A platform for research: civil engineering, architecture and urbanism
A platform for research: civil engineering, architecture and urbanism
A Survival Analysis Model for Sewer Pipe Structural Deterioration
Abstract: The structural state of sewer systems is often quantified using condition classes. The classes are based on the severity of structural defects observed on individual pipes within the system. Here, a survival analysis model was developed to predict the overall structural state of a sewer network based on camera inspection data from a sample of pipes in the system. The convolution product was used to define the survival functions for cumulative staying times in each condition class. An original calibration procedure for the sewer deterioration model was developed to overcome the censored nature of data (left censored and right censored) available for the calibration of sewer deterioration models. The exponential and Weibull functions were used to represent the distribution of waiting times in each deterioration state. Cross‐validation tests showed that the Weibull function led to greater uncertainty than the exponential function for the simulated proportion of pipes that are in a deteriorated state. Using various sample sizes for model calibration, these cross‐validation tests also showed that the model's results are robust to smaller calibration sample sizes. This confirms the model's potential for predicting the overall state of deterioration of a sewer network when only a small proportion of the pipes have been inspected.
A Survival Analysis Model for Sewer Pipe Structural Deterioration
Abstract: The structural state of sewer systems is often quantified using condition classes. The classes are based on the severity of structural defects observed on individual pipes within the system. Here, a survival analysis model was developed to predict the overall structural state of a sewer network based on camera inspection data from a sample of pipes in the system. The convolution product was used to define the survival functions for cumulative staying times in each condition class. An original calibration procedure for the sewer deterioration model was developed to overcome the censored nature of data (left censored and right censored) available for the calibration of sewer deterioration models. The exponential and Weibull functions were used to represent the distribution of waiting times in each deterioration state. Cross‐validation tests showed that the Weibull function led to greater uncertainty than the exponential function for the simulated proportion of pipes that are in a deteriorated state. Using various sample sizes for model calibration, these cross‐validation tests also showed that the model's results are robust to smaller calibration sample sizes. This confirms the model's potential for predicting the overall state of deterioration of a sewer network when only a small proportion of the pipes have been inspected.
A Survival Analysis Model for Sewer Pipe Structural Deterioration
Duchesne, Sophie (author) / Beardsell, Guillaume (author) / Villeneuve, Jean‐Pierre (author) / Toumbou, Babacar (author) / Bouchard, Kassandra (author)
Computer‐Aided Civil and Infrastructure Engineering ; 28 ; 146-160
2013-02-01
15 pages
Article (Journal)
Electronic Resource
English
A Survival Analysis Model for Sewer Pipe Structural Deterioration
| Online Contents | 2013
Data-driven bi-level sewer pipe deterioration model: Design and analysis
| Elsevier | 2020
SEWER DETERIORATION PREDICTION SYSTEM AND SEWER DETERIORATION PREDICTION METHOD
| European Patent Office | 2018