This fact combined with the application of data analysis techniques such as outlier detection and machine learning using back-end software allow for faster response times to possible contamination events, collection and analysis of large spatio temporal data sets, and diagnostic capabilities for water treatment systems, as just a few of the potential benefits of realtime water quality monitoring. The work presented in this manuscript describes the design, procurement, building, and testing of an IoT-enabled monitoring system prototype for POU applications with these goals in mind. The proposed system consists of a sensor package that includes basic water quality parameters of pH, EC, ORP, Dissolved Oxygen , and temperature. The system uses a Raspberry Pi 4 Model B microprocessor. It includes a software program that enables data transfer from the sensors to an online repository where any collaborators who have access can view the data.
Preliminary algorithms that utilize reference correlations, statistical analysis techniques, hydroponic bucket and physical and chemical constants to calculate and predict other water quality parameters such as corrosivity risk have also been developed.Wireless sensor networks have been applied to an array of environmental monitoring problems. Those focused on water quality monitoring cover a range of applications for environmental waters, drinking water, and wastewater. Some primary questions addressed by previous work include lowering the costs of wireless sensor networks, optimizing power usage for proposed systems, developing software for contamination detection, and improving solutions for data transmission. The works presented utilize an array of programming languages, sensors, circuitry, and communication protocols that all serve to make real-time data monitoring accessible for operators and consumers. Often environmental waters are prone to various perturbances, requiring the need for real-time monitoring. Rivers and marine coasts have been sites for the deployment of smart wireless sensor networks; Adu-Manu et al. employed wireless sensor nodes and used an energy-efficient data transmission schedule to obtain real-time data for pH, conductivity, calcium, temperature, fluoride, and dissolved oxygen.
The author’s system architecture includes a Libelium Waspmote that uses communication modules such as 3G/4G, General Packet Radio Service, long-range 802.12.4/ZigBee, and Wideband Code Division Multiple Access connectivity to transmit data to the cloud. A low-cost Arduino-based sonde was designed in the marine environment that used the Arduino Mega 2560 Mega and ArduinoUno platforms for two design configurations The first design was for a Lagrangian style drifter that Lockridge et al. deployed for 55 hours to measure salinity and temperature using Atlas Scientific K 1.0 Conductivity Probe and the Atlas Scientific ENV-TMP temperature probe. The data collected by the drifter was compared to values from the Dauphin Island Sea Lab Weather Station YSI 6600 sonde. The RMS error was 1.35 ppt for the salinity and was 0.154 °C for the temperature measurements. Salinity and temperature regressions were also performed and found to be highly correlated with R2 = 0.96 for salinity and R2 = 0.99 for temperature. The first data logger demonstrated consistently higher salinity results than the second, whereas temperature tracked similarly for both data loggers over the entire deployment. The system contributed significantly to verifying sensor performance for harsh aquatic environments. However, the work only stored data locally as .csv files and did not transmit data wirelessly or in real-time.
For IoT-enabled systems, some of the most straightforward system configurations are proposed by Vijayakumar & Ramya and Pasika & Gandla . Work by Vijayakumar presents a low-cost system for developing a real-time water quality monitoring system in the IoT environment. The core controller used is the Raspberry Pi B+, and the system has temperature, turbidity, pH, conductivity, and dissolved oxygen sensors. The Raspberry Pi runs on a LINUX kernel and uses an external USR-WIFI232-X-V4.4 module to transfer the data to the internet and is visible on the ioBridge Server. Pasika et al. propose a system consisting of four sensors with the Arduino Mega microcontroller. Their approach is designed in Embedded-C and accomplishes data transmission using the ESP8266 Wi-Fi module. Authorized users can access the data using a User ID and password to log into a Thing Speak server. The information is gathered, stored, analyzed, and transmitted in real-time.Similar systems to those mentioned previously build upon similar design principles but include additional features incorporated into the software designs, such as power-saving capabilities, user alerts, and contamination detection. Power saving capabilities are primarily emphasized for more rugged environments where connectivity issues and power loss can be concerns. One such area is Malawi, the site for a proposed integrated sensor network. The solution was to develop a low power gateway node that reduces energy consumption using a wake-up mechanism that triggers waking and sleeping modes. A gateway solution built upon the ALIX2 embedded Linux board provided a way to interconnect different networks to address connectivity issues. Parra et al. also addressed power-saving algorithms for their fish farm monitoring wireless sensor network, which was achieved by only sending data if the difference between predesignated reference and threshold values were exceeded. Thus, reducing the amount of data moved to the external web server employed in their system and saving computing power. Threshold values can also be employed for systems alerting users of parameter exceedance of WHO or EPA guidelines. This is the case in the IoT-enabled system developed by Geetha & Gouthami which measures conductivity, turbidity, water level, and pH. The hardware design consists of a TI CC3200 single-chip microcontroller with an in-built WIFI module, and ARM Cortex can connect to the nearest Wi-Fi hotspot. The data collected by the sensors is sent to the cloud, and if the values exceed threshold limits obtained from WHO, then an alert is sent to the mobile application developed as part of the proposed system. The programming for the software design that shows the real-time updates was done using ENERGIA IDE, and data are stored in the Ubidots cloud, which includes a real-time dashboard to analyze data, control devices, and shares the data through public links. Lambrou et al. further developed the event detection capabilities of wireless sensor networks by designing and developing a low-cost network embedded system consisting of a central measurement node, a control node, and a notification node. The system contained sensors to measure temperature, turbidity, electrical conductivity, and pH. These three subsystems served to collect water quality measurements from the sensors, implemented contamination event detection algorithms, and stored the measured data in a local database, visualized the data in the form of charts, and sent email/SMS alerts for contamination events. The system architecture utilizes a PIC MCU , an ARM processor, and a ZigBee RF transceiver. The data is posted to the web using the Pachube IoT platform. For the contamination event detection algorithms, the first system was denoted as the Vector Distance Algorithm with a risk indicator function estimated on the Euclidean distance between the normalized sensor signal vector and the normalized control signal vector of clean water. The second was the polygon area algorithm which calculated a separate risk indicator function estimated on the ratio of the polygon area formed by the vector components of each sensor on a two-dimensional spider graph. To validate the event detection algorithms, intentional contamination was performed using E. Coli bacteria and Arsenic, stackable planters which was added to potable water at various concentrations at discrete time intervals. Overall, the PAA had better performance for both contaminants due to fewer false alarms.It is with these developed wireless sensor networks in mind that the prototype in this work has been uniquely designed to combine, improve upon, and fill in gaps from the existing technologies and methods for applications specific to point-of-use water filtration applications.
After investigating the current hardware and software utilized for wireless sensor networks, a key research question to be addressed was to determine the sensors and algorithms that were best suited for smart point-of-use water quality monitoring. Therefore, some of the existing relationships between various sensor parameters and water quality concerns were reviewed to better guide the sensor selection process. Here, we present some of the findings on what water quality information can be related to pH, ORP, DO, and EC. For acute effects from biological contaminants, it is imperative that sufficient disinfection levels are maintained. In drinking water, this is primarily achieved with Chlorine and Chloramine, which previous research has attempted to link to ORP measurements. Experiments have been conducted that investigate the ability of ORP to be a predictor for kill level of organisms including total coliform, E. Coli, and enterococci by measuring the ORP in conjunction with Chlorine dose for wastewater . It was found that the ORP increased with an increase in chlorine added, total chlorine, and free chlorine. However, the regression slopes were relatively low indicating low sensitivity of ORP measurements as a function of each chlorine species. As more chlorine was added, a second increase in ORP was observed. From this point, Monochloramine is oxidized to Dichloramine, which has a higher redox potential and produces H+ leading to a lower solution pH. Once all the Dichloramine was oxidized, free chlorine became available, and the third increase in ORP was observed, indicating the chlorine break point after which viable counting of microorganisms was not observed. The control system implemented could continuously detect the break point and determine the correct chlorine does for inactivating microorganisms in the water. The proposed system ultimately can easily detect the shifting point where dominating chemical species change by monitoring points along with the ORP and pH profiles during chlorination to achieve proper disinfection. ORP measurements have also been explored for monitoring and controlling water disinfection for the produce washing industry . ORP ranges between 600 and 700 show that free-floating decay and spoilage bacteria, as well as pathogenic bacteria such as E. coli or Salmonella species, are killed within 30 seconds. Recent research in commercial and model post harvest water systems has shown that, if necessary, ORP criteria can be relied on to determine microbial kill potential across a broad range of water quality. However, it is important to note the limitation of ORP and pH because the effect of pH on chlorine speciation, one must use caution in not having a false sense of adequate disinfection rates at high pH’s. In general, a ten-fold increase in total or free chlorine concentrations does not result in a corresponding proportional increase in ORP millivolts. Their results showed that good water quality likely results in measurements of 650 to 700 ORP if the water pH is 6.5 to 7. Raising the pH to 8.0 lowers the ORP value, as more hypochlorite ions are present. Maintaining constant pH but adding more chlorine raises the ORP to a plateau of about 950 ??. Finally, Myron L Company describes their correlation for predicting free available chlorine using ORP and pH levels. The correlation was obtained from a series of experiments where an exact amount of chlorine in the form of laboratory-grade bleach was added to DI water in a closed system and measuring pH and ORP to create calibration curves for their Myron L Ultrameter II 6PFCE Water Quality Meter. From some of the listed works, it is evidenced that ORP can be used as an effective indicator for disinfection levels in water and has the potential to protect users from potential biological contamination using realtime measurements of ORP in a wireless sensor network. Other correlations between the variables measured in the wireless sensor network proposed in this work and water quality have been discovered utilizing statistical data analysis methods and machine learning techniques. In 2014, Zhang et al. focused on robust online clustering and modified pixel-based adaptive segmentation and concluded that the MoPBAS method was suitable for detecting anomalous sensor readings and event clustering from salinity and turbidity measurements, which can assist in addressing root environmental causes and significance levels of disturbance events. Forough et al. used an alternative machine learning technique known as the support vector machine model topredict a water quality index using pH, DO, TDS, temperature, Nitrate, phosphate, BOD, turbidity, and fecal coliform data obtained from water samples collected over 11 months. Their SVM model, developed in MATLAB, successfully explained 87% of the variability in total WQI, with Nitrate being the most important attribute influencing the WQI as calculated from the sensitivity ratio.