behavior. We will need a differential equation for the angular velocity of the steam engine flywheel and drive shaft and a differential equation for the angle of deflection between the flyball connecting arms and the vertical shaft about which the flyballs revolve. Furthermore, the equations need to be coupled to account for the mechanical linkage between the governor and the throttle valve. The Flyball Motion. The model for the flyball governor follows from taking into account all of the forces acting on the flyball and applying Newton’s law, ma = F. The angle between the flyball connecting arm and the vertical shaft about which the flyballs revolve will be denoted by θ. Assuming that the angular velocity of the vertical shaft and the rotational speed of the engine have the same value, , there is a centrifugal acceleration acting on the flyballs in the outward direction of magnitude 2 L sin θ (see Figure 5.P.5). Ω θ mΩ 2L sin θ cos θ L m –mg sin θ mΩ 2L sin θ –mg FIGURE 5.P.5 The angle of deflection of the centrifugal pendulum is determined by the opposing components of gravitational force and centrifugal force. CONFIRMING PAGES 15:52 P1: KUF/OVY JWCL373-05-II P2: OSO/OVY QC: SCF/OVY T1: SCF JWCL373-Brannan-v1 October 21, 2010 Projects 391 Recall from calculus that the magnitude of this acceleration is the curvature of the motion, 1/L sin θ, times the square of the tangential velocity, 2 L 2 sin2 θ. Taking into account the force due to gravitational acceleration and assuming a damping force of magnitude γ θ , we obtain the following equation for θ: m Lθ = −γ θ − mg sin θ + m2 L sin θ cos θ. (1) Equation (1) results from equating components of inertial forces to impressed forces parallel to the line tangent to the circular arc along which the flyball moves in the vertical plane determined by the pendulum arm and the vertical shaft (Problem 1). Angular Velocity of the Steam Engine Flywheel and Drive Shaft. The equation for the rotational speed of the flywheel and drive shaft assembly of the steam engine is assumed to be d J = −β + τ, (2) dt where J is the moment of inertia of the flywheel, the first term on the right is the torque due to the load, and the second term on the right is the steam generated torque referred to the drive shaft. Linearization About the Operating Point. In order to use the feedback control concepts of Section 5.9, it is necessary to determine the linear equations that are good approximations to Eqs. (1) and (2) when θ and are near their equilibrium operating points, a mathematical technique known as linearization. The equilibrium operating point of the steam engine will be denoted by 0 , the equilibrium angle that the flyball connecting arm makes with the vertical will be denoted by θ 0 , and the equilibrium torque delivered to the engine drive shaft will be denoted by τ 0 . Note that in the equilibrium state, corresponding to θ = 0, θ = 0, and = 0, Eqs. (1) and (2) imply that g = L20 cos θ0 . (3) τ0 = β0 . (4) and To linearize Eqs. (1) and (2) about θ 0 , 0 , and τ 0 , we assume that θ = θ0 + φ, = 0 + y, and τ = τ0 + u, where φ, y, and u are perturbations that are small relative to θ 0 , 0 , and τ 0 , respectively. Note that φ represents the error in deflection of the flyball connecting arm from its desired value θ 0 , in effect, measuring or sensing the error in the rotational speed of the engine. If φ > 0, the engine is rotating too rapidly and must be slowed down; if φ < 0, the engine is rotating too slowly and must be sped up. Substituting the expressions for θ, , and τ into Eqs. (1) and (2), retaining only linear terms in φ, y, and u, and making use of Eqs. (3) and (4) yield φ + 2δφ + ω02 φ = α0 y (5) J y = −βy + u, (6) and where δ = γ /2m L, ω02 = 20 sin2 θ0 , and α0 = 0 sin 2θ0 (Problem 3). The Closed-Loop System. Regarding the error in rotational speed, y, as the input, the transfer function associated with Eq. (5) is easily seen to be α0 G(s) = 2 . s + 2δs + ω02 CONFIRMING PAGES 15:52 P1: KUF/OVY JWCL373-05-II 392 P2: OSO/OVY QC: SCF/OVY T1: SCF JWCL373-Brannan-v1 Chapter 5 October 21, 2010 The Laplace Transform Thus (s) = G(s)Y (s), where (s) = L{φ(t)} and Y (s) = L{y(t)}. Regarding u as the input, the transfer function associated with Eq. (6) is H (s) = 1 Js + β so that Y (s) = H(s)U(s), where U (s) = L {u(t)}. The closed-loop system is synthesized by subtracting (s) at the summation point, as shown in Figure 5.P.6. F + Σ K FIGURE 5.P.6 Y H – G Block diagram of the feedback control system corresponding to the Watt governor linearized about the operating point. Thus a positive error in deflection of the flyball connecting arm causes a decrease in steam generated torque and vice versa. A proportional gain constant K has been inserted into the feedback loop to model the sensitivity of the change in steam generated torque to the error (s) in the deflection of the flyball connecting arm. Physically, K can be altered by changing the location of the pivot point of the lever arm that connects the sliding collar on the governor to the vertical rod attached to the throttle valve. Note that any external input affecting engine speed is represented by F(s). Project 3 PROBLEMS In the following problems, we ask the reader to supply some of the details left out of the above discussion, to analyze the closed-loop system for stability properties, and to conduct a numerical simulation of the nonlinear system. 1. Work out the details leading to Eq. (1). 2. Give physical explanations for the meaning of Eqs. (3) and (4). 3. Derive the linearized system (5) and (6). 4. Show that the transfer function of the closed-loop system linearized about the operating point is HK (s) = s 2 + 2δs + ω02 . (J s + β)(s 2 + 2δs + ω02 ) + K 5. Use the Routh criterion to show that if the gain factor K is sufficiently large, H K (s) will have two poles with positive real parts and the corresponding closed-loop system is therefore unstable. Derive an expression for K c , that value of K at which the real parts of a pair of conjugate complex poles of H K (s) are equal to 0. 6. The Nonlinear Feedback Control System. Using the relations (3) and (4), show that Eqs. (1) and (2) can be expressed in terms of φ and y as m Lφ = −γ φ − m20 L cos(θ0 ) sin(θ0 + φ) + m(0 + y)2 L sin(θ0 + φ) cos(θ0 + φ) (i) and J y = −βy − K φ, (ii) where the negative feedback loop has been incorporated into the nonlinear system. 7. Simulations. Consider the following parameter values expressed in SI units: m = 12, J = 400, L = 12 , β = 20, g = 9.8, θ0 = π/6, γ = 0.01, with 0 determined by the relation (3). (a) Using the above parameter values, construct a root locus plot of the poles of H K (s) as K varies over an interval containing K c . Verify that CONFIRMING PAGES 15:52 P1: KUF/OVY JWCL373-05-II P2: OSO/OVY QC: SCF/OVY T1: SCF JWCL373-Brannan-v1 October 21, 2010 Projects a pair of poles cross from the left half plane to the right half plane as K increases through K c . (b) Conduct computer simulations of the system (i), (ii) in Problem 6 using the above parameter values. Do this for various values of K less than K c and greater than K c while experimenting with different values of y(0) = 0 to represent departures from the equilibrium operating point. For simplicity, you may always assume that φ(0) = φ (0) = 0. Plot graphs of the 393 functions φ(t) and y(t) generated from the simulations. Note that if φ(t) wanders outside the interval [−θ0 , π − θ0 ], the results are nonphysical. (Why?) Are the results of your simulations consistent with your theoretical predictions? Discuss the results of your computer experiments addressing such issues as whether the feedback control strategy actually works, good choices for K, stability and instability, and the “hunting” phenomenon discussed above. CONFIRMING PAGES 15:52