AN APPROACH FOR PREDICTING LATENT INFRASTRUCTURE FACILITY DETERIORATION

A pavement deterioration model predicts the performance of a pavement over time as a function of traffic, pavement characteristics and environmental factors. The most important performance characteristics of a pavement are its ability to bear traffic loads and its ability to provide a smooth ride. However, there is no unambiguous approach to directly measure these characteristics. Therefore, pavement performance is considered to be unobservable. The problem of designing pavement deterioration models is the problem of defining the above unobservable characteristics in terms of what is observed, i.e., in terms of measured extents and severities of different damage components. One approach that has been applied is to regress subjective judgements of inspectors evaluating the pavement on a prespecified set of damage measurements. The two most widely used indices of pavement performance in the United States -- the PCI and the PSI -- have been derived in this fashion on the basis of engineering judgement. A method of getting around the reliance on a subjective condition index has been to model the deterioration over time of a small set of observed damage components. The methodology presented in this paper is a statistical technique that both avoids judgemental measures and provides a basis for a choice between different damage measurements and alternative damage models. Measures of latent pavement performance are statistically estimated from observed pavement damage. No constraints are placed on the number or type of measurements required, so the methodology is flexible enough to include different measurement techniques and data collection strategies. The estimation procedure simultaneously fits a deterioration model and a performance index calibration model to data, thereby producing much better fits to data than traditional deterioration models. Another major problem is accounting for the fact that observed damage is the result of two simultaneously occurring processes: a process of deterioration and the maintenance performed by the highway agency on the pavement. In this paper, a deterioration model is first estimated using the PSI as the performance index and a simultaneous equation specification. This model is then compared against a model using a latent performance variable, where the latent variable is constructed from the same damage measurements as the PSI. Subsequently, models involving multiple latent variables are developed.

  • Corporate Authors:

    Massachusetts Institute of Technology

    Department of Civil Engineering, 77 Massachusetts Avenue
    Cambridge, MA  United States  02139
  • Authors:
    • Ben-Akiva, Moshe
    • Ramaswamy, R
  • Publication Date: 1992

Media Info

  • Features: Appendices; Figures; References; Tables;
  • Pagination: n.p.

Subject/Index Terms

Filing Info

  • Accession Number: 00622407
  • Record Type: Publication
  • Files: TRIS
  • Created Date: May 31 1992 12:00AM