Digital Engineering, Analytics and Artificial Intelligence

Introduction and Context

Hydraulic structures—dams, spillways, barrages, weirs, canals, tunnels, pumping stations and levees—are increasingly designed, constructed and managed as information-rich systems. Their safety and performance depend not only on physical dimensions and materials but also on the quality, timeliness and interpretation of data. Digital engineering integrates computer-aided design (CAD), building information modelling (BIM), geographic information systems (GIS), remote sensing, computational fluid dynamics (CFD), finite-element analysis (FEA), digital twins and Internet of Things (IoT) monitoring into a coordinated engineering workflow.

Artificial intelligence (AI) extends this workflow by identifying patterns in large datasets, detecting anomalies, forecasting hydraulic and structural behaviour, and supporting predictive maintenance. However, digital tools do not replace engineering judgement. A model may be computationally sophisticated but physically inappropriate, poorly calibrated or vulnerable to corrupted data. Consequently, postgraduate engineers must understand both the capabilities and limitations of digital methods.

This chapter treats digitalisation as an engineering decision system rather than as a collection of software products. The central principle is that every digital result must remain traceable to physical reality, governing equations, field observations and accountable professional decisions.

Learning Objectives

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

1. Explain the roles and relationships of CAD, BIM, GIS, remote sensing, CFD and FEA in hydraulic engineering.

2. Design an information workflow from survey and investigation through operation and maintenance.

3. Distinguish numerical simulation, statistical analytics, machine learning and digital twins.

4. Apply suitable data and performance metrics to anomaly detection, forecasting and predictive maintenance.

5. Evaluate model verification, validation, calibration and uncertainty.

6. Identify cybersecurity, privacy, interoperability and data-governance requirements.

7. Critically assess AI applications in hydraulic-structure safety and asset management.

1. Digital Engineering as a Systems Approach

Digital engineering is the structured creation, management and use of digital information throughout an asset’s life cycle. It includes geometric information, material properties, design assumptions, construction records, monitoring observations, inspection findings, maintenance actions and operational constraints.

A useful distinction is:

• Data are recorded observations, such as reservoir level or piezometric pressure.
• Information is interpreted data, such as an identified rise in uplift pressure.
• Knowledge is understanding of mechanisms, such as seepage caused by drain blockage.
• Decision support connects knowledge to action, such as inspection, drawdown or remedial grouting.

A digital workflow should preserve a chain of custody:

measurement → quality control → storage → modelling → interpretation → decision → recorded action.

Breaking this chain can create dangerous confidence in inaccurate outputs. For example, a highly detailed three-dimensional model cannot compensate for an incorrect foundation permeability or an incorrectly surveyed spillway crest.

1.1 The asset information model

A hydraulic-structure information model normally contains:

• spatial reference systems and survey control;
• geometry and topology of hydraulic and structural components;
• materials, construction stages and design parameters;
• hydrological and meteorological time series;
• boundary conditions and operational rules;
• instrumentation locations, calibration records and uncertainty;
• inspection images, defects and remedial works; and
• roles, approvals, revisions and access permissions.

The model should have a defined level of information need. Excessive detail increases cost and can obscure critical variables, whereas insufficient detail prevents reliable analysis. The required detail for a construction quantity take-off differs from that needed for seepage diagnosis or emergency response.

2. CAD, BIM and GIS in Hydraulic Structures

2.1 Computer-aided design

CAD supports two-dimensional drafting and three-dimensional geometric representation. It is valuable for alignment drawings, reinforcement details, hydraulic profiles, construction sequencing and fabrication information. CAD layers, coordinate systems, naming conventions and revision control must be standardised. A drawing should identify its datum, units, scale, issue status and responsible checker.

CAD is primarily a representation tool. It does not automatically ensure that a design satisfies hydraulic, structural, geotechnical or safety requirements. Geometrically correct drawings may still contain disconnected drainage paths, incorrect invert levels or unbuildable reinforcement arrangements.

2.2 Building information modelling

BIM links geometry with non-geometric information and relationships. In a dam or barrage project, objects may include gates, embedded parts, drains, galleries, turbines, pumps, concrete blocks and monitoring instruments. Each object can contain manufacturer data, inspection history, replacement intervals and links to calculations.

BIM improves coordination through clash detection and controlled information exchange. It can reveal, for example, that an access passage conflicts with a gate hoist, that a drainage pipe intersects reinforcement, or that an instrument cannot be installed after concrete placement.

BIM should be treated as a managed information environment rather than merely a three-dimensional model. Important management practices include:

• a common data environment;
• agreed object classification and naming;
• permissions and approval workflows;
• model federation rather than uncontrolled file duplication;
• clear responsibility for each information field; and
• a documented exchange format and revision history.

2.3 GIS and spatial analysis

GIS manages geographically referenced data. It is particularly useful for catchment delineation, flood-risk mapping, sediment sources, canal command areas, landslide susceptibility, downstream population exposure and emergency planning.

Raster data represent continuous fields such as elevation or rainfall; vector data represent points, lines and polygons such as instruments, canals and inundation boundaries. Spatial analysis must account for coordinate reference systems, horizontal and vertical datums, resolution and positional uncertainty.

GIS can integrate hydraulic-model results with land use and population data. A flood-inundation polygon, however, is meaningful only when its terrain model, boundary conditions, roughness assumptions and vertical datum are known. Overlaying datasets with inconsistent datums can produce apparently precise but physically misplaced results.

2.4 Integration of CAD, BIM and GIS

CAD and BIM generally describe assets at engineering scale, whereas GIS describes their regional and landscape context. Integration requires consistent coordinates, unique asset identifiers and common metadata. A gate in BIM should correspond to the same gate in the maintenance database and the same mapped feature in GIS.

A practical asset identifier might encode facility, subsystem and component, but identifiers should remain stable even when geometry or ownership changes. Stable identification is essential for joining inspection, sensor and maintenance records.

3. Remote Sensing and Survey Data

Remote sensing supplies observations without direct physical contact. Common sources include satellite imagery, aerial photography, radar, LiDAR, multispectral sensors, thermal cameras and unmanned aerial vehicles.

Applications include:

• reservoir-area and water-level estimation;
• flood-extent mapping;
• shoreline and bank erosion;
• sediment plume tracking;
• vegetation and land-use change;
• deformation mapping using interferometric synthetic aperture radar (InSAR);
• inspection of inaccessible faces and spillway surfaces; and
• digital elevation model generation.

Each product has limitations. Optical imagery is affected by cloud, haze and illumination. Radar may be affected by layover, vegetation and decorrelation. LiDAR can provide high-resolution elevation but may not penetrate water reliably and requires careful classification. UAV surveys require flight planning, control points, permissions and safe operation.

Remote-sensing observations should be validated against ground control, gauge readings or conventional surveys. A change detected in imagery may represent genuine deformation, a change in water level, vegetation effects, processing artefacts or a datum error. Remote sensing is therefore best viewed as complementary evidence, not an unquestionable replacement for inspection.

4. Computational Fluid Dynamics

CFD numerically solves governing equations for fluid motion over a discretised domain. For incompressible Newtonian flow, the continuity equation is

∇·u = 0,

and the momentum equation can be written as

ρ(∂u/∂t + u·∇u) = −∇p + μ∇²u + ρg + f,

where ρ is fluid density, u is velocity, t is time, p is pressure, μ is dynamic viscosity, g is gravitational acceleration and f represents other body forces. These equations assume a continuum fluid and require appropriate initial and boundary conditions.

In hydraulic engineering, CFD may be used to study spillway aeration, energy dissipation, gate flow, hydraulic jumps, sediment transport, intake vortices and local scour. Turbulence is commonly represented through Reynolds-averaged models, large-eddy simulation or direct numerical simulation. The latter is generally too computationally expensive for practical full-scale structures, while simplified turbulence models require calibration and judgement.

CFD quality depends on:

• domain extent and boundary conditions;
• mesh resolution and quality;
• free-surface treatment;
• turbulence and multiphase models;
• time-step selection;
• convergence criteria; and
• comparison with physical or field data.

A converged solution is not necessarily a correct solution. Mesh-independence testing examines whether results change materially as the mesh is refined. Verification asks whether the equations were solved correctly; validation asks whether the equations and model represent the real system adequately.

5. Finite-Element Analysis of Hydraulic Structures

FEA divides a continuum into elements connected at nodes. The unknown field, such as displacement, is approximated using shape functions. In linear elasticity, the assembled equilibrium equation is commonly expressed as

K u = F,

where K is the global stiffness matrix, u is the vector of unknown nodal displacements and F is the applied load vector. The relationship assumes small deformation, appropriate material behaviour and correctly specified supports and loads.

FEA is used for stress and deformation in concrete dams, arch structures, retaining walls, tunnel linings, foundations and embedded components. Coupled analyses may include seepage, thermal effects, dynamic loading and soil–structure interaction.

For hydraulic structures, common load cases include self-weight, hydrostatic pressure, uplift, sediment pressure, temperature, earthquake effects and construction stages. The engineer must distinguish characteristic, factored and accidental actions according to the governing code and design basis.

Important checks include:

• mesh convergence;
• realistic boundary conditions;
• material constitutive models;
• contact and joint behaviour;
• tensile cracking assumptions;
• uplift and seepage representation;
• sensitivity to uncertain parameters; and
• comparison with hand calculations or independent models.

Numerical stress concentrations at sharp corners may be mathematical artefacts. Decisions should therefore be based on physically meaningful regions, fracture or cracking criteria and engineering interpretation rather than on a single extreme nodal value.

6. IoT Monitoring and Digital Twins

IoT monitoring connects sensors, communications, data storage and analytical services. Typical measurements include reservoir level, gate position, vibration, temperature, joint opening, uplift pressure, pore pressure, seepage discharge, rainfall and downstream water level.

A reliable monitoring system requires sensor selection based on the failure modes of concern. Measurement frequency should reflect the process dynamics. A slowly changing foundation pressure may not require the same sampling rate as gate vibration. Sampling too slowly can miss events; sampling too rapidly can produce unnecessary data and communication burden.

Sensor records should include timestamp, unit, location, calibration status, quality flag and uncertainty. Automated quality checks may identify missing values, impossible ranges, flat-line signals, abrupt discontinuities and communication failures.

A digital twin is a continuously updated digital representation of a physical asset that combines a model, observations and operational context. It is more than a static BIM model and more than a sensor dashboard. A useful twin should support questions such as:

• Is observed seepage consistent with the current reservoir level?
• Has the structural response changed from its established baseline?
• What is the likely consequence of a gate-control failure?
• Which inspection or maintenance action should be prioritised?

Digital twins may be descriptive, diagnostic, predictive or prescriptive. The last category requires particularly strong validation because it can influence operational decisions directly.

7. Analytics and Artificial Intelligence

7.1 Descriptive, diagnostic and predictive analytics

Descriptive analytics summarises what has happened. Diagnostic analytics investigates why it happened. Predictive analytics estimates what may happen, and prescriptive analytics evaluates possible actions.

Hydraulic-structure data commonly contain seasonality, operational changes, missing records and correlated measurements. A simple baseline model can sometimes be more reliable than a complex AI model. For example, uplift pressure may be represented initially as a function of reservoir level, temperature and time trend. More complex models should be justified by demonstrable improvement and not by novelty alone.

7.2 Anomaly detection

An anomaly is an observation or pattern inconsistent with expected behaviour. Detection methods include:

• threshold and rule-based alarms;
• residual analysis from physics-based models;
• statistical control charts;
• clustering;
• principal component methods; and
• supervised or unsupervised machine learning.

For a measured variable y and predicted value ŷ, a residual may be defined as

r = y − ŷ.

A residual is meaningful only when model uncertainty, sensor error and operating conditions are considered. A fixed threshold can create false alarms during normal seasonal changes, while an excessively broad threshold can conceal deterioration.

Anomaly detection should distinguish sensor faults from asset faults. Redundant sensors, cross-variable checks and inspection records help make this distinction. An alarm should lead to a defined response pathway: verification, escalation, inspection, operational adjustment or emergency action.

7.3 Forecasting

Forecasting may concern inflow, reservoir level, flood peaks, seepage, deformation or equipment failure. Models range from time-series methods to neural networks and hybrid physics-informed systems. Forecast accuracy should be assessed on data not used for training.

Useful measures include mean absolute error (MAE),

MAE = (1/n) Σ|yᵢ − ŷᵢ|,

where n is the number of observations, yᵢ is the observed value and ŷᵢ is the forecast. MAE has the same units as the response and is easily interpreted. Other measures may be appropriate when large errors have disproportionate consequences. Accuracy metrics must be accompanied by uncertainty intervals and performance under rare or extreme conditions.

7.4 Predictive maintenance

Predictive maintenance estimates the condition or remaining useful life of equipment and recommends intervention before failure. For gates and pumps, relevant variables may include motor current, vibration, opening time, hydraulic pressure, temperature and maintenance history.

A maintenance decision should consider probability of failure, consequence of failure, inspection cost, outage cost and safety requirements. A statistically accurate model is not automatically a good maintenance policy. The decision threshold must reflect risk and the availability of safe fallback actions.

7.5 Physics-informed and hybrid models

Purely data-driven models may fail outside the range of training data, especially during unprecedented floods or unusual operating conditions. Physics-informed and hybrid models combine conservation laws, numerical simulation and data-driven correction. Such models can improve physical plausibility and reduce data requirements, but they introduce additional assumptions and may be difficult to audit.

8. Verification, Validation, Calibration and Uncertainty

These terms should not be treated as interchangeable:

• Verification asks whether implementation correctly solves the intended mathematical model.
• Validation asks whether the model adequately represents the real system for its intended use.
• Calibration estimates parameters using observations.
• Uncertainty analysis evaluates the effect of imperfect knowledge, measurement error and model limitations.

Validation data must be independent where possible. Using the same observations for calibration and validation can create an unjustified impression of accuracy. For AI systems, data should be separated into training, validation and test sets, with time-based separation often preferable for monitoring applications.

Uncertainty may arise from hydrological inputs, material properties, sensor bias, boundary conditions, geometry, model structure and future climate or operational conditions. Results should therefore be reported with assumptions, ranges and confidence or credible intervals where appropriate.

Engineers should use uncertainty-informed thresholds rather than treating a point prediction as fact. A warning is more defensible when it states both the estimated condition and the confidence in that estimate.

9. Cybersecurity, Data Governance and Ethics

Connected infrastructure creates a cyber-physical risk. Threats include unauthorised access, ransomware, sensor spoofing, denial of service, manipulation of gate commands and theft of sensitive infrastructure information.

Basic controls include:

• network segmentation between operational technology and office systems;
• strong authentication and least-privilege access;
• encryption in transit and at rest;
• secure backups and tested recovery procedures;
• patch and vulnerability management;
• monitoring of logs and unusual commands;
• secure configuration of remote access; and
• manual fallback modes for critical operations.

Data governance defines ownership, stewardship, retention, quality, access, classification and permitted use. Metadata and audit trails are essential for regulatory investigations and safety decisions. Personal or commercially sensitive data must be protected, particularly where cameras, worker records or downstream population information are involved.

AI governance should address explainability, bias, accountability, model drift and human oversight. A black-box prediction should not be allowed to override a statutory safety procedure without an accountable engineer and an approved operating protocol. Automated systems should fail safely and provide operators with intelligible reasons for alarms.

10. Illustrative Case: Digital Monitoring of a Concrete Gravity Dam

Background

Consider an illustrative, fictional concrete gravity dam with a gated spillway, foundation drains and a downstream population centre. The dam has vibrating-wire piezometers, uplift-pressure cells, joint meters, seepage weirs, reservoir-level sensors and gate-position encoders. Records exist for twelve years, but several instruments have intermittent gaps.

Problem

Operators observe a gradual increase in seepage during high reservoir levels. One piezometer also shows an apparent upward trend. It is unclear whether the change represents foundation deterioration, drain blockage, sensor drift or a normal seasonal response.

Analysis

The engineering team first establishes data quality. Sensor timestamps are aligned, units are checked, calibration certificates are reviewed and impossible values are removed without deleting the original records. A physics-based baseline relates uplift and seepage to reservoir level, temperature and recent rainfall. Residuals are then examined by season and operating state.

The apparent piezometer trend is not reproduced by adjacent instruments. Its signal contains periods of constant readings and abrupt steps, suggesting sensor or communication problems. In contrast, seepage at one weir increases beyond the model’s uncertainty band and is accompanied by a change in the pressure difference between two drainage galleries.

Approach and solution

The team does not rely on the AI alarm alone. It performs a manual inspection, checks drain flows, surveys relevant joints and compares the result with a finite-element seepage model. The model indicates that partial obstruction of a drainage path could produce the observed pressure distribution. A camera inspection identifies sediment accumulation near a drain inlet.

The drain is cleaned under an approved procedure, and monitoring continues at increased frequency. The digital twin records the event, intervention, post-maintenance response and updated asset condition. The anomaly-detection model is retrained only after the engineering team confirms that the post-maintenance behaviour represents the desired baseline.

Outcome

Seepage reduces toward the established relationship with reservoir level, while the unreliable piezometer is replaced. The principal benefit is not merely early detection; it is the integration of sensor quality control, physical modelling, field inspection and controlled maintenance.

Learning points

1. A sensor anomaly is not necessarily a structural anomaly.

2. Independent instruments and physical models improve diagnosis.

3. Maintenance records are valuable training data.

4. Human review remains essential for safety-critical decisions.

5. A digital twin must represent uncertainty and intervention history.

11. Implementation Strategy and Best Practices

A successful programme should begin with decisions and failure modes, not software procurement. Recommended stages are:

1. Define safety, operational and maintenance questions.

2. Develop an asset register and information requirements.

3. Establish survey, coordinate, naming and metadata standards.

4. Inventory existing data and assess quality.

5. Select sensors and models linked to credible failure mechanisms.

6. Pilot one high-value use case, such as gate condition monitoring.

7. Establish verification, validation and acceptance criteria.

8. Integrate outputs into operating and emergency procedures.

9. Train staff in interpretation, cybersecurity and data stewardship.

10. Review performance, model drift and benefits periodically.

Interoperability is promoted through documented application programming interfaces, open or well-specified exchange formats and persistent identifiers. Vendor lock-in should be assessed in terms of data portability, long-term support and access to audit information.

Contemporary developments include cloud-edge architectures, satellite-based near-real-time flood mapping, autonomous inspection platforms, foundation models for engineering documents and physics-informed machine learning. These developments offer efficiency but also increase the need for governance, validation and skilled human supervision.

Challenges and Limitations

Digital engineering faces several persistent limitations:

• legacy drawings and undocumented modifications;
• incomplete, biased or non-representative monitoring data;
• sensor drift and communication failures;
• uncertainty in extreme hydrological events;
• incompatible software and proprietary formats;
• high initial cost and specialist skill requirements;
• model drift as assets age or operating rules change;
• false alarms and alarm fatigue;
• cyberattack and data tampering; and
• overconfidence in visually impressive models.

The most serious limitation is often organisational rather than computational. If operators do not trust the system, if maintenance records are not entered, or if alarms have no response procedure, additional analytics will not improve safety. Digital transformation must therefore include governance, competence, change management and funding for sustained operation.

Chapter Summary

Digital engineering connects geometry, physical models, observations and decisions across the life cycle of hydraulic structures. CAD supports controlled representation; BIM adds object-based information and coordination; GIS provides spatial context; remote sensing extends observation over large or inaccessible areas; CFD models fluid behaviour; and FEA models structural response. IoT systems provide continuous monitoring, while digital twins integrate models, data and operational context.

AI can support anomaly detection, forecasting and predictive maintenance, but its value depends on representative data, suitable failure-mode understanding, independent validation and clear response procedures. Verification, validation, calibration and uncertainty analysis must be distinguished. Physics-based and hybrid models are often preferable where extrapolation and safety are important.

Cybersecurity, data governance, interoperability, explainability and human accountability are essential because digital systems influence real hydraulic operations. The appropriate objective is not maximum automation but trustworthy, auditable and risk-informed engineering decision support.

Key Terms

CAD: Computer-aided design used to create and manage engineering drawings and geometry.

BIM: Building information modelling linking asset geometry with structured information and relationships.

GIS: Geographic information system for storing, analysing and displaying spatially referenced data.

Remote sensing: Observation of Earth or assets using satellites, aircraft, UAVs or other non-contact sensors.

CFD: Computational fluid dynamics using numerical methods to solve fluid-flow equations.

FEA: Finite-element analysis dividing a continuum into elements to approximate its response.

IoT: Internet of Things comprising connected sensors, communications and data services.

Digital twin: An updated digital representation of a physical asset connected to observations and context.

Anomaly detection: Identification of observations or patterns inconsistent with expected behaviour.

Forecasting: Estimation of future values or states using data and models.

Predictive maintenance: Maintenance planned using predicted condition or failure risk.

Verification: Checking that a model or software implementation solves its intended formulation correctly.

Validation: Assessing whether a model represents the real system adequately for its intended purpose.

Calibration: Estimation or adjustment of model parameters using observations.

Model drift: Decline in model performance as data, assets or operating conditions change.

Data governance: Rules for data ownership, quality, access, security, retention and accountability.

Interoperability: Ability of systems to exchange and use information consistently.

Residual: Difference between an observed value and a model prediction.

Physics-informed model: A model incorporating governing physical principles into data-driven analysis.

Review Questions

Short Answer Questions

1. Distinguish CAD from BIM in hydraulic-structure practice.

2. Why must remote-sensing results be checked against ground or gauge observations?

3. Define verification, validation and calibration.

4. What is the purpose of a digital twin?

5. Identify four cybersecurity controls for connected hydraulic infrastructure.

Long Answer Questions

1. Explain how CAD, BIM and GIS can be integrated across the life cycle of a dam.

2. Discuss the principal modelling and validation issues in CFD analysis of a spillway.

3. Explain how FEA can support assessment of a concrete hydraulic structure and identify its major assumptions.

4. Critically evaluate AI applications in anomaly detection and predictive maintenance.

5. Develop a governance framework for an IoT-enabled hydraulic-structure monitoring system.

Analytical / Application Questions

1. A seepage sensor rises during high reservoir levels, but neighbouring instruments remain stable. Develop a step-by-step diagnostic process that separates sensor failure from genuine seepage deterioration.

2. Propose a digital-twin architecture for a gated barrage, identifying data sources, models, alarms, cybersecurity controls and human decision points.

3. An AI forecasting model performs well during ordinary floods but poorly during extreme events. Analyse possible causes and recommend a safer hybrid modelling and validation strategy.

Multiple Choice Questions with Answers

Question 1

Which statement best distinguishes a digital twin from a static BIM model?

A. A digital twin contains only two-dimensional drawings

B. A digital twin is continuously connected to observations and operational context

C. A digital twin cannot include physical models

D. A digital twin is used only during construction

Correct Answer: B. A digital twin is continuously connected to observations and operational context

Question 2

In the residual expression r = y − ŷ, what do y and ŷ represent?

A. Density and viscosity

B. Mesh size and time step

C. Observed and predicted values

D. Pressure and discharge

Correct Answer: C. Observed and predicted values

Question 3

Which activity is primarily associated with model verification?

A. Comparing predictions with field behaviour

B. Checking whether the numerical implementation solves the intended equations

C. Selecting a maintenance budget

D. Interviewing operators

Correct Answer: B. Checking whether the numerical implementation solves the intended equations

Question 4

Which technology is most directly suited to regional flood-extent mapping?

A. Remote sensing integrated with GIS

B. Reinforcement detailing in CAD

C. Finite-element stress analysis only

D. Gate motor lubrication records

Correct Answer: A. Remote sensing integrated with GIS

Question 5

What is a major risk of using a fixed anomaly threshold without considering operating conditions?

A. It eliminates all false alarms

B. It may create alarms during normal seasonal variation

C. It automatically validates the sensor

D. It prevents model drift

Correct Answer: B. It may create alarms during normal seasonal variation

Question 6

For the incompressible-flow continuity equation ∇·u = 0, the principal assumption is that the fluid is

A. Compressible with changing density

B. Incompressible in the model domain

C. Always inviscid

D. Always turbulent

Correct Answer: B. Incompressible in the model domain

Question 7

Which practice best supports cybersecurity in operational technology?

A. Using one shared administrator password

B. Removing all access logs

C. Network segmentation and least-privilege access

D. Connecting every device directly to the public internet

Correct Answer: C. Network segmentation and least-privilege access

Question 8

Why should AI test data generally be independent of training data?

A. To increase the number of model parameters

B. To provide a more credible estimate of performance on unseen data

C. To remove the need for calibration

D. To guarantee zero prediction error

Correct Answer: B. To provide a more credible estimate of performance on unseen data

Question 9

A CFD solution is converged but disagrees with laboratory measurements. The most appropriate conclusion is that

A. convergence proves the model is valid

B. the disagreement requires investigation of model assumptions and validation

C. the measurements must automatically be wrong

D. mesh refinement is never relevant

Correct Answer: B. the disagreement requires investigation of model assumptions and validation

Question 10

Which response best distinguishes a sensor anomaly from an asset anomaly?

A. Ignore all isolated alarms

B. Use redundant measurements, quality checks and physical inspection

C. Increase the alarm threshold permanently

D. Replace the hydraulic model with a spreadsheet

Correct Answer: B. Use redundant measurements, quality checks and physical inspection