1. Foundations of Machine Dynamics and Control

Introduction and Context

Machine dynamics and control provides the conceptual bridge between mechanical design and the actual motion of a machine. A mechanism may satisfy its geometric requirements and possess adequate strength, yet still perform poorly because of vibration, resonance, excessive settling time, instability, or inadequate tracking. Dynamics explains how forces, torques, inertia, damping, and compliance produce motion. Control explains how that motion can be measured, compared with a desired objective, and regulated through purposeful actuation.

At postgraduate level, the subject is not limited to solving equations of motion. It requires the engineer to decide what should be modelled, which variables are important, how assumptions affect predictions, and whether a passive or active design is appropriate. The same physical machine can be represented as a rigid-body mechanism, a flexible multibody system, a state-space model, or a closed-loop control system, depending on the design question.

A useful engineering workflow therefore begins with the physical system, proceeds through modelling and dynamic analysis, and ends with validation and control implementation. The foundations developed in this chapter support later study of vibration, mechanism dynamics, robotics, servo drives, feedback control, and mechatronic machine design.

Learning Objectives

After studying this chapter, the reader should be able to:

1. distinguish statics, kinematics, kinetics, dynamics, and control;

2. classify mechanical systems by linearity, time dependence, energy behaviour, and causality;

3. determine degrees of freedom and select suitable generalized coordinates;

4. formulate equilibrium and explain transient, steady-state, and stability behaviour;

5. interpret open-loop and closed-loop feedback structures;

6. relate poles, eigenvalues, damping, bandwidth, and performance measures;

7. develop a machine dynamics–control design workflow;

8. identify modelling limitations, implementation risks, and professional responsibilities.

From Motion Description to Controlled Behaviour

Kinematics describes position, velocity, and acceleration without considering the forces that cause motion. Kinetics relates motion to force and torque through laws such as Newton’s second law and the Euler rotational equation. Dynamics is often used in the broader sense to include both kinematics and kinetics, together with inertial, elastic, and dissipative effects. Statics is the special case in which acceleration is zero and the resultant forces and moments are in equilibrium.

Control adds a functional objective. A machine is not merely required to move; it may need to follow a prescribed trajectory, maintain speed under changing load, reject disturbances, or position a tool within a specified tolerance. The controlled variable is commonly called the output, while the desired value is the reference or command. Sensors measure the output, a controller computes an action, and an actuator produces force, torque, flow, or electrical input.

A simple rotational drive illustrates the distinction. For a motor shaft with inertia J, viscous damping c, applied motor torque T_m, and load torque T_L,

J(d²θ/dt²) + c(dθ/dt) = T_m - T_L.

This equation is a dynamic model. If a controller adjusts T_m so that θ follows a commanded position θ_r, the model becomes part of a closed-loop control system. The physical equation has not changed; the information flow and applied input have.

Important Concepts and Terminology

A system is a collection of interacting elements with defined inputs, outputs, and internal variables. A plant is the physical process being controlled. An input is an externally imposed quantity, while an output is a measured or otherwise observed response. A disturbance is an input not intentionally commanded, such as friction variation, cutting force, or load torque.

A state is the minimum set of variables required, together with future inputs, to determine future system behaviour. For a second-order mass–spring–damper system, displacement and velocity form a possible state vector. The state vector is written as x(t), and a general continuous-time state model is

ẋ = f(x,u,t), y = g(x,u,t).

For a linear time-invariant system, this becomes

ẋ = Ax + Bu, y = Cx + Du,

where A, B, C, and D are system matrices. The matrix A governs internal dynamics; its eigenvalues are particularly important for stability.

The degree of freedom (DOF) is the number of independent coordinates needed to specify the configuration of a system. A free rigid body in three-dimensional space has six DOF: three translations and three rotations. A pendulum constrained to rotate about a fixed pin has one. DOF is a geometric and constraint-related concept; it should not be confused automatically with system order, although the two may coincide in simple mechanical models.

A generalized coordinate is an independent coordinate selected to describe configuration. Generalized forces include ordinary forces and moments represented consistently with those coordinates. Poor coordinate selection can introduce unnecessary constraints, singularities, or numerical ill-conditioning.

Equilibrium is a state in which the system remains unchanged when subjected to the specified inputs and constraints. For a static mechanical system, net force and net moment vanish. In a dynamic system, equilibrium may be a constant state or, in a rotating frame, a steady operating condition. Linearization about an equilibrium is central to small-signal analysis.

A transient response is the portion of a response associated with initial conditions or changes in input and which generally decays or evolves with time. The steady-state response is the limiting behaviour after transient effects have disappeared, if such a limit exists. A system may have a steady periodic response rather than a constant one when subjected to a periodic input.

Classifying Dynamic Systems

Classification is useful because it determines which analytical tools and assumptions are appropriate.

A system is continuous-time when variables evolve continuously, whereas a discrete-time model updates at sampling instants. A digital controller commonly interacts with a continuous mechanical plant through sampling, computation, and zero-order-hold actuation.

A linear system obeys superposition: the response to a sum of inputs equals the sum of the individual responses, and scaling an input scales its response. A nonlinear system does not generally satisfy this property. Common nonlinearities include Coulomb friction, backlash, saturation, dead zones, geometric stiffness, contact, and aerodynamic drag. Linear models remain valuable when they represent local behaviour near an operating point, but their domain of validity must be stated.

A time-invariant system has parameters and governing relationships that do not explicitly change with time. A system with changing mass, temperature-dependent stiffness, or scheduled controller gains may be time-varying. A system is causal when its output depends only on present and past inputs. Physical machines are causal; noncausal mathematical filters are useful for offline analysis but cannot be implemented directly in real time.

Systems may also be classified as lumped-parameter or distributed-parameter. A lumped model represents inertia, stiffness, and damping at discrete elements. A flexible shaft or beam is fundamentally distributed, possessing theoretically many modes, although a finite modal approximation may be adequate for a specified frequency range.

Energy provides another important classification. An energetic storage element stores kinetic or potential energy; damping and friction dissipate energy; actuators and external forces supply energy. Passive mechanical components cannot generate net energy, whereas active control can inject energy and therefore requires careful stability analysis.

Degrees of Freedom, Constraints, and Modelling Choice

Determining DOF is often the first substantive modelling decision. In a planar mechanism, a free rigid body has three DOF. Each independent lower-pair constraint typically removes one or more relative motions, but the result depends on the joint type and on special geometric configurations. Gruebler–Kutzbach-type mobility formulas can provide a first estimate, but they must be checked against redundant constraints, overconstraints, and singular positions.

A machine may have apparent coordinates that are not independent. For example, the positions of several links in a four-bar mechanism are related by loop-closure equations. One may retain all link coordinates and impose constraints using Lagrange multipliers, or eliminate dependent coordinates and use a minimal set. The first approach preserves a systematic formulation; the second often produces a smaller numerical model.

DOF also depends on the purpose of analysis. A robot arm may have six rigid-body joint coordinates for positioning, but if link flexibility is relevant, elastic deflections introduce additional states. A motor-drive model may neglect structural modes for low-bandwidth speed control, yet those modes become essential when a high-bandwidth position loop is designed.

Equilibrium, Stability, and Response

Equilibrium alone does not guarantee acceptable behaviour. Consider a mass–spring–damper system:

mẍ + cẋ + kx = F(t).

For free motion, the characteristic equation is ms² + cs + k = 0. With positive mass, stiffness, and damping, the natural response decays. The undamped natural frequency is ω_n = √(k/m), and the damping ratio is ζ = c/(2√(km)). The damping ratio indicates the qualitative transient form: underdamped systems oscillate while decaying, critically damped systems return rapidly without oscillation in the ideal model, and overdamped systems return without oscillation but more slowly.

Stability concerns the response to perturbations. An equilibrium is locally asymptotically stable if sufficiently small disturbances produce responses that return to equilibrium as time tends to infinity. It is unstable if some arbitrarily small disturbance produces a response that departs from the equilibrium. Marginal cases, such as an ideal undamped oscillator, neither decay nor grow and require careful interpretation.

For a continuous linear time-invariant system, internal asymptotic stability requires all poles of the transfer function or eigenvalues of A to lie in the open left half of the complex plane, subject to proper treatment of pole-zero cancellations and internal modes. In discrete time, stable poles lie inside the unit circle. This criterion is powerful but applies to the model and operating condition used; unmodelled flexible modes, saturation, delays, and parameter variation can invalidate conclusions.

Performance is described using measurable criteria. For a step response, rise time indicates how quickly the output initially responds, peak time identifies the first maximum, overshoot measures excursion beyond the final value, and settling time measures the time to enter and remain within a specified tolerance band. Steady-state error is the long-term difference between reference and output. Bandwidth characterizes the frequency range over which a system tracks inputs adequately, while robustness expresses the ability to retain acceptable performance despite uncertainty and disturbances.

Feedback as an Engineering Architecture

In open-loop control, the command is generated without measuring the actual output. Open-loop operation is simple and can be effective when the plant is predictable, but it cannot automatically correct load changes or modelling errors.

In closed-loop control, a sensor measures the output and the controller acts on an error signal. For a standard negative-feedback loop with forward transfer function G(s) and feedback transfer function H(s), the closed-loop transfer function is

T(s) = G(s)/[1 + G(s)H(s)].

The denominator shows why feedback changes stability and dynamics, not merely accuracy. Increasing loop gain may reduce tracking error and disturbance sensitivity over a useful frequency range, but it can also increase overshoot, excite unmodelled modes, amplify measurement noise, or cause instability through delay and phase lag.

A proportional–integral–derivative controller is commonly represented by

C(s) = K_p + K_i/s + K_d s,

with practical derivative filtering. Proportional action responds to present error, integral action accumulates error and can remove steady-state error for suitable plants, and derivative action anticipates error variation and can improve damping. None is universally beneficial: integral action can cause windup during actuator saturation, while derivative action is sensitive to noise.

Feedback architecture should be selected with the physical machine in mind. Cascaded current, speed, and position loops are common in electric drives because inner loops can regulate faster variables and present a simpler effective plant to outer loops. Feedforward can improve tracking when the required motion and load are known, but feedback remains important for correcting uncertainty.

A Machine Dynamics–Control Design Workflow

A reliable workflow is iterative rather than strictly linear.

First, define the performance objective and operating envelope. Position accuracy, allowable vibration, response time, payload, energy consumption, safety constraints, and disturbance conditions should be quantified. Ambiguous objectives produce ambiguous designs.

Second, identify the physical architecture: bodies, joints, actuators, sensors, transmissions, structural members, contact interfaces, and likely disturbances. Select generalized coordinates and establish sign conventions.

Third, formulate the dynamic model using Newton–Euler equations, energy methods, virtual work, or constrained multibody methods. Include only justified simplifications. Record parameters, units, assumptions, and neglected phenomena.

Fourth, determine equilibrium points and linearize where appropriate. Examine natural frequencies, modes, poles, controllability, observability, and likely resonances. Validate the model against analytical limiting cases, experiments, or higher-fidelity simulation.

Fifth, select the control architecture and design controller gains. Evaluate time-domain and frequency-domain performance, actuator limits, sensor noise, computational delay, and robustness to parameter variation.

Sixth, implement progressively: simulation, hardware-in-the-loop, laboratory testing, and controlled field operation. Instrumentation should measure not only the commanded output but also safety-critical variables such as temperature, current, vibration, and travel limits.

Finally, compare measured results with predictions and update the model. Differences are not merely errors to be hidden; they reveal friction, compliance, backlash, saturation, or disturbances that may require redesign.

Applied Case: Precision Positioning of a Flexible Drive

Background

Consider an illustrative machine consisting of a servo motor driving a precision carriage through a long, lightly stiffened belt. The carriage must follow repeated position commands while carrying variable payloads. A rigid-body model predicts satisfactory settling time, but prototype tests show oscillation after rapid moves.

Situation and Analysis

The motor inertia and carriage mass form the dominant low-frequency dynamics, while belt elasticity introduces a second mode. The initial controller was tuned using only the rigid-body model. Its bandwidth lies close to the belt’s resonance, so the controller supplies energy at precisely the frequency where the mechanical system stores and exchanges elastic energy. The measured response exhibits overshoot and a long decay, indicating insufficient effective damping and poor robustness.

The problem is not solved simply by increasing proportional gain. Higher gain improves low-frequency stiffness but further excites the flexible mode. Sensor placement also matters: a motor-mounted encoder does not directly measure carriage position, so belt stretch remains partly invisible to the feedback loop.

Approach

The design team extends the model to include motor inertia, carriage mass, belt stiffness, damping, actuator limits, encoder quantization, and computational delay. A frequency-response test identifies the flexible resonance. The position-loop bandwidth is reduced below the resonance, while a faster inner velocity loop is retained. A notch filter is introduced cautiously to attenuate the identified mode, and feedforward acceleration is added to reduce tracking error during commanded motion. Anti-windup logic is included because the motor saturates during aggressive moves.

A second carriage encoder is considered for direct output feedback. Before adoption, observability, sensor cost, alignment, and failure modes are assessed. The revised controller is tested in simulation, then under gradually increased payload and command speed.

Outcome

The illustrative redesign produces a slower but better-damped response, with substantially reduced oscillation and more predictable performance across payload conditions. The remaining limitation is that the notch filter depends on resonance frequency; temperature and payload changes may shift that frequency. Consequently, acceptance testing includes parameter variation rather than a single nominal test.

Learning Points

The case demonstrates that control performance is constrained by physical dynamics. A controller cannot compensate indefinitely for inadequate stiffness, unmodelled modes, sensor placement, or actuator limits. Model fidelity should be matched to control bandwidth, and performance specifications should include robustness, not only nominal rise time.

Practical, Professional, and Digital Applications

The principles of this chapter apply to machine tools, industrial robots, vehicle suspensions, turbines, winders, hard-disk drives, semiconductor stages, prosthetic devices, and automated production systems. In each case, the engineer must connect mechanical design variables—mass, inertia, stiffness, damping, friction, geometry, and transmission ratio—to control variables such as gain, bandwidth, sampling rate, and filtering.

Computer-aided engineering tools support multibody simulation, finite-element modal analysis, system identification, control synthesis, and hardware-in-the-loop testing. Digital twins can combine measured data with a dynamic model for condition monitoring and predictive maintenance. Data-driven and adaptive methods can estimate changing parameters or compensate for nonlinearities, but they do not remove the need for physical constraints, stability arguments, validation, and fail-safe design. Machine-learning components should be bounded by supervisory logic when their actions affect safety-critical motion.

Challenges, Limitations, and Professional Responsibility

The most common modelling mistake is false precision: a mathematically detailed model may still be wrong if friction, joints, contact, or boundary conditions are poorly represented. Conversely, an unnecessarily complex model can obscure dominant effects and make controller design impractical. Linearization can conceal saturation, backlash, bifurcation, and large-amplitude behaviour. Numerical simulation can also give misleading confidence when initial conditions, solver tolerances, or parameter values are unrealistic.

Control systems introduce risks including unstable tuning, sensor failure, actuator saturation, cyber-physical interference, and unexpected interactions between machines. Safety functions should not depend on a single software loop or an unverified learned model. Engineers must document assumptions, trace requirements to tests, maintain calibration records, protect access to programmable controllers, and comply with applicable machinery, electrical, functional-safety, and electromagnetic-compatibility requirements.

Best practice is to maintain a model hierarchy, use dimensional checks and energy checks, identify uncertainty explicitly, test extreme operating conditions, separate safety interlocks from performance control, and review both nominal and failure behaviour. The correct design is not necessarily the fastest one; it is the one that achieves the required function with acceptable stability, robustness, safety, maintainability, and cost.

Current Developments and Emerging Directions

Modern machine dynamics increasingly involves cyber-physical systems in which mechanical structures, embedded computation, networks, and sensors are designed together. Real-time estimation, active vibration control, digital commissioning, and condition-based maintenance are becoming routine in advanced manufacturing. Flexible robots and collaborative machines require control methods that account for compliance and human interaction rather than treating the mechanism as perfectly rigid.

Research directions include nonlinear and robust control, online system identification, energy-aware motion planning, distributed sensing, and physics-informed machine learning. These developments strengthen rather than replace foundational modelling. The engineer who understands equilibrium, DOF, poles, feedback, and uncertainty is better positioned to judge when an advanced method is justified and when a simpler, verifiable design is preferable.

Chapter Summary

Machine dynamics describes how forces, torques, inertia, stiffness, damping, constraints, and disturbances determine motion; control regulates that motion toward a specified objective. Systems can be classified by linearity, time dependence, causality, energy behaviour, and lumped or distributed representation. Degrees of freedom and generalized coordinates establish the structure of a model, while equilibrium and linearization support local analysis.

Transient response, steady-state behaviour, poles, damping, bandwidth, and stability are central performance concepts. Feedback can improve accuracy and disturbance rejection but may also excite resonances or create instability. A sound workflow defines requirements, builds and validates an appropriate model, analyses modes and stability, designs a controller, tests implementation limits, and iterates using measured evidence. Professional practice requires robustness, safety, documentation, and awareness of model limitations.

Key Terms

Dynamics: Study of motion together with the forces and moments causing it.

Kinematics: Description of position, velocity, and acceleration without considering causative forces.

Kinetics: Relation between motion and applied forces or torques.

Plant: Physical process or machine acted upon by a controller.

State: Minimum information needed to determine future behaviour with known inputs.

Degree of freedom: Number of independent coordinates required to specify configuration.

Generalized coordinate: Independent coordinate selected to describe a system configuration.

Equilibrium: Condition in which the system remains unchanged under specified conditions.

Transient response: Time-dependent response associated with initial conditions or input changes.

Steady-state response: Long-term response after transient effects have vanished or settled.

Stability: Property describing whether perturbations decay, persist, or grow.

Natural frequency: Frequency at which a system tends to oscillate in free motion.

Damping ratio: Dimensionless measure of damping relative to critical damping.

Feedback: Use of measured output information to influence system input.

Disturbance: Uncommanded input that affects system behaviour.

Bandwidth: Frequency range over which specified tracking or response quality is maintained.

Controllability: Ability to drive system states using available inputs.

Observability: Ability to infer internal states from available measurements.

Robustness: Ability to maintain acceptable behaviour despite uncertainty and variation.

Review Questions

Short Answer Questions

1. Distinguish kinematics, kinetics, and dynamics.

2. What is meant by a generalized coordinate?

3. Define transient response and steady-state response.

4. Why can feedback both improve performance and reduce stability margin?

5. State the continuous-time pole condition for asymptotic stability of a linear time-invariant system.

Long Answer Questions

1. Explain how degrees of freedom and constraints influence the formulation of a machine model.

2. Discuss the classification of dynamic systems by linearity, time dependence, causality, and energy behaviour.

3. Derive and interpret the main parameters of a mass–spring–damper system.

4. Explain the machine dynamics–control design workflow from requirements definition to validation.

5. Discuss the roles of proportional, integral, and derivative action, including their practical limitations.

Analytical / Application Questions

1. A servo system has excellent rigid-body step-response performance but oscillates when a flexible transmission is installed. Identify the likely mechanism, propose modelling changes, and recommend two control-design responses.

2. A machine must track position accurately under variable load, but its actuator saturates during rapid commands. Explain how saturation affects feedback behaviour and propose suitable design and testing measures.

3. A designer proposes increasing loop gain to eliminate steady-state error. Analyse the possible benefits and risks, considering resonance, noise, delay, robustness, and integral action.

22. Multiple Choice Questions with Answers

Question 1

Which statement best distinguishes dynamics from kinematics?

A. Dynamics describes motion without forces, while kinematics describes forces

B. Dynamics relates motion to forces and moments, while kinematics describes motion variables

C. Dynamics applies only to static systems, while kinematics applies only to flexible systems

D. Dynamics is limited to control algorithms, while kinematics is limited to sensors

Correct Answer: B. Dynamics relates motion to forces and moments, while kinematics describes motion variables

Question 2

A planar rigid body that is free to move has how many degrees of freedom?

A. 1

B. 2

C. 3

D. 6

Correct Answer: C. 3

Question 3

For a continuous-time linear time-invariant system, asymptotic stability generally requires its poles to lie in the:

A. Open right half-plane

B. Open left half-plane

C. Unit circle only

D. Imaginary axis only

Correct Answer: B. Open left half-plane

Question 4

Which feature most directly distinguishes closed-loop control from open-loop control?

A. Use of a mechanical transmission

B. Use of feedback from a measured output

C. Operation with continuous-time signals

D. Presence of an actuator

Correct Answer: B. Use of feedback from a measured output

Question 5

For a mass–spring–damper system, increasing damping while mass and stiffness remain constant generally:

A. Increases oscillatory overshoot in the ideal model

B. Reduces oscillatory behaviour and changes the transient response

C. Removes the system’s inertia

D. Makes the system noncausal

Correct Answer: B. Reduces oscillatory behaviour and changes the transient response

Question 6

Why may increasing feedback gain destabilize a machine with flexible transmission?

A. It eliminates all sensor information

B. It can increase excitation near an unmodelled resonance

C. It always reduces bandwidth to zero

D. It converts every plant into an open-loop system

Correct Answer: B. It can increase excitation near an unmodelled resonance

Question 7

Which variable pair can serve as a state for a simple mass–spring–damper system?

A. Mass and stiffness

B. Displacement and velocity

C. Force and damping coefficient only

D. Rise time and bandwidth

Correct Answer: B. Displacement and velocity

Question 8

What is a principal disadvantage of integral action in a controller?

A. It cannot respond to accumulated error

B. It may cause windup when the actuator saturates

C. It always increases measurement noise directly

D. It removes the need for feedback

Correct Answer: B. It may cause windup when the actuator saturates

Question 9

In the illustrative flexible-drive case, why was a rigid-body model inadequate?

A. It contained too many elastic modes

B. It neglected the belt resonance relevant to the control bandwidth

C. It included sensor noise twice

D. It assumed the machine was discrete-time

Correct Answer: B. It neglected the belt resonance relevant to the control bandwidth

Question 10

Which practice is most appropriate when validating a machine dynamics and control design?

A. Test only the nominal parameter set

B. Avoid documenting modelling assumptions

C. Compare predictions with measurements across relevant operating conditions

D. Increase controller gain until oscillation appears

Correct Answer: C. Compare predictions with measurements across relevant operating conditions