Due to the changes in soil moisture content and deformation caused by weather changes

there is a correlation between weather temperature and failures of urban water distribution pipelines. Based on the failure records of pipelines and weather temperatures of a northern China city

the correlations between different weather temperature indicators and water mains failures were analyzed. The failure prediction models of water distribution pipelines considering weather temperature factors were established using error back propagation neural network (BPNN) and genetic expression programming (GEP) methods. According to the failure records of water distribution pipelines

pipeline geographic information

and weather temperature records of the case city in the past 11 years

the correlations among pipeline failures and six weather factors were analyzed

including average temperature

freezing indicator

maximum increase

maximum decrease

maximum increase rate

and maximum decrease rate. BPNN and GEP were used to establish the implicit and explicit relationships between the number of pipeline failures (i.e.

the dependent variable) and four explanatory variables (the selected weather temperature indicator

the diameter

age

and length of pipelines). The explicit and implicit models were used to predict the number of pipeline failures in the case city in the next year. The determination coefficients of the prediction results of the model without considering weather factors were 0.65 and 0.60 respectively

while those considering weather factors were 0.78 and 0.88

where the prediction accuracy increased by 13% and 28%. Therefore

it is reasonable and effective to establish a failure prediction model of water distribution pipelines by considering weather factors.