Experimental Design, Testing and Quality Analytics

Introduction and Context

Concrete performance is established not by a single test result but by a defensible chain of decisions: defining the engineering question, selecting representative materials, controlling preparation and curing, measuring relevant properties, quantifying uncertainty, and comparing evidence with acceptance criteria. Advanced concrete technology therefore requires more than familiarity with compressive-strength testing. It requires an integrated approach to laboratory experimentation, field investigation, statistical quality control, durability assessment and technical decision-making.

Concrete is intrinsically variable. Variability arises from cement and aggregate characteristics, moisture condition, batching, mixing, transport, placement, compaction, curing, specimen preparation, testing equipment and operator practice. A high-quality investigation must distinguish genuine material or process effects from random measurement variation and systematic bias. It must also recognize that a test result is an estimate of a property, not the property itself.

This chapter presents a framework for planning and interpreting investigations of concrete strength and durability. It is intended for postgraduate students and practicing engineers who must design experiments, evaluate quality data, specify acceptance procedures and make decisions under uncertainty. Applicable standards should always be consulted; examples in this chapter use general statistical principles and do not replace project-specific standards such as those issued by ASTM, EN, BS, IS or other national bodies.

Learning Objectives

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

1. formulate testable questions and select suitable experimental designs;

2. develop sampling plans for laboratory and field concrete;

3. distinguish destructive, non-destructive and partially destructive tests;

4. calculate and interpret mean, standard deviation, coefficient of variation and confidence intervals;

5. apply control charts and acceptance criteria appropriately;

6. identify sources of measurement uncertainty and bias;

7. interpret strength and durability data without overstating conclusions; and

8. prepare a technically defensible quality-investigation report.

1. From Engineering Question to Experimental Design

1.1 Define the decision before choosing the test

A sound investigation begins with a decision statement. Examples include:

• Is a proposed mixture suitable for a specified characteristic strength?
• Has a change in supplementary cementitious material altered early-age strength?
• Does concrete placed in a marine exposure zone meet durability requirements?
• Is a low rebound number evidence of inadequate in-place strength?

The decision determines the response variable, specimen type, age, exposure condition, sample size and acceptance rule. Testing without a defined decision frequently produces large quantities of data that do not answer the engineering question.

A testable research question should identify:

• the population or process being studied;
• the factor or intervention of interest;
• the response variable;
• the comparison or control condition;
• the relevant age and exposure; and
• the decision threshold.

For example: “At 28 days, does replacing 30% of Portland cement with a specified fly ash reduce mean compressive strength below the project requirement, when water-to-binder ratio, aggregate grading and curing are controlled?”

1.2 Variables and experimental factors

The response variable is the measured outcome, such as compressive strength, splitting tensile strength, chloride migration coefficient, sorptivity or mass loss. Factors are controllable or observable influences, such as water-to-binder ratio, binder type, curing regime, temperature, air content and age. Levels are the selected values of a factor.

Factors may be:

• quantitative, such as water-to-binder ratio or curing temperature;
• categorical, such as binder type or curing method;
• fixed, when specific levels are deliberately selected; or
• random, when levels represent a broader population, such as batches or production days.

A nuisance variable affects the response but is not the principal subject of study. Blocking is used to control nuisance variation. For example, tests performed on different days may be blocked by day so that day-to-day laboratory variation is not confused with the effect of mixture composition.

1.3 Replication, randomization and blocking

Three principles are fundamental.

Replication means independently preparing or sampling multiple experimental units. Repeated readings on the same specimen are not equivalent to independent replication because they do not capture specimen-to-specimen variation.

Randomization assigns treatments or test order randomly, reducing systematic association between treatment and uncontrolled influences. If all control specimens are tested in the morning and all modified specimens in the afternoon, time-of-day effects may be confounded with treatment effects.

Blocking groups relatively homogeneous units, such as batches, casting days or aggregate sources. Comparisons are then made within blocks whenever possible.

A technical replicate is a repeated measurement of one specimen or sample. A biological or production replicate is an independently produced concrete unit. The latter is generally more valuable for estimating process variability.

1.4 Common experimental designs

A completely randomized design is suitable when experimental units are reasonably homogeneous and one or more treatments are assigned at random. A randomized block design is preferable when a known nuisance factor, such as batch or day, is important.

A factorial design investigates two or more factors simultaneously. For two factors, A and B, the design can estimate:

• the main effect of A;
• the main effect of B; and
• the interaction between A and B.

An interaction exists when the effect of one factor depends on the level of another. For example, the influence of curing temperature may differ for ordinary Portland cement and slag cement. Studying each factor separately could miss this interaction.

A response-surface design is useful when the objective is optimization over quantitative factors, such as water-to-binder ratio, paste volume and chemical-admixture dosage. It requires careful selection of factor ranges and should not be used to extrapolate beyond the investigated region.

Mixture experiments require special treatment because mixture proportions are constrained by a total, usually 100%. Increasing one component necessarily reduces at least one other component. Conventional independent-factor analysis can therefore be misleading; mixture-design methods are more appropriate.

1.5 Avoiding confounding

Confounding occurs when the effect of one factor cannot be separated from another. If a low-water mixture also uses a different aggregate and is cured at a different temperature, an observed strength change cannot be attributed uniquely to water-to-binder ratio. The remedy is to hold irrelevant factors constant, randomize their order, block their effects or include them explicitly in the design.

2. Sampling and Specimen Management

2.1 Population, lot and sample

The population is the complete set of material or concrete units to which conclusions are intended to apply. A lot is a defined quantity produced under sufficiently similar conditions. A sample is the portion selected for testing. A specimen is the prepared test unit, such as a cylinder, cube, prism or drilled core.

Representativeness is more important than convenience. Samples should be selected using a documented procedure that covers relevant production times, locations and batches. Taking all samples from the first truck or from easily accessible surface concrete can introduce serious bias.

2.2 Field sampling principles

Field samples should be obtained from concrete representative of the placement under investigation. The sampling plan should specify:

• batch or delivery identification;
• sampling location and time;
• number and size of increments;
• method of combining increments;
• tests performed on fresh concrete;
• specimens cast and their intended ages; and
• transport, protection and curing conditions.

Fresh concrete properties such as temperature, slump or slump flow, air content, density and unit water addition should be recorded with batch information. A strength specimen without traceable batch data has limited diagnostic value.

For hardened concrete, cores should be located using drawings, reinforcement scans and the investigation objective. Core orientation, diameter, length-to-diameter ratio, drilling direction, moisture condition, defects and trimming details must be recorded. Cores containing reinforcement, major voids or visible damage may be unsuitable for particular tests, but exclusion must be documented rather than used selectively to improve results.

2.3 Sample size and power

Sample size should reflect the expected variability, minimum practically important difference and desired probability of detecting that difference. A simple planning expression for comparing two independent means is:

\[

n \approx \frac{2(z_{1-\alpha/2}+z_{1-\beta})^2\sigma^2}{\Delta^2}

\]

where \(n\) is the approximate number of independent observations per group, \(\alpha\) is the two-sided significance level, \(1-\beta\) is the desired power, \(\sigma\) is the assumed common standard deviation, and \(\Delta\) is the difference to be detected. The expression assumes independent observations, approximately normal errors, equal variance and a comparison of two means. It is a planning approximation, not an acceptance formula.

Concrete programs often require more units than a purely statistical calculation suggests because specimens may be damaged, rejected for valid reasons or needed for multiple ages and durability tests. The unit of replication must be defined before testing. Three cylinders cast from one batch are not necessarily equivalent to three independently sampled batches.

2.4 Chain of custody and data integrity

Each sample should have a unique identifier linked to mixture, batch, location, date, age and curing history. Records should include equipment identification, calibration status, operator, deviations, environmental conditions and raw readings. Data should not be transcribed selectively or altered without an audit trail. Photographs of failure modes and specimen condition are often valuable evidence.

3. Testing Concrete: Principles and Limitations

3.1 Destructive tests

The compressive-strength test is a controlled failure test. For a specimen with maximum applied load \(P\) and loaded cross-sectional area \(A\), nominal compressive strength is:

\[

f_c=\frac{P}{A}

\]

where \(f_c\) is in pascals when \(P\) is in newtons and \(A\) is in square metres. The equation assumes that the specified specimen geometry, loading rate, alignment, end preparation and moisture condition are controlled in accordance with the governing standard. It does not by itself correct for specimen shape, size, slenderness or capping; such corrections, where permitted, must follow the relevant standard.

Flexural strength, splitting tensile strength and fracture-energy tests provide different information from compressive strength. No single strength test fully characterizes structural performance. A high compressive strength may coexist with low tensile resistance or poor shrinkage behavior.

Destructive durability tests may measure mass loss, expansion, permeability-related transport, electrical resistivity, chloride migration, carbonation depth or resistance to freeze-thaw cycles. Accelerated tests can be useful for comparison, but acceleration may alter the governing mechanism. For example, an imposed electrical field in a migration test is not identical to natural chloride diffusion.

3.2 Non-destructive and partially destructive tests

Rebound hammer, ultrasonic pulse velocity, impact-echo, penetration resistance, pull-off and maturity methods can support investigation without removing large quantities of concrete. Their principal uses are screening, uniformity mapping, locating anomalies and selecting areas for confirmatory cores.

Non-destructive tests usually require project-specific correlation with destructive results. A universal conversion from rebound number or pulse velocity to compressive strength is unreliable because correlations depend on aggregate type, moisture, age, surface condition, carbonation and mixture composition. A correlation established on laboratory cubes should not automatically be applied to old, in-place structural concrete.

The maturity method estimates strength development from the combined effect of time and temperature. A strength–maturity relationship is valid only for concrete with sufficiently similar mixture constituents and curing history to the calibration specimens. It should not be used as an unconditional substitute for strength testing.

4. Statistical Description of Concrete Data

For observations \(x_1,x_2,\ldots,x_n\), the sample mean is:

\[

\bar{x}=\frac{1}{n}\sum_{i=1}^{n}x_i

\]

The sample standard deviation is:

\[

s=\sqrt{\frac{\sum_{i=1}^{n}(x_i-\bar{x})^2}{n-1}}

\]

The denominator \(n-1\) provides an unbiased estimate of population variance under independent random sampling. The coefficient of variation is:

\[

CV=100\frac{s}{\bar{x}}\%

\]

provided that the mean is positive and meaningfully different from zero. CV is useful for comparing relative variability, but it can be misleading when means are small or when the distribution is strongly non-normal.

The standard error of the mean is:

\[

SE=\frac{s}{\sqrt n}

\]

It describes uncertainty in estimating the population mean, not the scatter of individual future results. For approximately normal data with unknown population standard deviation, an approximate 95% confidence interval for the mean is:

\[

\bar{x}\pm t_{0.975,n-1}\frac{s}{\sqrt n}

\]

where \(t_{0.975,n-1}\) is the 97.5th percentile of the Student’s t distribution with \(n-1\) degrees of freedom. The interval describes uncertainty in the estimated mean under the stated sampling assumptions; it does not mean that 95% of individual concrete results lie inside the interval.

Worked example: strength statistics

Suppose six independent 28-day cylinder results, in MPa, are 42, 45, 44, 47, 41 and 45. Their mean is:

\[

\bar{x}=264/6=44.0\text{ MPa}

\]

The squared deviations from the mean are 4, 1, 0, 9, 9 and 1, giving a sum of 24. Thus:

\[

s=\sqrt{24/5}=2.19\text{ MPa}

\]

and

\[

CV=100(2.19/44.0)=4.98\%

\]

The standard error is \(2.19/\sqrt6=0.89\) MPa. With five degrees of freedom, \(t_{0.975,5}\) is approximately 2.571, so the approximate 95% confidence interval is:

\[

44.0\pm2.571(0.89)=44.0\pm2.29

\]

or approximately 41.7 to 46.3 MPa. This interval concerns the mean represented by the six observations; it is not an acceptance limit for every individual specimen.

Distributional considerations and outliers

Concrete strength data may be approximately normal within a stable process, but small samples make distributional assumptions difficult to verify. Durability measurements such as permeability or crack width may be skewed and may require transformation or robust summaries. An outlier should never be removed merely because it is inconvenient. Investigate specimen damage, transcription error, equipment malfunction, sampling error and genuine process causes. Exclusion requires a pre-established rule or documented technical evidence.

5. Quality Control and Acceptance Criteria

5.1 Process control versus acceptance testing

Process control monitors production to identify drift and prevent defects. Acceptance testing determines whether a defined lot meets contractual or regulatory requirements. These purposes differ. A process-control result may trigger adjustment even when the lot technically passes acceptance; conversely, a lot may pass a limited acceptance rule while the process shows undesirable trends.

Acceptance criteria must specify the characteristic being assessed, sampling frequency, specimen age, test method, allowable variability and treatment of nonconforming results. The criterion may concern a characteristic strength, a minimum individual result, a mean of a defined group or a durability threshold. Engineers must not substitute a convenient criterion for the one specified by the governing code or contract.

5.2 Control charts

A Shewhart chart plots results in time or production order against a central line and control limits. For individual observations, limits are commonly estimated from the process mean and an estimate of within-process variation. Control limits are statistical signals, not specification limits. A process can be statistically stable but produce concrete that fails the specification; it can also be statistically unstable while current results happen to meet the specification.

Signals include a point beyond a control limit, sustained runs on one side of the centre line and systematic trends. Rules should be selected before examining data to avoid overinterpreting random patterns. Control charts require rational subgrouping: observations grouped together should represent similar short-term conditions, while differences between subgroups reveal process changes.

5.3 Capability and process performance

When specification limits and a stable process are available, capability indices may be informative. For a normally distributed process with lower and upper specification limits LSL and USL:

\[

C_p=\frac{USL-LSL}{6\sigma}

\]

and

\[

C_{pk}=\min\left(\frac{USL-\mu}{3\sigma},\frac{\mu-LSL}{3\sigma}\right)

\]

where \(\mu\) and \(\sigma\) are the process mean and standard deviation. \(C_p\) measures potential capability if centred; \(C_{pk}\) also reflects off-centre operation. These indices assume a stable process, meaningful specification limits and an approximately normal distribution, or a justified alternative analysis. They should not be calculated mechanically for a changing process or a small, nonrepresentative dataset.

5.4 Acceptance risk

Acceptance decisions involve producer’s risk—the probability of rejecting acceptable concrete—and consumer’s risk—the probability of accepting unacceptable concrete. A small sample can provide weak protection against both risks. The sampling plan should therefore be designed in relation to the consequences of failure, reliability requirements and applicable code provisions. Safety-critical infrastructure may justify more extensive investigation than a low-consequence application.

6. Uncertainty Analysis

Measurement uncertainty combines contributions from repeatability, reproducibility, calibration, resolution, specimen geometry, environmental conditions and sampling. A result should be reported with sufficient information to indicate how confidently it represents the measurand.

If independent input quantities \(x_i\) contribute to a result \(y=f(x_1,x_2,\ldots,x_m)\), the first-order combined standard uncertainty is approximately:

\[

u_c(y)=\sqrt{\sum_{i=1}^{m}\left(\frac{\partial f}{\partial x_i}u(x_i)\right)^2}

\]

where \(u(x_i)\) is the standard uncertainty of input \(x_i\). This expression assumes sufficiently small uncertainties and independence. If inputs are correlated, covariance terms must be included:

\[

u_c^2(y)=\sum_i c_i^2u^2(x_i)+2\sum_{i<j}c_ic_j\,u(x_i,x_j)

\]

where \(c_i=\partial f/\partial x_i\) and \(u(x_i,x_j)\\) is covariance.

An expanded uncertainty is often reported as \(U=ku_c\), where \(k\) is a coverage factor selected for the desired coverage probability and distributional assumptions. For approximately normal conditions, \(k\) near 2 is commonly associated with roughly 95% coverage, but it is not universally exact.

For compressive strength \(f=P/A\), uncertainty arises from load measurement and area measurement, as well as from specimen preparation and sampling. Statistical variation between specimens is conceptually distinct from instrument uncertainty, although both affect confidence in a decision. A rigorous investigation reports both where relevant.

7. Interpreting Strength and Durability Evidence

7.1 Strength interpretation

Strength depends on age, moisture, curing, specimen geometry, loading rate, end condition and direction of loading. Results should therefore be compared only when test conditions are compatible. A lower core strength than a standard-cured cylinder does not automatically prove deficient concrete; in-place curing, drilling damage, size effects and moisture condition may contribute.

When investigating an apparent deficiency, use a hierarchy of evidence:

1. verify records, calculations and specimen identity;

2. check testing-machine calibration and laboratory procedure;

3. examine fresh-concrete and placement records;

4. review spatial and temporal patterns;

5. perform suitable non-destructive mapping;

6. obtain targeted cores or other confirmatory samples; and

7. undertake structural assessment if material nonconformity is confirmed.

A single low value should prompt investigation, not an automatic conclusion.

7.2 Durability interpretation

Durability is exposure-specific. Water absorption, sorptivity, electrical resistivity, chloride transport, carbonation depth, sulfate expansion and freeze-thaw resistance measure different mechanisms. A low value in one test does not guarantee performance under all exposures.

For carbonation, a commonly used empirical relationship is:

\[

x_c=k\sqrt t

\]

where \(x_c\) is carbonation depth, \(t\) is exposure time and \(k\) is an empirical coefficient for a defined concrete and environment. This relationship is an approximation; it assumes a particular exposure regime and should not be extrapolated indiscriminately to changing humidity, sheltering, cracking or cyclic wetting.

Durability data should be interpreted with exposure history, temperature, humidity, cracking, cover depth and curing quality. Accelerated testing is most defensible when it has been correlated with field performance or is used for comparative screening within a controlled program.

7.3 Correlation is not causation

A correlation between low resistivity and high chloride content, for example, may be affected by pore-solution chemistry, moisture and temperature. Regression can describe association but does not establish mechanism. Models should be validated using independent data and evaluated for prediction error, not judged only by a high coefficient of determination.

8. Illustrative Case Study: Investigation of Variable 28-Day Strength

This is a fictional illustrative case.

Background

A contractor produced 40 MPa ready-mixed concrete for a transfer slab. The specified acceptance procedure required results at 28 days, while the project team observed greater-than-usual variation in three consecutive production weeks.

Problem

Twenty-four cylinder results had a mean of 41.8 MPa and a standard deviation of 4.6 MPa. Two results were below 35 MPa. The contractor proposed deleting both results as “abnormal.” The engineer needed to determine whether the issue was random testing variation, a process shift or genuine nonconformity.

Analysis

The investigation first checked specimen identifiers, curing records, machine calibration, loading rates and failure modes. Both low specimens were traceable and had satisfactory failure patterns. No transcription error was found. Results were then plotted by production date and batch. The lower values clustered on afternoon batches.

Fresh-concrete records showed that afternoon ambient temperature was higher and slump was frequently increased at site by adding water. The water additions were not consistently recorded. A review of aggregate moisture measurements showed that the moisture correction had been missed on two days.

The team separated results by production day and treated day as a potential block. It also inspected the relationship between measured slump, added water and strength. The data suggested a process-related increase in effective water-to-binder ratio rather than a random outlier problem. Because the sample was observational and confounded by several operational changes, the team did not claim that temperature alone caused the reduction.

Approach and solution

The corrective program included:

• verification and recalibration of aggregate moisture sensors;
• a prohibition on undocumented water addition;
• recording of all site adjustments;
• additional fresh-concrete sampling from morning and afternoon batches;
• duplicate specimens from independently selected batches;
• a short factorial laboratory study of water-to-binder ratio and curing temperature; and
• a control chart based on production order rather than report order.

The two low results were retained. A targeted core investigation was not initially required because subsequent process-control data stabilized and structural records indicated that the affected batches had not been placed in critical zones. If later evidence had shown persistent deficiency, cores and structural analysis would have been necessary.

Outcome

Over the next four weeks, the mean increased to 43.6 MPa and the standard deviation decreased to 2.8 MPa. No result was excluded. The improvement was associated with corrected moisture control and controlled water addition, although the team appropriately treated this as process evidence rather than a formal causal experiment.

Learning points

The case demonstrates that acceptance data can also diagnose production processes. Outlier removal without investigation can conceal a real problem. Production order, traceability and fresh-concrete records are essential. Statistical stability and specification compliance are related but different questions. Finally, the proper response to uncertain evidence is targeted additional information, not unsupported certainty.

9. Professional, Regulatory and Ethical Considerations

Testing laboratories should operate under competent personnel, documented methods, calibrated equipment and appropriate quality systems. Personnel must report deviations, preserve raw data and distinguish measured results from interpretation. Selective testing, undisclosed specimen rejection, back-calculation of convenient results and alteration of records are serious ethical breaches because they can affect public safety.

Acceptance criteria should be aligned with the governing contract, design code and national standards. Engineers must identify which edition and method apply, because specimen dimensions, curing requirements and statistical rules can differ among standards. Where standards conflict or are incomplete, the basis for professional judgment should be recorded.

Confidentiality and data governance are also relevant. Digital laboratory systems and artificial-intelligence tools can assist with anomaly detection, image classification of failures, automated transcription and predictive maintenance. However, AI outputs require validation, traceability and human review. A model trained on one aggregate, climate or testing machine may not generalize to another. AI should support—not replace—the engineer’s responsibility for sampling validity, safety decisions and compliance.

10. Best-Practice Workflow

A defensible investigation can follow this sequence:

1. define the engineering decision and population;

2. identify response variables, factors, confounders and acceptance thresholds;

3. select a design with replication, randomization and blocking where appropriate;

4. prepare a sampling and chain-of-custody plan;

5. verify methods, calibration, personnel competence and environmental controls;

6. collect fresh and hardened-concrete data with full traceability;

7. inspect data for transcription errors, missingness, trends and process order;

8. quantify variation and uncertainty, without confusing standard deviation with standard error;

9. compare results with the correct contractual or code criterion;

10. investigate nonconformity using independent evidence;

11. report limitations, assumptions and decisions transparently; and

12. use findings to improve the process, not merely to classify individual results.

Challenges and Limitations

Concrete investigations are limited by small samples, heterogeneous materials, changing field conditions and imperfect reference methods. Laboratory control can improve repeatability while reducing realism. Accelerated durability tests may not reproduce field mechanisms. Non-destructive correlations can be site-specific. Statistical models can create false confidence when data are nonrepresentative or autocorrelated. Acceptance criteria may be unable to distinguish a localized defect from a widespread process problem.

These limitations do not make testing ineffective; they define how evidence should be qualified. Strong conclusions require convergence among independent sources: test results, production records, spatial patterns, physical inspection and engineering analysis.

Chapter Summary

Experimental design in advanced concrete technology begins with a clearly defined engineering decision. Representative sampling, independent replication, randomization, blocking and control of confounding are essential to credible conclusions. Destructive tests provide direct performance evidence but are affected by specimen preparation and test conditions; non-destructive methods are valuable for screening and mapping but require suitable calibration.

The mean describes central tendency, standard deviation describes specimen-to-specimen variation, standard error describes uncertainty in the estimated mean, and confidence intervals quantify estimation uncertainty under stated assumptions. Control limits are not specification limits, and process capability indices are meaningful only for stable and appropriately modelled processes. Measurement uncertainty should include relevant instrument, method, sampling and environmental contributions.

Strength and durability results must be interpreted in relation to age, curing, exposure, geometry and mechanism. A low result should initiate a structured investigation rather than automatic rejection or unjustified deletion. Quality analytics is therefore both statistical and engineering: it integrates numbers with materials science, field records, standards, professional judgment and ethical responsibility.

Key Terms

Population: The complete set of units to which an investigation intends to generalize.

Lot: A defined quantity produced under sufficiently similar conditions.

Sample: A portion selected from a population or lot for investigation.

Specimen: A prepared test unit, such as a cylinder, cube, prism or core.

Replication: Independent repetition of an experimental unit or production sample.

Randomization: Random allocation or ordering used to reduce systematic bias.

Blocking: Grouping similar units to control a known nuisance source of variation.

Factor: An input or condition that may influence a measured response.

Response variable: The measured outcome of an experiment or investigation.

Confounding: Inability to separate the effects of two or more factors.

Standard deviation: A measure of dispersion among individual observations.

Standard error: The estimated uncertainty of a sample mean.

Coefficient of variation: Standard deviation expressed as a percentage of the mean.

Confidence interval: An interval procedure used to estimate a population parameter under stated assumptions.

Control chart: A time-ordered graphical tool for detecting process instability.

Specification limit: A contractual or technical boundary for acceptable performance.

Measurement uncertainty: Quantified doubt associated with a measured result.

Destructive test: A test that damages or fails the specimen.

Non-destructive test: A test intended to assess concrete without significant damage.

Acceptance risk: The possibility of accepting unacceptable concrete or rejecting acceptable concrete.

Review Questions

Short Answer Questions

1. Distinguish replication from repeated measurement of the same specimen.

2. Why is randomization important in concrete experiments?

3. Define standard error and explain how it differs from standard deviation.

4. Why should a non-destructive strength correlation be calibrated for the project?

5. State two sources of measurement uncertainty in compressive-strength testing.

Long Answer Questions

1. Explain how a factorial experimental design can reveal interaction between binder type and curing temperature.

2. Describe a complete field sampling and chain-of-custody procedure for investigating suspected low-strength concrete.

3. Explain the difference between process control and acceptance testing, including the roles of control charts and specification limits.

4. Discuss the assumptions and limitations of using capability indices for concrete production.

5. Present a structured method for interpreting durability data from an accelerated laboratory test.

Analytical / Application Questions

1. A set of eight independent strengths has a mean of 46 MPa and a standard deviation of 3.2 MPa. Calculate the coefficient of variation and standard error, and explain what each result means.

2. A project shows low rebound numbers in one slab but normal values elsewhere. Design a confirmatory investigation using records, mapping, cores and statistical reasoning.

3. A laboratory study changes water-to-binder ratio and curing temperature simultaneously, with no control mixture. Identify the confounding problem and propose a revised experimental design.

Multiple Choice Questions with Answers

Question 1

Which principle primarily reduces systematic association between treatment and uncontrolled influences?

A. Randomization

B. Replication

C. Extrapolation

D. Censoring

Correct Answer: A. Randomization

Question 2

For compressive strength calculated as f_c = P/A, what does A represent?

A. Specimen age

B. Loaded cross-sectional area

C. Applied strain

D. Aggregate volume

Correct Answer: B. Loaded cross-sectional area

Question 3

For the six strengths 42, 45, 44, 47, 41 and 45 MPa, what is the sample mean?

A. 42.0 MPa

B. 43.0 MPa

C. 44.0 MPa

D. 45.0 MPa

Correct Answer: C. 44.0 MPa

Question 4

What does the standard error of a sample mean primarily describe?

A. Scatter of individual future results

B. Uncertainty in estimating the population mean

C. The specification limit

D. Machine loading rate

Correct Answer: B. Uncertainty in estimating the population mean

Question 5

Which statement correctly distinguishes control limits from specification limits?

A. They are always numerically identical

B. Control limits describe process behaviour, whereas specification limits define required acceptability

C. Specification limits are calculated only from standard deviation

D. Control limits apply only to durability tests

Correct Answer: B. Control limits describe process behaviour, whereas specification limits define required acceptability

Question 6

A non-destructive test is most defensibly used for estimating in-place strength when it has been:

A. Applied without regard to moisture

B. Correlated with suitable project-specific destructive results

C. Performed only on laboratory cubes

D. Converted using a universal equation

Correct Answer: B. Correlated with suitable project-specific destructive results

Question 7

In the approximate sample-size formula for comparing two means, increasing the difference Δ to be detected generally causes the required sample size to:

A. Increase proportionally to Δ

B. Remain unchanged

C. Decrease approximately with 1/Δ²

D. Become equal to the standard deviation

Correct Answer: C. Decrease approximately with 1/Δ²

Question 8

Which situation is an example of confounding?

A. Testing specimens in random order

B. Using independent samples from several batches

C. Changing aggregate type and water-to-binder ratio at the same time

D. Recording specimen temperature

Correct Answer: C. Changing aggregate type and water-to-binder ratio at the same time

Question 9

A process capability index C_pk reflects:

A. Only the number of specimens

B. Both process spread and displacement from specification limits

C. Only instrument resolution

D. The concrete’s water absorption mechanism

Correct Answer: B. Both process spread and displacement from specification limits

Question 10

Why should an apparent outlier in concrete-strength data not be deleted automatically?

A. All low results are always correct

B. It may indicate a genuine process or sampling problem

C. Outliers are never possible in concrete testing

D. Deletion always increases uncertainty

Correct Answer: B. It may indicate a genuine process or sampling problem