Abstract
Under certain conditions a sewer pipe may completely lose its structural integrity and 'collapse'; resulting in capital, social and environmental costs. A number of integrated methods have been used to identify and understand the factors associated with an increased likelihood of sewer collapse; namely, a series of sewer collapse site visits, the development of an Interaction Matrix (IM) based on expert knowledge, a statistical investigation and experimental work. The statistical investigation involved the integration of 6 previously disparate data sources into a single data set representing 636km of sewer. Analysis of the data using the logistic regression method allowed the formation of univariate and multivariate models to investigate the association of 18 independent variables with sewer condition. In addition, first order interactions of the independent variables were considered. Of the 18 explanatory terms investigated, 14 were found to be statistically significant in univariate analysis and 10 were retained by stepwise multivariate logistic regression. Further to this 7 first order interaction terms were found to be significant. The site visits, IM method and statistical investigation returned similar results in that all methods identified a wide range of parameters as being important in determining the likelihood of an existing rigid sewer pipe collapsing. The methods were not able to highlight a parameter, or small group of parameters, as being of particular importance. A laboratory based soil box test facility for the structural testing of small diameter sewer pipes was designed and constructed. Experimental work was conducted to investigate the performance of sewer pipes under repeated cyclical loading conditions when laid within beddings of varying compaction levels. Pipes within compacted material were found to perform significantly better, over both short and long time periods, than those in uncompacted material. Suggestions were made regarding the potential for UK water companies to develop predictive models for sewer structural condition as part of overall sewer rehabilitation strategies.