The Effects of Socioeconomic Status, Geological, Engineering, and Climate Factors on Machine Learning-Dependent Forecast of Water Pipe Failure

1. Introduction

Having a reliable and safe water supply is essential to the WSN's administration. Water distribution pipes, which move water from water treatment plants to customers, are the fundamental components of a WSN. This corrosion is especially bad for subsurface water pipes constructed in US urban centers that date back to the 19th century. The more than 700 water main breaks that happen every day in the United States and Canada result in the loss of over 2 trillion gallons of drinkable water annually. A burst water pipe may cause enormous financial losses as well as harm to the environment or society. According to US Water Service Agency projections, over the next 20 years, the replacement costs of the country's present WSNs and their anticipated expansions would come to a total of more than $1 trillion. Due to these horrible issues, management is under pressure to provide proactive support for loss reduction by adopting management practices for long-term improvement and reliable pipe failure estimate models.

The key to developing precise prediction models is to determine the significant variables—also referred to as input variables—that influence pipe failure. Several factors that might lead to pipe breaking have been assessed over the last several decades using experimental testing, finite element models, and historical data analysis. A recent review suggests that these factors may be broadly classified into three categories: physical, operational, and environmental. The physical characteristics of the pipe that are most often considered are its diameter, age, length, and material. Kettle and Goulter, for example, used statistical techniques to ascertain the relationship between pipe sizes and break likelihood.

The likelihood of longer pipes failing has been shown by Tai P. et al. (2023). The operational factor that has received the greatest research attention is the frequency of previous failures. These studies demonstrate that the probability of a pipe failing is often connected with the number of previous line failures. Water pressure is another common operational element for pipes in the WSNs. It is shown that there is a positive correlation between internal water pressure and the chance of breaking in metal and cement pipes. Pipe problems might also be caused by external factors. The factors include traffic levels, soil classifications, and climate. Furthermore, many of these details are often quite unknown. Numerous meteorological factors, such as temperature and precipitation, have been linked to pipe problems in previous studies. The results indicated that larger temperature variations may increase the risk of pipe collapse. It's critical to understand the interactions between the chance of pipe collapse and these three distinct types of contributing factors. Apart from the above mentioned elements, it is becoming evident that collaboration with diverse situations, such as social and economic concerns, has to be considered while executing forecasting for WSNs. For instance, the effects of configuration collapse and population-related data were considered in recent study on the reliability and adaptability of communities. However, these factors have seldom ever been taken into account by the pipe failure prediction model in use today. In the meanwhile, shareholders are paying close attention to how the major contributing reasons of pipe failures are interpreted in order to make well-informed decisions about the allocation of resources. This is on top of developing trustworthy and practical methods for evaluating pipe failure. Even while previous studies examined how many characteristics affected the chance of a pipe collapsing, it is still unclear what the comparative relevance—or degree of the effect—is. Thus, interpretability plays an equally important role in the development of a pipe collapse prediction model.