Age effects on iron -based pipes in water distribution systems

Laboratory testing was performed on aged pipes with varying degrees of turberculation in order to characterize the relationship between reduced pipe diameter and Hazen-­‐Williams C. These results, combined with a manipulation of the Hazen-­‐ Williams equation, provided a simple and direct method for addressing potential changes in pipe diameter. This method was then applied to a network model using EPANET 2. The network modeling application demonstrates the magnitudes and types of errors that can be introduced to water age in a network model by ignoring possible changes in the flow area of aged pipes. Specifically, it was found that the largest error introduced was in the pattern of fluctuations that occurs as different source waters moved through the distribution system. Introduction Water distribution networks are complex systems made up of pipes, valves, storage tanks, and pumps -­‐ along with many other parts. Due to the complexity of water distribution networks, it is very difficult to comprehend their operation just by reviewing a network schematic. Network models have become a valuable tool in Coauthored by Ryan T. Christensen, Steven L. Barfuss, P.E., and Michael C. Johnson, Ph.D., P.E. 17 understanding the day to day and extreme event operations of networks and for investigating the response of a network to various scenarios. In recent years there has been rapid growth in the usage of network models; still, there are several difficulties faced in developing accurate models. One of the most significant difficulties is the acquisition of adequate data (e.g. nodal heads, flow distribution, pipe characteristics, etc.) in order to accurately characterize pipe networks. This challenge arises as a result of incomplete records, undocumented changes, and because pipe properties such as roughness and flow area may change as certain types of pipes age. The degree to which age based degradation occurs is highly dependent on water quality, pipe material, and the type of coating applied to a pipe (Colebrook and White 1937a; Williams and Hazen 1960; Lamont 1981; Sharp and Walski 1988). The reconciliation of inaccurate network data is generally accomplished through a process of calibration whereby physical data (such as nodal head values) are measured within a network and compared to simulated values obtained from a network model. By comparing the measured values to the simulated values and adjusting the physical attributes of the network model it is possible to improve the correlation between the network model and the actual network by identifying and rectifying inconsistencies in the network model. Throughout this process, it is of vital importance to ensure that any changes made to the input parameters of a distribution network are justifiable based on the supporting evidence (Hirrel 2008). 18 While developing an accurate hydraulic model is a necessary part of modeling water quality, network models that will be used to model water quality have special requirements versus those that will only be used to model hydraulic conditions. Network models that focus on modeling hydraulics are commonly calibrated by comparing the nodal heads within the hydraulic model to heads measured in the pipe network and adjusting the friction coefficients of pipes within the network so that the modeled and measured values are within an acceptable tolerance of error. During this process, potential changes in pipe diameter are normally given little consideration (Boxall et al. 2004; Walski 2004). Although this method is usually satisfactory for situations in which a model will be used for simulating pressures and fire flows, this procedure does not account for possible changes in flow area resulting from pipe age and as a result may not model the appropriate water velocities needed for the accurate modeling of water age. Moreover, many water quality problems including disinfectant decay, disinfection by-­‐product formation, and taste and odor problems have been associated with the residence time of water in distribution systems (AWWA and EES 2002). Research by Hallam et al. (2002) and Clark and Haught (2005) has also indicated that chlorine decay rate is a function of velocity. While this research is not commonly applied in the current network models, future models may account for this dependence which would further reinforce the importance of accurately modeling water velocity. In order to address these issues, several studies have suggested the need to adjust the 19 diameter of aged pipes that have significantly reduced flow areas when modeling water quality (Skipworth et al. 2002; AwwaRF 2004; Boxall et al. 2004; AWWA 2005). For example, Boxall et al. (2004) recommended the assumption that a 1-­‐mm effective roughness height is equal to a 2-­‐mm loss in diameter. While this approach may be accurate in some cases, it is important to remember that the early tests upon which the idea of effective roughness height is based were performed using sand grains as the roughness elements (Nikuradse 1933; Prandtl 1933; Colebrook and White 1937b). As a result, other types of roughness elements may not have a direct physical correlation with effective roughness height. Still, a correlation between roughness and area reduction would be very useful for aged pipes. However, such a correlation cannot be obtained from roughness testing performed on arbitrary surfaces, but instead requires the testing of actual aged pipes. Interest in water quality modeling is increasing. One portion of a 1999 survey commissioned by the AWWA Engineering and Computer Applications Committee sought to determine current and planned applications for water quality modeling (AWWA 2005). Among the applications cited by survey respondents were: replacing water quality monitoring with modeling, obtaining operational information, investigation of water age, and locating and sizing storage tanks. Growth in the field of water quality modeling is expected to continue. However, the potential changes that can occur in aged pipes continue to be a challenge in developing accurate network models, especially with respect to accurately modeling water age. The objective of this paper is 20 to provide guidance that will enable system modelers to improve their estimates of water age in networks with degraded pipes. Laboratory testing has been used to assess the changes that occur in aged pipes. The laboratory test results are presented and the effects of these changes on modeling water age were explored by applying the findings of the laboratory testing to a pipe network. Laboratory Testing Laboratory testing was conducted in order to evaluate the hydraulic and physical characteristics of sections of aged pipe. The primary objective of the physical testing was to investigate methods for estimating the amount of area reduction and to determine the amount of headloss present in the aged pipes. Eleven aged pipe sections were obtained from water utilities and subjected to physical testing. The pipe sections varied in age from 25 to 50 years and in nominal diameter from 0.75-­‐in to 4-­‐in. The majority of the pipe sections varied in length from 20 to 40 pipe diameters. Most were acquired when utilities were repairing or replacing a pipe section. Testing was performed in order to determine an effective diameter and to evaluate the roughness of each pipe section. For this purpose, the internal volume of each pipe section was determined by filling the pipe sections with water and then measuring the volume of the water. An average pipe diameter was then back calculated using the volume of the water and the pipe length with the assumption of a circular pipe cross section. The diameter ratio, d/D, was obtained by dividing this calculated average diameter (d) by the inside pipe diameter (D) in its new condition as obtained from a table of standard 21 pipe diameters for new pipe (ASME 2004). The standard pipe diameter was used because it is generally easy to obtain and in many cases represents the best available estimate for the original pipe diameter that would be available to a distribution system modeler. Table 2 provides a summary of the aged pipe sections tested during this study. %. Because the velocity for a given flow rate is inversely proportional to the square of pipe diameter, a 5% reduction in diameter will result in an increase of 10.8% in velocity for Table 2: Description of Aged Pipes Pipe Section Pipe Description D (in) d (in) d/D Percent Reduction in Flow Area 1 galvanized steel 0.824 0.767 0.931 13 2 galvanized steel 0.824 0.777 0.943 11 3 galvanized steel 0.824 0.781