Construction

Quality Assurance in Materials and Construction

Appendix D

Use of a Statistical Evaluation of Project Test Results to Monitor the Overall Process

Although many practices were noted during this review involving State DOTs, the Team is highlighting one practice noted in the Alabama Division. This practice involved the Alabama Division Office personnel proactively implementing the intent of TA 6120.3 to investigate a concern with their QA program.

Most process reviews focus on verifying compliance with specifications, regulations, and good practice. Another type of review is to take the data acquired on projects and analyze it to monitor the overall process.

The Alabama Division Office has performed three reviews by analyzing asphalt test results. The Alabama Division Office requested copies of all asphalt test results on numerous projects throughout the State. They then entered the data into the computer and analyzed it using established statistical methods. For these reviews all of the test results for each asphalt mix were evaluated together.

One component of the review was comparing the overall quality of the material accepted to the pay factors established by the contract specifications. For each asphalt mix, the Division Office performed a quality level analysis to determine the percent of material that was within certain evaluation ranges. This was done to determine if the existing contract methods for determining pay factors for material quality were reasonable. For laboratory air voids the Division Office calculated the percent of material between 3 percent and 5 percent air voids. They found that 17 of 25 mixes evaluated had more than 25 percent of the material outside of a 3 percent to 5 percent acceptance range. They also found that the contract method of evaluating the quality of the material had awarded a bonus to 16 of these mixes.

Another component of the review was comparing the contractor's test results with the Agency's test results. This comparison was on a mix basis, not a lot by lot basis. According to the contract, the results were compared on the project using a sample by sample comparison. By statistically comparing all the results for each mix a recurring bias that is undetectable on a sample by sample comparison becomes apparent. In this analysis, 18 of the 23 mixes evaluated failed either the t-Test or the F-test at 99 percent confidence.

For the initial review, the Division Office used QL-PAY, a computer program used by Federal Lands to establish pay factors based on percent-within-limits (PWL) and statistically compare two sets of test results using the t-test and F-test. This program can be obtained by contacting any of the Federal Lands Offices. For the follow-up review, they used an Excel spreadsheet that utilized the t-test and F-test functions built into Excel. The spreadsheet only performs the t-test and F-test; it does not perform a quality analysis. The spreadsheet has since been enhanced by the Resource Center to be more user friendly. A copy of the spreadsheet can be obtained from the Resource Center by contacting Rob Elliott or Tom Harmon.

This type review is a very effective tool the FHWA Division Office can use to monitor the effectiveness of the State's QC/QA procedures. The data can be analyzed at various levels; mix, project, district, contractor or statewide. This is a good way to verify if there are problems that the State's materials quality control and quality assurance system is not catching. It is also a way to assess the quality of the material being accepted.

For further information on this successful practice, contact Mr. Steve Mills, Operations and Pavements Engineer, Alabama Division Office.