American Water (Canada)

Artificial intelligence (AI) and machine learning (ML) have been completely transformed by deep learning (DL), which provides unmatched power in handling massive, unstructured information from a variety of fields. CNNs, neural networks with recurrent connections (RNNs), generative models, (DRL), and deep learning via transfer are the main deep learning models that are covered in detail in this article. The structure, uses, advantages, and drawbacks of each model are thoroughly investigated. In this work, we use the Fruit-360 dataset, a benchmark picture classification dataset, for an empirical investigation. Six well-known deep learning architectures are compared and evaluated: Bidirectional LSTM, CNN, Simple RNN, (LSTM), (GRU), and Bidirectional GRU. Analysis is done on performance parameters including accuracy, precision, recall, and computing efficiency to have a better understanding of how well they work on various tasks. The purpose of this survey is to give scholars, professionals, and enthusiasts a thorough grasp of the various uses and capacities of deep learning models. The study provides insights that help progress the discipline and direct the selection of suitable models for certain real-world problems.

AWWA Water Science is an interdisciplinary journal that publishes original, refereed (peer-reviewed) research on the science, engineering, and social aspects of water. AWS publishes research exploring theoretical, experimental, and practical approaches that advance fundamental understanding and application of all aspects of water works and any source water that could conceivably become a potable water supply.

Abstract The occurrence of taste and odor (T&O) compounds in drinking water can lead to public concern about quality when left unaddressed, therefore suitable treatment is needed. This study evaluated the treatability of 18 T&O compounds using chlorine, permanganate, and ozone in MilliQ and river water. Nine T&O compounds showed complete oxidation (>99% removal) in all water matrices by ozone. Most of the remaining T&O compounds were removed to >50% by ozone. Alkalinity, pH, and natural organic matter (NOM) had a significant impact on the decay of ozone. Chlorine and permanganate were ineffective (removal <30%) for 2‐methylisoborneol (2‐MIB), geosmin, hexanal, pyrazines, and haloanisoles. Although chlorine achieved complete removal for alkyl sulfides and indole, it was only partially effective for aldehydes and ketones. Permanganate was more effective than chlorine in oxidizing unsaturated aldehydes and ketones. Increased NOM and pH reduced oxidation effectiveness for chlorine, but not permanganate.

Asbestos cement (AC) pipe was used as pipe material starting from 1930s. It was widely used during 1950s–1960s, and it phased out in the US in 1980s due to health concerns and the availability of PVC pipe. It was estimated that 600,000 mi AC pipe were installed in the United States and Canada. Several studies show that AC pipes are experiencing accelerated failure in recent years. To prioritize the AC main replacements, it is important to understand the performance of AC pipe in the water system. This paper analyzes the failures of AC pipe using the most recent 12-year main break data. AC pipes in coastal area and inland area were analyzed separately to understand their break rates. The results of this analysis could help utilities better understand the performance of AC pipe and prioritize AC pipe replacement.

Legionella was monitored in reclaimed water systems and were detected by culture in 50% of 115 samples, while 80% of the samples were positive by quantitative polymerase chain reaction (qPCR). The lowest level of Legionella was in a system that practiced biological nutrient removal and maintained an average 0.3 mg/L free chlorine residual. The difference between qPCR and ethidium monoazide treated samples was used as an indicator of viability, and the greatest difference was in a chloraminated system, where the molecular method indicated a 9% viability. Ninety‐six percent of the species were L. pneumophila, and 87% were identified as L. pneumophila serotype 1. Total amoebae (summed as cysts or trophozoites) were detected in 100% of samples from all six systems, and were mostly (50–95%) in the active trophozoite phase. In the chloraminated system, 87% of the mesophilic amoebae and 66% of the thermophilic amoebae were in the cyst phase.

Physical hardening of a drinking water plant does not eliminate vulnerabilities at the plant or in the distribution systems, which can extend over hundreds of miles. Hazardous agents may be introduced into drinking water either intentionally or accidentally through cross connections and backflows that overcome system pressures. Currently water utilities typically have knowledge of such events after the compromised water quality has been detected via taste/odor issues or public health impacts in drinking water consumers. A comprehensive approach is required to protect distribution system water quality by employing technologies that facilitate “real time” feedback and provide tools to indicate an early warning of unanticipated changes in water quality. The approach being evaluated by the United Stated Environmental Protection Agency in their Water Security Initiative consists of integrating multi-streams of information (i.e., water quality, syndromics, eye witness, and law enforcement). Using the Water Security Initiative as the platform, this chapter will discuss a conceptual model that is capable of integrating information from various technologies (i.e., Automatic Meter Readers capable of backflow, leak/tamper detection) in distribution system pipes to convey multi-streams of information to software-assisted alarms, which can then integrate information from hydraulic models to trigger automated sampling at strategically selected sites for laboratory-based verification of intentional or accidental contamination events. Following water quality aberrations, the utility needs to initiate a Consequence Management Plan (CMP). The CMP needs to be a “living document” designed in a manner that is intuitive, self-explanatory, and is capable of guiding utilities through the most appropriate data collection, analysis, communications, and mitigation steps. This chapter will provide an overview of the wide and varied processes that a utility needs to navigate to return an impacted system back to normal operation as quickly and safely as possible.

Pressure transients in drinking water pipelines (i.e., surges) may cause hydraulic pressure gradients, resulting in the potential for intrusion of pathogens present in the external environment into the distribution system. The objectives of this study were to determine the occurrence of indicator microorganisms and pathogens in the vicinity of potable water pipelines and assess the potential for intrusion attributable to transient distribution system pressure changes. As part of an earlier study (Kirmeyer et al, 2001), soil and water samples were collected at sites immediately exterior to drinking water pipelines at eight locations in six states. Samples were then tested for occurrence of total and fecal coliforms, Clostridium perfringens, Bacillus subtilis, coliphage, and enteric viruses. Indicator microorganisms and enteric viruses were detected in more than 50% of the samples examined. Monitoring of pressure transients at a large distribution system indicated that pressure transients occurred frequently, although negative pressures were detected on only one occasion. The results of this study suggest that during negative‐ or low‐pressure events, microorganisms may enter the treated drinking water through pipeline leaks.

In the past, we measured the filter freeboard (the measurement from the top of the filter media to the top of the washwater trough) quarterly. To accomplish this task, we had to first drain the filter, then, standing on top of the washwater troughs, measure at various places to get an average. This practice was obviously hazardous because someone had to straddle the washwater troughs and enter a confined space to take the measurements. Adding to our difficulties, we did not have a good way to measure filter bed expansion during a filter backwash. We constructed a gadget to perform this task in a safer, more efficient manner. A list of materials and construction information is provided.

A finite-element algorithm is developed to analyze the dynamic response of multiple, jointed concrete pavements to moving aircraft loads. In the finite-element idealization, the pavement-subgrade system is idealized by thin-plate finite elements resting on a Winkler-type viscoelastic foundation represented by a series of distributed springs and dashpots. The dowel bars at the transverse joints are represented by plane frame elements. The dowel-pavement interaction effects are accounted for by employing contact elements between the pavement and the dowel bar. Keyed joint or aggregate interlock joint is assumed for the longitudinal joint and is represented by vertical spring elements. The dynamic aircraft-pavement interaction effects are considered in the analysis by modeling the aircraft by masses supported by spring-dashpot systems representing the landing gear of the aircraft. The accuracy of the computer code developed is verified by the available experimental and analytical solutions. A parametric study is conducted to investigate the effects of various parameters on the dynamic response of pavements.