Major Project / Dissertation

Chapter 11
Safety, Standards, Ethics and Governance

Bachelors Program in Electrical Engineering
Major Project / Dissertation   •   Bachelor

Safety, Standards, Ethics and Governance

Introduction

An electrical engineering major project is not complete when a prototype operates or a simulation produces an attractive graph. It is complete only when the work is safe, technically credible, legally defensible, ethically conducted and responsibly communicated. Safety, standards, ethics and governance therefore form an integrated professional discipline rather than a final compliance checklist.

Electrical engineers work with hazards that may be immediate, such as electric shock, arc flash, fire and mechanical injury, and hazards that may be less visible, such as electromagnetic interference, privacy loss, biased automated decisions and insecure control systems. A project may also affect people who did not design it: operators, maintenance staff, users, bystanders, patients, communities and future engineers who rely on its documentation.

Standards provide shared technical expectations, but they do not replace engineering judgement. A standard may specify a test method, insulation requirement, electromagnetic compatibility limit or quality process; the engineer must still determine which standards apply, identify residual risks and document justified decisions. Similarly, ethical conduct extends beyond avoiding plagiarism. It includes honest reporting, responsible data management, disclosure of conflicts of interest and accountability for foreseeable consequences.

Learning Objectives

By the end of this chapter, students should be able to:

  1. Identify electrical, electromagnetic, environmental, software and human hazards in a project.
  2. Apply the hierarchy of controls and basic risk-assessment methods.
  3. Explain the purpose and limitations of IEC, IEEE, national and sector-specific standards.
  4. Design an EMC assessment strategy involving emissions, immunity, grounding, shielding and filtering.
  5. Apply principles of research integrity, data governance, cybersecurity and privacy.
  6. Recognize responsible-AI issues, conflicts of interest and professional accountability.
  7. Produce a defensible safety, compliance and governance record for a major project.

1. Safety as a Systems Property

1.1 Hazard, risk and harm

A hazard is a source or condition with the potential to cause harm. Risk combines the likelihood of an unwanted event with the severity of its consequence. A useful qualitative model is:

R = L × S

where R is a risk rating, L is likelihood and S is severity. This expression is a prioritization aid, not a precise probability law. Likelihood and severity must be defined consistently, and the rating should be supported by evidence rather than arbitrary numbers. Some organizations use a risk matrix instead of multiplication, particularly where catastrophic events must receive high priority even when they are unlikely.

Electrical hazards include shock, burns, arc flash, overcurrent, unexpected energization, stored energy, rotating machinery, battery failure, electromagnetic exposure and fire. Project-specific hazards may arise from high-voltage supplies, capacitors, inductors, lithium batteries, lasers, radio-frequency transmitters, soldering, chemicals, cooling systems and embedded actuators.

1.2 Energy and shock protection

Electric shock severity depends on current through the body, its path, duration, waveform, frequency and individual conditions. A simplified relation is Ohm’s law:

I = V/R

where I is current in amperes, V is voltage in volts and R is resistance in ohms. The relation is only an approximation for a human body because body impedance varies with skin condition, contact area, frequency, current path and applied voltage. It must never be used to infer that a particular voltage is automatically safe.

Protection should be layered. Typical measures include insulation, barriers, enclosures, protective earthing, bonding, fuses, circuit breakers, residual-current devices, isolation transformers, reduced-energy design, interlocks, emergency stops and safe discharge paths. Stored energy must be considered after the supply is disconnected. A capacitor stores approximately:

E = ½CV²

where E is energy in joules, C is capacitance in farads and V is voltage in volts. This equation assumes an ideal capacitor charged to voltage V. For example, a 100 μF capacitor charged to 400 V stores E = 0.5 × 100×10⁻⁶ × 400² = 8 J. Eight joules can produce a dangerous arc or mechanical reaction; the capacitor must therefore be discharged, verified and guarded.

1.3 Hierarchy of controls

The hierarchy of controls prioritizes controls from most to least reliable:

  1. Elimination: remove the hazardous energy or operation.
  2. Substitution: use a lower-voltage, lower-energy or less hazardous alternative.
  3. Engineering controls: enclosures, interlocks, current limiting, isolation and automatic shutdown.
  4. Administrative controls: procedures, permits, training, labels and supervision.
  5. Personal protective equipment: insulated gloves, face shields, arc-rated clothing and eye protection.

PPE is important but should not be the primary control when an engineering solution is practicable. A competent person should define test boundaries, authorized operators, isolation procedures and emergency arrangements. Before energization, inspect wiring, clearances, protective conductors, current limits, fuses, mechanical stability and measurement equipment. Use the one-hand practice only where appropriate and never treat it as a substitute for de-energization and isolation.

1.4 Risk assessment and safe experimental practice

A project risk assessment should identify the activity, hazard, possible consequence, existing controls, initial risk, additional controls, responsible person and residual risk. Controls should be verified rather than merely listed. A residual risk is acceptable only when it is justified against institutional policy, legal requirements and the project’s context.

A practical laboratory sequence is:

  • define the operating envelope;
  • identify normal, abnormal and maintenance conditions;
  • isolate all energy sources before modification;
  • verify absence of voltage using an appropriate tested instrument;
  • control stored energy;
  • establish barriers and warning signs;
  • conduct a pre-start inspection;
  • energize progressively with current limiting;
  • monitor temperature, current, vibration and unexpected behavior;
  • record incidents, near misses and corrective actions.

Safety documentation should be version controlled. A change from a 12 V prototype to a 400 V inverter is not a minor revision; it requires a new assessment, updated procedures and possibly specialist approval.

2. Standards and Conformity

2.1 The role of standards

A standard is a consensus-based technical document or rule that establishes requirements, recommendations, terminology, test methods or performance criteria. IEC standards are widely used in electrotechnology; IEEE standards address many electrical, electronic, computing and power-engineering topics; national standards and regulations provide jurisdiction-specific legal requirements. Other important families include ISO management standards, CISPR EMC standards, occupational safety rules and sector requirements for medical, automotive, aerospace, energy or telecommunications equipment.

Standards may be:

  • regulatory or mandatory, when adopted into law, a contract or an approval scheme;
  • normative, containing requirements that must be met within the document’s scope;
  • informative, providing explanation or guidance;
  • voluntary, but commercially or professionally expected.

The current edition, amendments, national deviations and scope must be checked. A standard’s title alone is insufficient evidence that it applies. Engineers should maintain a compliance matrix identifying each requirement, evidence, test method, responsible person and status.

Compliance-matrix fieldTypical content
RequirementExact clause or legal obligation
ApplicabilityWhy it applies or does not apply
Design evidenceSchematic, calculation, component certificate or software record
VerificationInspection, analysis, test or demonstration
ResultPass, fail, conditional or open
Owner and dateResponsible person and evidence date

2.2 Standards do not replace judgement

Compliance with a standard does not guarantee absolute safety. A standard may not address an unusual use environment, foreseeable misuse or a newly introduced technology. Conversely, a project may use a standard as a design reference without claiming formal certification. Students must distinguish between “designed with reference to” and “certified compliant with.” Only an authorized conformity process can support a formal certification claim where law or contract requires it.

A sound process begins with requirements capture: destination markets, supply voltage, environment, user population, intended function, foreseeable misuse, interfaces, wireless operation and maintenance conditions. The project team then identifies applicable laws, standards and institutional procedures, performs a gap analysis, designs controls and retains objective evidence.

3. Electromagnetic Compatibility

3.1 Meaning of EMC

Electromagnetic compatibility means that equipment can function satisfactorily in its electromagnetic environment without introducing intolerable electromagnetic disturbance to other equipment. EMC has two complementary dimensions:

  • emissions: unwanted conducted or radiated energy produced by equipment;
  • immunity: ability to operate correctly when exposed to electromagnetic disturbance.

A design can fail either dimension. For example, a switching converter may disturb a radio receiver, while a control system may reset during an electrostatic-discharge event.

3.2 Coupling paths

Interference reaches a victim through conduction, radiation, capacitive coupling or inductive coupling. Typical sources include fast digital edges, power converters, motors, relays, radio transmitters and ESD events. Effective EMC design begins at the source and considers the complete source–path–victim model.

Common controls include minimizing loop area, controlling edge rates, separating noisy and sensitive circuits, using continuous reference planes, filtering cables, applying appropriate shielding, bonding enclosures and selecting common-mode or differential-mode filters correctly. Cable shields require a deliberate termination strategy; an ineffective or pigtail connection can increase inductance and reduce high-frequency performance.

3.3 Decibels and measurement

Power ratios are commonly expressed as:

dB = 10 log₁₀(P₂/P₁)

For equal impedances, voltage or current ratios may be expressed as:

dB = 20 log₁₀(V₂/V₁) = 20 log₁₀(I₂/I₁)

The 20-log form assumes the compared impedances are equal or that the relevant power relationship justifies it. P₁ and P₂ are powers; V₁, V₂ are voltages; and I₁, I₂ are currents. A 20 dB reduction in equal-impedance voltage corresponds to a voltage ratio of 0.1, not 0.01.

EMC testing must specify detector type, bandwidth, antenna or probe, cable arrangement, grounding, test environment and equipment configuration. A laboratory pre-compliance test can identify problems early, but it is not necessarily a substitute for an accredited or legally recognized conformity test.

4. Research Integrity and Evidence

Research integrity requires honesty, accuracy, transparency, care and respect. Misconduct commonly includes fabrication, falsification and plagiarism. Questionable practices include selective deletion of inconvenient measurements, undisclosed image manipulation, p-hacking, inappropriate authorship, duplicate publication and presenting simulation results as experimental results.

A defensible engineering report distinguishes:

  • measured data from calculated values;
  • raw data from processed data;
  • assumptions from observations;
  • correlation from causation;
  • uncertainty from error;
  • exploratory findings from confirmatory claims.

Measurements should include instrument identity, calibration status, range, resolution, sampling conditions, environmental conditions and repeated trials where meaningful. A simple mean is:

x̄ = (1/n)∑ᵢ₌₁ⁿ xᵢ

where xᵢ is the ith observation and n is the number of observations. The mean does not by itself establish accuracy; systematic error, uncertainty and outliers require separate consideration. Data should not be removed merely because it weakens a preferred conclusion. Any exclusion rule must be defined and justified.

Authorship should reflect substantial intellectual or practical contribution, while supervision, funding or routine assistance should not automatically establish authorship. Every author should approve the final work and accept responsibility for the contribution claimed.

5. Data Governance, Privacy and Cybersecurity

5.1 Data governance

Data governance assigns responsibilities for collection, access, quality, retention, sharing, security and disposal. A project data-management plan should answer: What data are collected? Why are they necessary? Who owns or controls them? Where are they stored? How are they backed up? Who can access them? When are they deleted or archived?

The data life cycle includes collection, classification, storage, processing, sharing, retention and destruction. Access should follow least privilege: each person receives only the access needed for their role. Version control, checksums, audit logs and documented metadata improve reproducibility.

5.2 Privacy

Personal data may include names, identifiers, location, images, health information, voice recordings, device identifiers and inferred attributes. Privacy principles generally include purpose limitation, data minimization, transparency, accuracy, security, retention limitation and respect for individual rights. Applicable privacy law depends on jurisdiction and context; students must consult institutional ethics procedures and legal guidance rather than assuming that public availability means unrestricted use.

Anonymization seeks to prevent re-identification, while pseudonymization replaces direct identifiers but may remain reversible. Small datasets, distinctive behavior and linked databases can make re-identification possible. Consent must be informed, voluntary and appropriately documented; it is not valid if participation is coerced or the purpose is unclear.

5.3 Cybersecurity of electrical systems

Connected devices and control systems require security throughout their life cycle. Threats include unauthorized access, malware, insecure firmware updates, weak credentials, exposed debug ports, denial of service, manipulated sensor data and unsafe commands. Core controls include secure authentication, role-based access, encrypted communications where appropriate, signed updates, network segmentation, secure defaults, logging, vulnerability management, backup and recovery.

Security must be balanced with safety and availability. A control system that locks down without a safe fallback may create a physical hazard. Threat modeling should identify assets, actors, attack surfaces, consequences and mitigations. Students should never scan, exploit or test systems without explicit authorization and a defined scope.

6. Responsible AI and Automated Decision-Making

AI may be used to classify faults, optimize control, generate code, analyze signals or assist technical writing. Its use introduces risks of biased data, hallucinated technical claims, leakage of confidential material, non-reproducible outputs, insecure generated code and unclear accountability.

Responsible use requires:

  • defining the system’s intended purpose and limits;
  • verifying training and test data quality and representativeness;
  • testing performance across relevant operating conditions;
  • monitoring false positives and false negatives;
  • preserving human oversight for high-consequence decisions;
  • documenting model version, data, prompts, parameters and evaluation results;
  • protecting confidential and personal information;
  • disclosing material AI assistance.

An AI-generated circuit, citation, calculation or safety recommendation is an unverified suggestion, not authoritative evidence. The engineer remains responsible for checking it against primary sources, simulation, measurement and applicable standards. For safety-critical decisions, explainability, traceability and a controlled fallback are especially important.

7. Conflicts of Interest and Professional Accountability

A conflict of interest occurs when personal, financial, institutional or relational interests could improperly influence professional judgement, or reasonably appear to do so. Examples include evaluating a product supplied by a family member, accepting gifts from a vendor, hiding a failed test to protect funding, or selecting a component because of an undisclosed commercial relationship.

A conflict is not necessarily misconduct. The proper response is disclosure, assessment and management. Possible controls include independent review, recusal, separation of duties, transparent procurement and recording the decision rationale.

Professional accountability means accepting responsibility for decisions, limitations and consequences. Engineers should communicate risks in language that decision-makers can understand, escalate unsafe conditions, preserve evidence and correct errors promptly. A student should not sign or imply approval for work they did not perform or verify. Whistleblowing should follow institutional procedures where possible, while imminent threats to life or serious harm require urgent escalation to competent authorities.

8. Integrated Project Governance

A major project benefits from a governance plan that connects technical design with responsibility. At minimum, the plan should contain:

  1. scope, intended use and stakeholders;
  2. hazard register and risk controls;
  3. applicable standards and compliance matrix;
  4. test and calibration plan;
  5. data-management and cybersecurity plan;
  6. ethics approval or exemption record where required;
  7. change-control procedure;
  8. incident, near-miss and nonconformance process;
  9. approval points and named responsibilities;
  10. final limitations, residual risks and maintenance requirements.

A stage-gate approach may include concept approval, preliminary design review, safety review, prototype review, verification testing and release review. No gate should be treated as ceremonial. Evidence should be sufficient for the decision being made.

Applied Case: Illustrative Battery-Powered Environmental Monitor

Background

A student team develops a battery-powered monitor containing a microcontroller, wireless radio, temperature and air-quality sensors, and a small fan. The device is intended for deployment in occupied university rooms.

Problem

Initial tests show intermittent resets when the fan starts. The team also stores identifiable room-location data in a cloud service and proposes using an AI model to classify “poor air quality.” One member suggests omitting failed trials from the report because the final accuracy would otherwise decline.

Analysis

The fan creates a transient load and conducted disturbance. The wireless system introduces EMC and cybersecurity considerations. Room-level data may reveal occupancy patterns, and the classification model may produce misleading results if the training data do not represent different rooms and ventilation conditions. Omitting failed trials would be falsification if the trials are relevant to the stated evaluation.

Approach and solution

The team creates a hazard and governance register. It measures battery impedance, fan starting current and supply-voltage dips, then separates the fan supply using an appropriate filter and improves local decoupling. It verifies that the change does not create unacceptable conducted or radiated emissions. A watchdog and brownout recovery strategy are added, but the device also enters a defined safe state when sensor data are invalid.

The team minimizes data by storing a non-identifying device code rather than room occupants’ identities, obtains institutional approval where required, restricts cloud access and defines retention. The AI model is evaluated against a held-out dataset from several ventilation conditions. Confusion matrices, false-positive and false-negative behavior are reported, and the model is described as decision support rather than an autonomous health determination.

All trials, including failures, are retained with timestamps and reasons. The report distinguishes exploratory debugging from final verification. A supervisor independently reviews the safety assessment and the team discloses a supplier relationship involving one selected sensor.

Outcome

The monitor becomes more reliable, the data collection is proportionate, the AI claim is narrowed, and the final report is reproducible. The project does not claim formal product certification; instead, it identifies the standards and tests that would be required before deployment beyond the university.

Learning points

Technical performance, safety, privacy, cybersecurity and integrity are interconnected. A design change can alter EMC; a data choice can create ethical risk; and an honest negative result can be more valuable than an inflated performance claim.

Challenges and Limitations

Standards may be expensive, difficult to interpret or unavailable in a student laboratory. National adoption can differ from the international source, and compliance requirements may change during a project. Risk matrices can create false precision, while inexperienced teams may underestimate low-frequency catastrophic events. EMC behavior is highly dependent on physical layout and test configuration, so simulation alone is inadequate. Privacy protection is difficult when datasets are small or linkable. AI performance can deteriorate after deployment because operating conditions change. Finally, organizational pressure can discourage escalation of inconvenient findings.

These limitations reinforce the need for competent supervision, specialist consultation, conservative assumptions, independent review and clear statements of uncertainty. The correct professional response to an unresolved safety or compliance issue is to stop, document and escalate—not to conceal or rationalize it.

Contemporary Developments and Best Practices

Current engineering practice increasingly uses digital traceability, automated test records, secure software supply chains, lifecycle cybersecurity, model cards, environmental impact assessment and safety cases. Power electronics, electric vehicles, renewable-energy systems, industrial control and medical devices increasingly combine high energy, connectivity and software, making multidisciplinary review essential.

Best practice is to:

  • begin safety and compliance planning during requirements definition;
  • consult the current standard and its national adoption;
  • design out hazards before relying on procedures or PPE;
  • test normal, abnormal and foreseeable misuse conditions;
  • preserve raw data and document processing;
  • use independent review for high-consequence claims;
  • disclose conflicts, limitations and AI assistance;
  • apply least privilege and secure defaults;
  • maintain change control and configuration management;
  • communicate residual risk clearly to users and maintainers.

Chapter Summary

Electrical engineering projects must be safe, compatible with their electromagnetic environment, aligned with applicable standards and governed by ethical professional practice. Risk assessment begins with identifying hazards and applying the hierarchy of controls, with engineering controls preferred over administrative controls and PPE. Stored energy, unexpected energization and abnormal conditions require particular attention.

Standards provide requirements, test methods and shared language, but engineers must determine scope, current edition, legal status and evidence of conformity. EMC requires control of emissions and immunity through source, path and victim analysis, supported by appropriate measurement. Research integrity demands accurate reporting, traceable data and honest treatment of uncertainty.

Data governance protects quality, security and accountability across the data life cycle. Privacy requires purpose limitation, minimization, transparency and appropriate security. Cybersecurity must be integrated with physical safety and system availability. Responsible AI requires validation, disclosure, human oversight and recognition that generated outputs are not authoritative. Conflicts of interest should be disclosed and managed, while professional accountability requires escalation of serious risks and correction of errors. A governance plan, compliance matrix, hazard register and independent review transform these principles into practical project controls.

Key Terms

Hazard: A source or condition with potential to cause harm.

Risk: The combination of likelihood and severity of an unwanted consequence.

Residual risk: Risk remaining after controls have been implemented.

Hierarchy of controls: Ordered approach prioritizing elimination, substitution, engineering controls, administrative controls and PPE.

Protective earthing: Connection intended to provide a low-impedance fault-current path and reduce dangerous touch voltage.

Stored energy: Energy retained in capacitors, inductors, batteries, mechanical systems or other components after disconnection.

Standard: A technical document establishing requirements, guidance, terminology or test methods.

Conformity assessment: Process used to determine whether specified requirements have been fulfilled.

Electromagnetic compatibility: Ability to function acceptably in an electromagnetic environment without unacceptable disturbance.

Emissions: Electromagnetic energy unintentionally produced by equipment.

Immunity: Ability of equipment to operate correctly when exposed to electromagnetic disturbance.

Research integrity: Honest, accurate, transparent and responsible conduct of research.

Fabrication: Inventing data or results that were not obtained.

Falsification: Manipulating materials, processes or data so that the research record is misleading.

Data governance: Assignment of responsibility for data quality, access, security, retention and use.

Least privilege: Giving users or systems only the access necessary for their authorized function.

Privacy: Protection of people’s information, identity and reasonable control over its use.

Responsible AI: Development and use of AI with attention to validity, fairness, transparency, security, human oversight and accountability.

Conflict of interest: An actual, potential or perceived interest that could improperly influence judgement.

Professional accountability: Responsibility for technical decisions, communications, limitations and consequences.

Review Questions

Short Answer Questions

  1. Distinguish a hazard from a risk and give one electrical example of each.
  2. Why is PPE considered a lower-level control than elimination or engineering control?
  3. Explain the difference between electromagnetic emissions and immunity.
  4. What is the purpose of a compliance matrix?
  5. State two ways in which research data can be made more reproducible.

Long Answer Questions

  1. Explain how a student team should conduct a risk assessment for a high-voltage converter prototype.
  2. Discuss the relationship between IEC, IEEE, national standards and legal requirements in an engineering project.
  3. Describe an EMC design and verification strategy for a microcontroller-based motor controller.
  4. Explain how data governance, privacy and cybersecurity should be integrated into a connected sensing project.
  5. Discuss the ethical responsibilities of an engineer using AI-generated code or analysis in a safety-relevant application.

Analytical / Application Questions

  1. A capacitor of 220 μF is charged to 300 V. Calculate its stored energy and recommend controls before servicing it. State the assumptions used.
  2. A prototype passes a conducted-emissions pre-compliance test but resets during electrostatic discharge. Analyze possible coupling paths and propose a prioritized investigation plan.
  3. A project sponsor asks a team to remove failed trials from a performance report and offers future employment to the student who approves the report. Identify the integrity and conflict-of-interest issues and propose an accountable response.

Multiple Choice Questions with Answers

Question 1

Which control is highest in the hierarchy of controls?

A. Personal protective equipment

B. Administrative instruction

C. Hazard elimination

D. Warning label

Correct Answer: C. Hazard elimination

Question 2

A 100 μF capacitor is charged to 400 V. Approximately how much energy does it store?

A. 0.008 J

B. 0.8 J

C. 8 J

D. 80 J

Correct Answer: C. 8 J

Question 3

Electromagnetic immunity refers to the ability of equipment to:

A. Avoid producing any electromagnetic field

B. Operate acceptably when exposed to electromagnetic disturbance

C. Increase its transmitted radio power

D. Eliminate all conducted current

Correct Answer: B. Operate acceptably when exposed to electromagnetic disturbance

Question 4

Which statement best describes a compliance matrix?

A. A list of project costs only

B. A record linking requirements to evidence and verification status

C. A replacement for laboratory testing

D. A list of preferred component suppliers

Correct Answer: B. A record linking requirements to evidence and verification status

Question 5

Removing relevant failed trials to improve a reported result is most directly an example of:

A. Falsification

B. Anonymization

C. Calibration

D. Substitution

Correct Answer: A. Falsification

Question 6

Which principle gives a user only the access required for an authorized task?

A. Open access

B. Least privilege

C. Data duplication

D. Full transparency of credentials

Correct Answer: B. Least privilege

Question 7

For equal impedances, a 20 dB reduction in voltage corresponds approximately to a voltage ratio of:

A. 0.001

B. 0.01

C. 0.1

D. 10

Correct Answer: C. 0.1

Question 8

Which is the most appropriate response to an undisclosed financial relationship with a component supplier?

A. Hide it if the component performs well

B. Disclose and manage the conflict

C. Delete the supplier from the report

D. Allow the supplier to approve the test results

Correct Answer: B. Disclose and manage the conflict

Question 9

Why should AI-generated technical output be independently verified?

A. AI output is automatically legally certified

B. AI output may contain incorrect, biased or insecure recommendations

C. Verification prevents all future software updates

D. AI cannot process numerical information

Correct Answer: B. AI output may contain incorrect, biased or insecure recommendations

Question 10

Which statement about standards is correct?

A. Any standard automatically proves legal certification

B. Standards never require interpretation

C. The applicable scope, edition and legal status must be checked

D. A standard eliminates the need for risk assessment

Correct Answer: C. The applicable scope, edition and legal status must be checked

Major Project / Dissertation
Chapter 11 of 14