Contractor Resource Center

The following document summarizes the WVDOH quality assurance and quality control philosophy.

Note: This is a lengthy report. You may wish to save the report to your hardrive and read it at a later time.

This paper was originally presented to the 1996 Road Builders Clinic Coeur d’Alene, Idaho March 11-13, 1996.

A Systems Approach to Insuring Quality
Gary L. Robson, West Virginia Division of Highways
We have heard many times about the need to do more with less. Because of budget restraints, personnel cuts, or whatever, there is just not enough money or people available to do everything we need to do in our industry. While the reasons may be different, the problem is the same as it was some 30 years ago; The need to do more with less.

In the early 1960’s we recognized that the traditional approach to construction inspection and to prescription specifications would not be manageable with the projected construction schedules involving the building of our Interstate System. We decided to consider the methods that had been successfully developed by the Defense Industry and the Federal Government during the second World War.

Because of the massive industrial expansion at the beginning of our involvement in World War II, American industry and the various military Departments faced essentially this same dilemma, that is, how do we inspect and test such massive production using traditional acceptance techniques and still not seriously impede necessary production.

Industry and government combined their effort to develop the concept of shared responsibility.

Industry developed the ability, sometimes with prodding, to control their production process to provide a high probability of compliance with requirements and still produce on a massive scale.

The buyers could then provide on the spot checks of this process to insure themselves that the risk of an inferior product was minimized.
We decided these concepts could be easily adapted to the production of construction materials and to certain elements in the highway construction process itself.

These concepts were defined and identified in various ways , but ultimately evolved into a system designed to assure the quality of a unit of material or work.

This system consisted of a quality control function, which was the domain of the producer or manufacturer or contractor, and a responsibility to make the decision to accept or reject the work, which was the right and responsibility of the buyer.

This system of assuring quality would require that suitable materials be used in the work, that the appropriate equipment be used to perform the work, that standard procedures be used to control the equipment and materials, and that methods be devised to adequately measure the quality of the work and materials used.

It soon became apparent to us in West Virginia that the foremost requirement would be that both parties to this system be staffed by qualified technicians. Trained, qualified technicians must carry out the process control or quality control functions and equally qualified people must carry out the acceptance function.

These designed experiments produced thousands of test samples and sufficient data to not only measure the variability of each of the materials and processes involved but also allowed us to measure the components of variance so that we could concentrate our training efforts toward reducing the major contributors to the overall variance.

Once these parameters were defined for all of the materials we selected to study, we could begin to write the specifications in terminology that recognized variability and still provided the assurance of quality that we wanted.

An application of some of the knowledge gained could be very simply worded as follows. We determined that compressive strength test results from “good” concrete produced for bridge abutments (a design strength of 3000 pounds per square inch is specified for this type of work) averaged 4200 psi and produced a standard deviation of 650 psi.

This information together with a knowledge of statistics led to the conclusion that a specification for strength, therefore, should recognize that more than 97% of all concrete used in bridge abutments would have strength values greater than 3000 psi, but almost 3% of the concrete could have strength values less than 3000 psi. In this case the bridge abutment would still be judged as “good”. The bottom line was that we should not expect 100% of all concrete produced to have strength values greater than 3000 psi, unless we wanted to demand production of concrete that had an average strength value in excess of 5000 psi.

This same knowledge could be used to write process control requirements that were dependent upon the variability of the process itself. If a contractor’s process control produced concrete strength values with a standard deviation of 1000 psi, then the average strength value needed to be 5000 psi. If the process produced strength values with a standard deviation of 200 psi, then the average strength value needed only to be 3400 psi.

The specification developed for portland cement concrete is a good example of the statistically based specifications developed as a result of this research.

Based on our new knowledge of process variability and the criticality of structural members, we defined a poor lot of concrete as one that the strength averaged only 4 standard deviations above the defined stress. The designer defines the design stress of a concrete. A good lot is one that has a strength requirement that minimizes both the sellers and buyers risks.

These definitions of good and poor Lots of concrete were used to assess the risks to both the state, the buyer, and the contractor who is the seller.

Compressive strength requirements for portland cement concrete pavement were arrived at using these same statistical concepts.
Statistical comparison techniques were developed to allow us to compare acceptance test data with quality control test data to insure that both systems were accurately measuring the product characteristics. We still use this comparison technique today to compare our acceptance samples to contractor QC tests.

The specification for portland cement concrete for structures, which is based on the knowledge gained in these studies plus an equitable use of probability concepts, includes the following:

• The contractor must have trained, certified technicians,
• The contractor must prepare a Quality Control Plan,
• The contractor must conduct tests and inspections to document that his process is in control.
• The contractor must develop, test and submit his own mix design,
• The contractors process control variability dictates the degree of over design required in the mix design,
• The contractor must provide QC test results to DOH,
• The contractors quality control compressive strength test results, developed as a part of his quality control process, can be used by DOH as a part of the acceptance decision documentation,
• The contractors production process variability dictates the level of inspection of the DOH,
• The contractors production process variability dictates the cement factor of the approved mix design,
• The DOH acceptance techniques which allow for a reduction of traditional DOH test frequencies by up to 90%,
• The DOH acceptance techniques require a statistical comparison of contractor and DOH test results to insure compatibility,
• The DOH acceptance techniques require that the compressive strength test results used to substantiate acceptance be analyzed statistically. The specifications establish that 93% of all of the concrete involved in the analysis be expected to have a compressive strength greater than the design strength specified and that 99.87% be greater than the design stress plus one sigma, and
• The acceptance specification provides for reduced payment for non-specification lots of material.

The specification for Portland Cement Concrete (PCC) pavement includes these same Quality Control or Process Control requirements, however, acceptance is based upon cores used for measurement of strength and thickness and on ride quality. The strength and thickness measurements are analyzed statistically and probability concepts are applied.

I would like to digress for a few minutes to tell you about the systems that go along with such a QA based specification. As I mentioned earlier, our Quality Assurance system requires both the state and the contractor to make measurements of the process and the products produced from that process. This means that a lot of data are produced. This data are used to feed the systems that make our (quality Assurance system work. For example, we designed our systems to require that the actual standard deviation of the production process be used in assessing the acceptability of the contractors mix design. Our QA system also includes a plant rating system.

The data that drives this system comes from concrete batch tickets, project site test results, a contractor/producer laboratory, the States district laboratory and our Materials Testing Laboratory. Database parameters for the PCC database include identifying the class of concrete, the cement factor, the target air content and the measure air content, the target slump and the measured slump, and identifiers to make the connection between the batch information and compressive strength test documents.

The data entry procedures have evolved from the use of forty-column IBM port a punch cards, to mark sensing forms, to the entry of data through an on-line system at the district level.

Once the data are submitted it is processed and collated to provide data integrity.

If the coefficient of variation of data developed on concrete slump, air content, strength and aggregate/cement solids content are within the highest prescribed limits and the physical plant has met inspection requirements we classify the plant A1. If the process control is not quite as tight but still acceptable, we classify the plant A2. If the process is not in control or the physical plant is not acceptable we classify the plant B. We do not allow competitive production from a B plant, we require full time DOH inspection in an A2 plant and we only require inspection in an A1 plant 10 times a month.

The evaluation of this data is conducted monthly using the main frame and software written internally. A summation of data for the last two years is separated into three time frames. Statistical parameters of average, standard deviation, coefficient of variation and range are produced for the last two years, the last year and the last three months. A subsystem that is used to evaluate the source’s ability to maintain a satisfactory level of compressive strength is a part of this system. Each class of concrete produced is evaluated using the cement factor and compressive strength as variables. If the average strength deviates from a specified strength level for the class of concrete and cement factor, an adjustment in the cement content is required.

When a mix design is submitted, either at the start of a project or when some component of an existing mix is changed, the strength level and other parameters of the proposed mix are compared to the statistics generated from production. The variation of the production process dictates the amount of overdesign. We require that the laboratory strengths be 2 standard deviations higher than the design strength requirement.

There are numerous support programs that aid in the collation and analysis of the data from the various sources while others complement the plant evaluations. A few of the support programs are for the following: analysis of wheel load carrying capacity of rigid pavements, thickness evaluation, certified cement mill analysis, all test data input by item and project and operation of an independent assurance sampling program.

All of these programs extract data from the database of field and laboratory testing.

Now back to examples of other specifications we have developed. Our first statistically based asphalt concrete specifications were developed from data obtained during our initial studies of variability of gradation and asphalt content and of in place density tests using a nuclear density gauge.

The same concepts of shared responsibility were included. The contractor must have trained personnel, must submit a quality control plan (and live with his own process control commitment), must develop his own mix proportions and conduct quality control tests. Since our studies indicated that an acceptable process exhibited considerable variation in test data we designed our sampling programs for both process control and acceptance to recognize this variation. We determined that any decision of control or acceptance of asphalt content or gradation should be based on the average of four tests. Since it was an extreme burden to develop this amount of data daily we decided to base these decisions on the moving average of multiple, consecutive sublots.

Our current specifications for Asphalt Concrete are an end result specification. Process control is also very important in this updated specification because the producer is required to reproduce his mix design every day in production and we accept on in place density, thickness and ride quality.

There is always an interest in what we perceive to be the advantages and disadvantages of the system we have been using for the past 25 years. Obviously, since we have continued to subscribe to these concepts, we believe that the advantages far exceed the disadvantage, if there is one.

There are advantages to the contractor. Selection of materials proportions for mixtures allows some savings. Process control allows the contractor to minimize variation and therefore reduce probability of rejection or reduced payment. The cement concrete specifications allow for the reduction of the target cement factor for any given mix if the strength level and variability of strength is controlled; this can be a direct savings to a quality conscious producer. All of these specifications allow, even require, the contractor to control his work schedule. There is no waiting for the DOH inspector to show up or to finish his tests.
The advantages to the DOH include the fact that we can do more with less. Rather than concentrate significant jobsite manpower to sampling and testing activities, we can concentrate on inspection. Rather than assign two or more inspectors to a cement concrete plant or an asphalt concrete plant or to an aggregate production facility, we can concentrate our inspection so that those that have trouble controlling their process can receive the most of our attention. Those that control their process and meet our high standards require less of our efforts.
There are fewer disagreements over the quality of test data.

Since the contractor/producer conducts tests at a frequency the same as DOH acceptance frequency requirements, we can use his data to assist in our acceptance decisions. We have reduced our testing on jobsites to almost 10 percent of the testing required prior to implementing these quality assurance specifications.

It is estimated that we have accepted more than 20,000,000 tons of asphalt concrete using these concepts of shared responsibility. 4,000,000 plus cubic yards of portland cement concrete have been accepted using the process control system as a part of the evidence of compliance and some 150,000,000 cubic yards of embankment material have been accepted without 100 percent DOH testing and inspection.

We know that the quality of the materials used in construction of a bridge pier, as an example, is higher. The last time we had to assess a penalty for nonspecification strengths was so long ago that our people have to look for the records; memory fails them.

I trust that most of you will remember the 1990 European Asphalt Study Tour in which some of our federal, state and contractor personnel participated. That tour immediately produced a rather concise report on what the participants saw on their European trip. When AASHTO released this report coincident with the Michigan demonstration project on Stone Mastic Asphalt there was an accompanying news release. This news release reported that the trip participants believed that Europe’s major pavements are better than those in the US; a fact that some of us do not subscribe to.

One of the reasons cited for this perceived superiority is that “The contractor has much more control over the work ”.

The report discusses what we have learned to call shared responsibility. We believe we have better construction because we have allowed, even required, the contractor to decide how to produce his product with minimal DOH intervention as long as the product can be demonstrated to meet our standards.

If these experts believe, as we do, that allowing the contractor more control over his work will result in better roadways, what are we waiting for?