Control, Industrial Automation and Digital Operations

Introduction

Modern industrial systems combine feedback control, programmable automation, supervisory interfaces and operational analytics. A control engineer may identify a process model, tune a PID controller, implement interlocks in a programmable logic controller (PLC), configure an operator interface, archive measurements through SCADA and use those measurements to detect degrading equipment.

These activities form a connected engineering workflow. A controller cannot compensate for an incorrectly scaled sensor; a well-designed PLC program remains unsafe without appropriate interlocks; and predictive-maintenance models are unreliable when operational data lack context or quality. This chapter therefore integrates control theory with laboratory implementation and digital operations.

Learning Objectives

After completing this chapter, students should be able to:

1. Model continuous-time open- and closed-loop systems using transfer functions.
2. Evaluate transient response, steady-state error and stability.
3. Explain and implement proportional, integral and derivative control.
4. apply explicit saturation and anti-windup logic in a digital controller.
5. Estimate process models from experimental input–output data.
6. Design PLC sequencing, interlocks and analog scaling logic.
7. Configure basic HMI and SCADA functions.
8. Extract condition indicators from operational data.
9. Evaluate predictive-maintenance results critically and safely.

1. Dynamic Systems and Feedback Control

1.1 Open- and closed-loop operation

An open-loop system applies a control action without using the measured output to correct that action. Examples include a heater operated for a fixed time and a motor driven at a preset voltage. Open-loop control is simple, but it cannot automatically reject disturbances or compensate for changing process characteristics.

A closed-loop system measures the controlled output and compares it with a reference. The resulting error influences the actuator input. In negative feedback, corrective action is intended to reduce the error.

For a linear time-invariant system with zero initial conditions, let G(s) denote the complete forward-path transfer function and H(s) the feedback-path transfer function. The closed-loop transfer function is

Here, R(s) is the reference, Y(s) is the output and s is the Laplace variable in s⁻¹. The expression assumes negative feedback, compatible signal units around the summing junction and a well-posed linear loop. For unity feedback, H(s)=1.

Figure: Negative-feedback system in which the feedback path connects directly to the lower input of the summing junction and G(s) represents the complete forward path.

1.2 Mathematical models

A physical system is often first expressed through a differential equation. Consider a thermal process:

Here, \theta(t) is temperature rise in K, q(t) is heat input in W, C is thermal capacitance in J/K, and R_th is thermal resistance in K/W. Because C(d\theta/dt) and \theta/R_th both have units of watts, the equation is dimensionally consistent.

With zero initial temperature rise, its transfer function is

This is a first-order model with steady-state gain K=R_th and time constant \tau=R_thC seconds.

Models may be obtained from physical laws, experimental identification or a combination of both. Transfer functions are appropriate for linear time-invariant systems around an operating point. Strong nonlinearities, varying parameters and large operating-range changes may require gain scheduling, nonlinear models or state-space methods.

2. Dynamic Response and Stability

2.1 First- and second-order behavior

For the stable first-order system

the response to a step of amplitude A is

The output reaches 63.2% of its total change at t=\tau. A commonly used approximate 2% settling time is t_s\approx4\tau. These results apply only to a dominant first-order response without appreciable dead time.

A standard underdamped second-order denominator is

where \zeta is the dimensionless damping ratio and \omega_n is the natural frequency in rad/s. For 0<\zeta<1, the percentage overshoot following a step is

This formula assumes a standard second-order system without zeros that substantially alter the transient response.

2.2 Stability criteria

A continuous-time linear system is asymptotically stable when every closed-loop pole lies in the open left half of the complex s-plane. A discrete-time system is asymptotically stable when all closed-loop poles lie strictly inside the unit circle in the z-plane.

For low-order systems, poles can be calculated directly. The Routh–Hurwitz criterion determines the number of right-half-plane roots without explicitly calculating them. Frequency-response methods use gain margin and phase margin to indicate relative stability. Positive margins do not guarantee satisfactory time-domain performance, particularly for nonlinear or uncertain plants, but they provide useful robustness indicators for compatible loop-transfer models.

2.3 Worked example: proportional closed-loop control

A plant has

and is controlled by a proportional controller K_c=3 under unity negative feedback. The complete forward path is

Therefore,

The closed-loop pole is s=-7/5=-1.4\ \text{s}^-1, so the system is stable. Its time constant is 5/7=0.714\ \text{s}. For a unit-step reference, the final output is 6/7=0.857, giving a steady-state error of 1/7=0.143. Proportional feedback improves speed but does not eliminate the step error in this type-0 system.

3. PID Control

3.1 Control modes

The parallel-form PID controller is

Here, e(t)=r(t)-y(t) is the control error, and u(t) is the controller output. If error is measured in engineering units E and output in units U, then K_p has units U/E, K_i has units U/(E·s), and K_d has units U·s/E.

• Proportional action responds to present error and generally increases loop speed.
• Integral action accumulates past error and can eliminate constant steady-state error when the resulting closed loop remains stable and the actuator has adequate authority.
• Derivative action anticipates error change and may improve damping, but it amplifies high-frequency measurement noise.

Practical controllers commonly apply derivative action to the measured variable rather than to the error, avoiding a derivative kick when the setpoint changes. A filtered derivative contribution may be represented by

where \tau_f>0 is the derivative-filter time constant in seconds.

Using backward-Euler discretization with sample period T_s, the corresponding update is

The sample period should be short relative to the fastest dynamics that must be controlled, while remaining compatible with sensor bandwidth, noise and processor loading.

3.2 Saturation and anti-windup

Actuators have limits. If an integral term continues growing while the command is saturated, the controller may recover slowly after the error reverses. This is called integral windup.

A non-circular back-calculation algorithm can be executed in the following explicit order at sample k:

1. Form a provisional integral state:

1. Calculate the unconstrained command:

1. Apply actuator limits:

1. Commit the integral state for the next sample:

The sequence is operationally explicit: u_k^* is computed from the provisional state, saturation is applied, and only then is I_k stored. K_aw has units s⁻¹ when I and u use the same output units. Conditional integration—suspending integration when saturation and error drive in the same direction—is another valid strategy.

3.3 Tuning and verification

PID tuning may use empirical tests, frequency response, pole placement, optimization or model-based rules. Regardless of method, tuning must be verified over expected operating conditions. Aggressive tuning can produce excessive actuator travel, sensitivity to delay and poor robustness.

A sound laboratory assessment records:

• rise and settling times;
• overshoot and steady-state error;
• disturbance rejection;
• actuator saturation and rate limits;
• sensitivity to noise and model uncertainty;
• response under manual-to-automatic transfer.

4. Experimental System Identification

4.1 Identification workflow

System identification estimates a model from measured input and output data. The principal stages are:

1. define the operating region and model purpose;
2. select an excitation that is informative but safe;
3. synchronize and record inputs, outputs and disturbances;
4. remove invalid data and document filtering;
5. estimate model parameters;
6. validate the model using data not used for estimation.

A common process model is the first-order-plus-dead-time model:

Here, K=\Delta y/\Delta u is the steady-state process gain, L is dead time in seconds and \tau is the time constant in seconds. For a monotonic step response, \tau is the time from the end of the estimated dead time until the output reaches 63.2% of its final change.

4.2 Worked identification and PI design

A heater command increases from 30% to 40%. Temperature eventually rises from 60°C to 68°C. The observed dead time is 12 s, and the temperature reaches

at 52 s after the input step. Thus,

For the controller form u=K_c(e+(1/T_i)\int e\,dt), one approximate IMC-based PI rule for an FOPDT model is

where \lambda>0 is the desired closed-loop time constant. This simplified rule is an approximate design relationship, not a universal optimum. Choosing \lambda=40\ \mathrm{s} gives

and T_i=40\ \mathrm{s}, so K_i=K_c/T_i=0.0240\ \%/(^\circ\mathrm{C}\cdot s). The result must be tested with saturation, disturbances and independent validation data.

Model accuracy can be summarized by root-mean-square error:

where y_k is measured output, \hat y_k is predicted output and N is the number of validation samples. RMSE retains the output's engineering unit, but it does not reveal bias, autocorrelated residuals or poor performance in specific operating regions.

5. PLC-Based Industrial Automation

5.1 PLC scan operation

A PLC repeatedly performs an operating cycle that normally includes input acquisition, program execution, output update and diagnostics or communications. Exact behavior depends on the PLC architecture; high-speed tasks and interrupts may run asynchronously.

IEC 61131-3 programming languages include ladder diagram, function block diagram, structured text and sequential function chart. Language choice should reflect the task, maintainability requirements and organizational standards.

Control logic should distinguish among:

• permissives, which must be true before equipment may start;
• interlocks, which prevent or terminate hazardous operation;
• alarms, which notify operators of abnormal conditions;
• trips, which move equipment toward a defined safe state.

An emergency stop or safety-critical protective function should not rely solely on ordinary PLC logic or a non-safety-rated communication network. Applicable machinery, electrical and functional-safety requirements determine whether hardwired safety circuits or safety-rated controllers are required.

5.2 Analog scaling

For a linear transmitter with raw input x spanning x_min to x_max, the engineering value is

The formula assumes a linear calibrated sensor and consistent units. For a 4–20 mA temperature transmitter representing 0–100°C, a current of 13.6 mA gives

PLC logic should also identify under-range and over-range signals rather than blindly extrapolating them as valid measurements.

5.3 Sequence design

Sequential automation should define states and transitions explicitly. A motor-start sequence might contain IDLE, STARTING, RUNNING, STOPPING and FAULT states. Every transition should have a clear condition, timeout behavior and recovery rule. Outputs should be assigned deterministically so that an old command cannot persist unintentionally.

Laboratory tests should include normal operation, invalid command combinations, sensor failure, communication loss, power restoration and emergency shutdown. Simulation should precede connection to energized machinery where practicable.

6. HMI and SCADA Systems

An HMI provides local or remote visualization and operator interaction. SCADA extends this capability across controllers and remote assets by supporting data acquisition, supervisory commands, alarms, trends and historical storage.

Figure: Operational data flow from field devices through control and supervisory systems to maintenance decisions.

Effective HMI design emphasizes abnormal conditions rather than decorative realism. Displays should use consistent units, restrained color, readable trends, clear equipment states and confirmation for consequential commands.

Alarm management requires priorities based on consequence and required operator response. Nuisance, duplicate and chattering alarms reduce situational awareness. Each alarm should have a defined cause, consequence, response and rationalized setpoint.

SCADA communications may use protocols such as Modbus TCP, industrial Ethernet protocols or OPC UA. Interoperability does not itself provide security. Appropriate measures include network segmentation, least-privilege accounts, authenticated remote access, backups, patch governance, logging and controlled configuration changes.

7. Operational Data and Condition Monitoring

7.1 From measurements to indicators

Condition monitoring uses measurements to assess equipment health. Common signals include vibration, acoustic emission, motor current, oil debris, temperature and process performance.

For N sampled values x_k, the root-mean-square value is

RMS represents signal energy and retains the signal's unit. For vibration velocity, it may be expressed in mm/s. Thresholds must be selected for the machine type, measurement location and operating condition; a single limit is not universally valid.

The crest factor is

It is dimensionless and may highlight impulsive behavior. Spectral analysis can identify frequencies associated with shaft rotation, imbalance, misalignment, gear mesh or bearing defects. Valid interpretation requires adequate sampling frequency, anti-alias filtering and suitable record length.

A standardized baseline score is

where \mu and \sigma are the mean and standard deviation of a representative healthy baseline. This score assumes the baseline remains relevant. A three-standard-deviation threshold is not automatically appropriate for non-normal or highly autocorrelated industrial data.

7.2 Predictive maintenance

Condition-based maintenance responds to measured condition. Predictive maintenance additionally estimates future degradation or remaining useful life. Approaches include:

• physics-based degradation models;
• statistical trend and survival models;
• supervised machine learning;
• anomaly detection when failure labels are scarce;
• hybrid models combining engineering knowledge and data.

AI models should be evaluated with time-ordered validation rather than random splitting when future behavior is being predicted. Data leakage can produce unrealistically high accuracy. Changes in operating regime, sensor replacement or maintenance practice can create model drift.

A predictive model should support, not replace, accountable engineering judgment. False negatives may create safety risk; false positives may cause unnecessary shutdowns. Performance should therefore be evaluated using consequence-sensitive metrics rather than accuracy alone.

8. Illustrative Integrated Case: Pumping Station Modernization

Background

This fictional case concerns a water pumping station with a variable-speed pump, pressure transmitter, PLC and SCADA connection. Operators reported pressure oscillation, frequent high-pressure alarms and unexpected bearing maintenance.

Problem

The existing controller used aggressive integral action, had no documented anti-windup strategy and continued integrating when the drive reached its speed limit. The SCADA historian stored pressure but not controller mode, output saturation or pump operating state. Vibration measurements were reviewed manually and irregularly.

Analysis

A safe speed step from 45% to 50% increased steady pressure from 280 to 300 kPa. The estimated dead time was 2 s and the time constant was 8 s. The identified process gain was

Using the approximate IMC PI rule with \lambda=8\ \mathrm{s} gave

Thus, K_i=K_c/T_i=0.025\ \%/(kPa\cdot s). The PLC task period was 0.1 s, sufficiently shorter than the identified 8 s process time constant for this application, subject to final closed-loop testing.

Historical healthy vibration had mean RMS velocity 2.00 mm/s and standard deviation 0.15 mm/s under comparable speed and loading. A provisional statistical warning level was set at

This was treated as an investigation trigger, not a universal safety limit.

Approach

The engineering team:

1. implemented the explicit back-calculation sequence described earlier;
2. added speed, pressure and dry-running permissives;
3. configured safe behavior for invalid transmitter signals;
4. redesigned the HMI to display mode, setpoint, output and saturation;
5. rationalized pressure alarms and added delays against chattering;
6. archived pump state, speed, pressure, current and vibration features;
7. compared vibration only among similar operating regimes;
8. required maintenance review before generating a work order.

Outcome

Commissioning tests showed reduced pressure oscillation and faster recovery after actuator saturation. Trending identified a sustained vibration increase above the operating-condition baseline. Inspection found developing coupling misalignment before severe secondary damage occurred. Because the case is illustrative, these outcomes demonstrate the workflow rather than claim a guaranteed economic benefit.

Learning Points

• Controller tuning, saturation handling and actuator constraints must be considered together.
• Context tags are essential for interpreting operational data.
• Statistical warnings require engineering confirmation.
• HMI and historian design affect both control performance and maintenance quality.
• Safety functions must remain independent of non-safety analytics where required.

9. Laboratory Practice and Best Practices

A coherent laboratory sequence can include:

1. Open-loop characterization: apply safe input steps and estimate gain, delay and time constant.
2. Closed-loop testing: compare proportional, PI and PID responses using identical references and disturbances.
3. Digital implementation: select sample time, apply derivative filtering, saturation and anti-windup.
4. PLC automation: implement state logic, permissives, interlocks and fault recovery.
5. HMI/SCADA integration: configure tags, trends, alarms and historical storage.
6. Condition monitoring: calculate RMS and spectral indicators and establish an operating-condition baseline.

Before energization, students should verify wiring, signal ranges, grounding, actuator direction, emergency-stop operation and output limits. Changes should be version controlled. Raw data, calibration information, controller parameters and test conditions should be retained so results are reproducible.

Important limitations include model uncertainty, sensor drift, communication delay, quantization, nonlinear actuators, missing data and changing production conditions. No single laboratory test proves universal stability or reliability. Conclusions must remain within the tested operating envelope.

10. Governance, Ethics and Contemporary Developments

Industrial automation is cyber-physical: software errors can damage equipment or injure people. Engineers have professional duties to document assumptions, preserve audit trails, protect credentials and avoid bypassing safeguards for convenience.

Operational data may reveal employee behavior, production volumes or commercially sensitive information. Collection and retention should therefore have legitimate purposes, appropriate access controls and defined governance. Automated maintenance recommendations should be explainable enough for responsible review, especially when failure consequences are severe.

Contemporary developments include edge analytics, digital twins, wireless condition monitoring, interoperable information models and machine-learning-assisted diagnostics. Digital twins range from simple synchronized models to high-fidelity virtual representations. Their value depends on model validation, configuration control and trustworthy data—not merely on visualization sophistication.

Chapter Summary

Control and digital operations integrate mathematical models, feedback, automation and operational analytics. Transfer functions describe linear dynamic relationships, while closed-loop stability depends on pole locations and adequate robustness. PID controllers combine present, accumulated and predicted error effects, but practical implementation requires filtering, sampling, saturation management and an unambiguous anti-windup update sequence.

System identification converts experimental data into usable process models, which must be independently validated. PLCs implement deterministic control, sequencing and interlocks, while HMI and SCADA systems support visualization, alarms, supervisory control and historical data. Operational measurements can provide condition indicators and predictive-maintenance evidence, provided that sampling, context, validation and governance are adequate. Safe industrial practice requires technical verification, cybersecurity, documentation and accountable human judgment.

Key Terms

Open-loop control: Control without output-based corrective feedback.

Closed-loop control: Control in which measured output influences the control action.

Transfer function: Ratio of output to input transforms for a linear time-invariant system under stated initial conditions.

Negative feedback: Feedback arranged so that measured output is subtracted from the reference.

Stability: The property that system responses remain bounded or converge according to the applicable stability definition.

Time constant: Parameter describing the response speed of a first-order system.

Dead time: Delay between an input change and the beginning of its observable output effect.

PID controller: Controller combining proportional, integral and derivative actions.

Integral windup: Excessive integral accumulation during actuator saturation.

Anti-windup: Logic that limits or corrects integral accumulation during saturation.

System identification: Estimation of a dynamic model from measured data.

PLC: Industrial controller designed for reliable real-time logic and automation.

Permissive: Condition that must be satisfied before an operation may begin.

Interlock: Logic that prevents or terminates an unsafe or prohibited action.

HMI: Interface through which operators observe and interact with a process.

SCADA: Supervisory system for data acquisition, alarms, trends and remote control.

Historian: Time-series database used to store operational process data.

Condition monitoring: Assessment of equipment health using measured indicators.

Predictive maintenance: Maintenance strategy that forecasts degradation or future failure risk.

Digital twin: Maintained digital representation linked to the state or behavior of a physical asset.

Review Questions

Short Answer Questions

1. Distinguish between open-loop and closed-loop control.
2. State the continuous-time and discrete-time pole conditions for asymptotic stability.
3. Why is derivative action commonly applied to the measured variable?
4. What is the difference between a PLC permissive and an interlock?
5. Why must condition-monitoring data be grouped by operating condition?

Long Answer Questions

1. Derive the negative-feedback closed-loop transfer function and state its assumptions.
2. Explain proportional, integral and derivative actions, including their benefits and practical limitations.
3. Describe a complete experimental system-identification workflow for an FOPDT process.
4. Explain how PLC, HMI, SCADA and historian functions cooperate in an industrial automation system.
5. Discuss cybersecurity, safety, ethical and governance responsibilities in predictive industrial operations.

Analytical / Application Questions

1. A unity-feedback plant is G_p(s)=3/(4s+1) and the proportional gain is 2. Determine the closed-loop transfer function, pole, time constant and unit-step steady-state error.
2. Design PLC states, permissives, interlocks and alarm responses for an automatically filled liquid tank with high- and low-level switches.
3. Given synchronized motor current, speed and vibration data, propose a method to detect deterioration while minimizing false alarms caused by changes in load.

Multiple Choice Questions with Answers

Question 1

For a negative-feedback loop with complete forward path G(s) and feedback path H(s), what is the closed-loop transfer function?

A. G(s)/(1-G(s)H(s))

B. G(s)/(1+G(s)H(s))

C. H(s)/(1+G(s))

D. G(s)H(s)/(1+H(s))

Correct Answer: B. G(s)/(1+G(s)H(s))

Question 2

A plant G_p(s)=2/(5s+1) uses proportional gain 3 and unity negative feedback. What is the closed-loop pole?

A. s=-0.2\ \text{s}^-1

B. s=-0.6\ \text{s}^-1

C. s=-1.4\ \text{s}^-1

D. s=+1.4\ \text{s}^-1

Correct Answer: C. s=-1.4\ \text{s}^-1

Question 3

In the chapter's explicit back-calculation anti-windup algorithm, which operation occurs immediately before the revised integral state is committed?

A. The actuator command is saturated

B. The previous error is discarded

C. The setpoint is forced to zero

D. The derivative gain is doubled

Correct Answer: A. The actuator command is saturated

Question 4

For a monotonic first-order-plus-dead-time step response, the time constant is measured from the end of the dead time until the output reaches approximately what percentage of its final change?

A. 36.8%

B. 50.0%

C. 63.2%

D. 95.0%

Correct Answer: C. 63.2%

Question 5

A linear 4–20 mA transmitter represents 0–100°C. What temperature corresponds to 13.6 mA?

A. 40°C

B. 50°C

C. 60°C

D. 85°C

Correct Answer: C. 60°C

Question 6

Which activity is normally part of a PLC operating cycle?

A. Replacing field sensors physically

B. Reading inputs, executing logic and updating outputs

C. Retuning every controller automatically on each scan

D. Rewriting the PLC firmware from SCADA

Correct Answer: B. Reading inputs, executing logic and updating outputs

Question 7

What is the asymptotic-stability condition for the poles of a discrete-time linear system?

A. They must lie strictly inside the unit circle

B. They must lie in the open right half-plane

C. They must all lie on the unit circle

D. They must have positive real parts

Correct Answer: A. They must lie strictly inside the unit circle

Question 8

What is the RMS value of the two samples 3 and 4?

A. 3.00

B. 3.50

C. \sqrt{12.5}\approx3.54

D. 5.00

Correct Answer: C. \sqrt{12.5}\approx3.54

Question 9

Which industrial component is primarily intended to retain time-stamped operational measurements for later trend and maintenance analysis?

A. Historian

B. Contactor

C. Emergency-stop switch

D. Current transformer

Correct Answer: A. Historian

Question 10

A safe process test changes pump speed from 45% to 50% and pressure from 280 kPa to 300 kPa. What is the steady-state process gain?

A. 0.25 %/kPa

B. 4 kPa/%

C. 20 kPa/%

D. 55 kPa/%

Correct Answer: B. 4 kPa/%