STP362: Establishing Specification Limits by Statistical Data Analysis

    Carpenter, C. A.
    Assistant Chief, Materials Research Div., and Engineer, Bituminous Research Div., Bureau of Public Roads, Washington, D. C.

    Oglio, E. R.
    Assistant Chief, Materials Research Div., and Engineer, Bituminous Research Div., Bureau of Public Roads, Washington, D. C.

    Pages: 10    Published: Jan 1964


    Abstract

    Specification limits for test characteristics of road materials and for the control of processes such as compaction have usually been established on the basis of judgement reflecting experience in testing and working with materials. The limits have been set with the basic intention that compliance will be required. Generally, there is considerable operating room or built-in tolerance for the variations that occur as the operation of incorporating the materials in the job is carried out. As long as these normal variations in materials and processes do not result in failure to meet the specifications, quality is considered to be under control.

    Suppliers of materials and contractors tend to operate close to the specification limits when it is financially advantageous. In this case many individual test results and sometimes half or more of all test results fall outside the specification limits, because of at least four specific factors: (1) natural, inherent variations in the material, (2) variations due to sampling, (3) variations due to proportioning and processing, and (4) variations attributable to testing procedures and techniques.

    The Bureau of Public Roads has initiated a broad research and development project in cooperation with the State highway departments that participate, to determine the feasibility of establishing specification requirements through statistical analysis of large samples of test data. Early in November, 1962, a suggestion for such a project was sent to all members of the AASHO (American Association of State Highway Officials) Committee on Materials through the committee's secretary. Several States had already undertaken such studies and had accumulated and analyzed a considerable amount of data. The Bureau proposed to analyze any data that might be made available to it, using normal frequency distribution plots and related procedures as suggested in ASTM Manual of Quality Control of Materials (1).

    Many highway departments participated, expressed their interest in participating, or agreed to the need for it although unable to participate because of limited personnel.

    In order to develop illustrative material applying specifically to highway material, the Bureau made a very limited pilot study on one recent, Federal, construction project using the data from the control testing of the bituminous surfacing material. It distributed an informal report of this work to representatives of the State highway departments at the AASHO annual meeting early in December, 1962. The graphical material from this pilot study is presented in this paper to illustrate the potential of statistical attack.


    Paper ID: STP44576S

    Committee/Subcommittee: D04.94

    DOI: 10.1520/STP44576S


    CrossRef ASTM International is a member of CrossRef.