Cheat sheet: tech terms for the water and wastewater industry
To make informed choices about software, platforms & applications, it’s crucial to grasp key digital tech terms. Refresh your knowledge with this cheat sheet.
Navigating the digital solutions market for water and wastewater systems can be overwhelming. To make informed choices about software, platforms, and applications, it is crucial to grasp key digital tech terms and ask the right questions about software technologies.
Refresh your understanding of commonly used terms, categorized into three main groups:
- Tools and Methods: understand the commonly used methodologies in digital solutions.
- Automation and Control: differentiate the various digital methods used for controlling assets and instruments.
- Model types: a brief summary of different types of models used by digital solutions.
While we have covered essential concepts, there is more to explore in the digital solutions landscape. Have we missed any terms? Do these definitions resonate with you? Get in touch and share your thoughts and questions.
Tools and methods
Understand the commonly used methodologies in digital solutions.A digital twin is a virtual counterpart of real-world assets and systems, particularly valuable in asset-intensive environments. Our specific definition emphasizes a holistic digital representation of the physical system, seamlessly integrating data for actionable insights in infrastructure management. This concept relies on a core service or platform that continuously ingests, validates, and stores relevant data, offering modularity without the need for a monolithic system. Integration with AI and IoT enables real-time monitoring, predictive maintenance, and the discovery of untapped opportunities, revolutionizing how companies manage their infrastructure.
(Source: Digital Twins for Wastewater Infrastructure - White paper – Royal HaskoningDHV)
Artificial intelligence (AI) is a computer-based approach that involves automated analysis of data using specialized algorithms, allowing machines to comprehend, assess, and learn from information through mathematical methods. AI-driven systems can memorize behavioral patterns and adapt their responses to align with or influence these behaviors.
(Source: IoT for all)
Key AI components include machine learning (ML), deep learning, and natural language processing (NLP). In water management, AI finds practical use in detecting underground pipe leaks, analyzing CCTV footage for sewer faults, and utilizing unconventional data sources, such as smartphone videos and CCTV, to monitor rainfall intensity. These applications leverage AI's data-driven capabilities for improved water system management.
(Source: IWA)
Machine learning, a specialized field within artificial intelligence (AI), centers on computer methods and algorithms that enhance performance through experiential learning. This involves constructing computer models from data, enabling future predictions without necessitating extra software programming.
(Source: IWA)
Machine learning encompasses various techniques like deep learning, decision trees, k-means, linear regression, among others, offering versatile approaches for data analysis and prediction.
(Source: DEEP LEARNING, John D. Kelleher)
(Source: IWA)
(Source: Regulating ChatGPT and other Large Generative AI Models)
(Source: Cloudera)
(Source: An Explainable Artificial Intelligence Approach for Unsupervised Fault Detection and Diagnosis in Rotating Machinery)
Automation and Control
Differentiate the various digital methods used for controlling assets and instruments.
(Sources: ADVANCED PROCESS CONTROL, Mark J. Willis & Ming T. Tham & Inst Tools)
Data logging refers to the practice of gathering and archiving data over a duration to identify patterns or document data-related occurrences within a system, network, or IT setting. It facilitates the monitoring of all engagements involving the storage, access, or modification of data, files, or applications within a storage medium or application.
(Source: Techopedia)
(Source: PLCdev)
A practical illustration of PID control is seen in vehicle cruise control, where, for instance, as a car ascends a hill, its speed naturally decreases if the engine power remains constant. The PID algorithm managing cruise control promptly restores the car's speed to the desired level without delay, achieved by carefully increasing engine power output. (Source: PLCdev)
Digital control, introduced to wastewater treatment plants in the early 21st century via programmable logic controllers (PLC) and supervisory control and data acquisition systems (SCADA), revolutionized the industry. SCADA systems rely on real-time sensor data to optimize mechanical equipment operations, ensuring key variables meet predefined targets. This technology's strength lies in its reliability, automating most processes based on actual data. SCADA's broad applications extend to controlling and monitoring various industrial processes and equipment, offering remote data access and management. Often, it collaborates with PLCs, PID controllers, and other instruments to form a comprehensive system, though some components may operate autonomously.
(Source: Proces Control Experts)
The SCADA system continuously monitors and controls the industrial process, creating a feedback loop. As conditions change or issues arise, the system can respond in real-time, ensuring the process remains efficient, safe, and productive.
(Source: RT Engineering)
(Source: PDFSUPPLY)
(Source: Advanced Process Control: Beyond Single Loop Control, Cecil L. Smith)
(Source: Inst Tools)
Predictive control is a general term that refers to a class of control strategies where the controller makes decisions based on predictions of future system behavior. Although Predictive control and Model Predictive Control (MPC) are terms that have been used as synonyms in literature, MPC originally involves the use of an optimization function aligned with a constrained dynamical system that simulates the process that is being controlled.
(Source: MathWorks)
Predictive control is a broader term that includes other approaches like data-based learning techniques that are used to predict the behaviour of control targets like regression models and machine learning methods. The use of data-driven models is more common on situations where it is hard to build a good model given large and complex processes involved.
(Source: Predictive control of a water distribution system based on process historian data)
Model types
A brief summary of different types of models used by digital solutions.
(Source: Between the Poles of Data-Driven and Mechanistic Modeling for Process Operation & A review of mechanistic and data-driven models of aerobic granular sludge)
(Source: Between the Poles of Data-Driven and Mechanistic Modeling for Process Operation & A review of mechanistic and data-driven models of aerobic granular sludge)
Predictive control can take advantage of hybrid models, which merge both data-driven and mechanistic modeling techniques. These hybrid models blend the mathematical depiction of a system's internal mechanisms with parameter fitting or regression using process data. It's important to acknowledge that in the literature, the term "hybrid modeling" has been commonly employed, but its definition is somewhat unclear since it can encompass various other modeling approaches. For instance, calibrating a mechanistic model can be viewed as a hybrid modeling approach, as it retains the reflection of all pertinent internal mechanisms. Another approach involves generating data points from a mechanistic model and utilizing these points to train a data-driven model. However, the outcome of this hybrid modeling strategy is a purely data-based model, devoid of information about the system's inner structure.
(Source: Between the Poles of Data-Driven and Mechanistic Modeling for Process Operation)
