The recent boom in AI (Artificial Intelligence), and especially the media attention surrounding it, have sparked a fierce debate about the use of AI in different sectors. The water sector is historically not big on innovation, but with its global importance and future stress caused by climate change, there may very well be a push to innovate further.
An article from 2022 (see the academic resources section for more information) discussed the potential role AI could play in process engineering. In this article we will set out some key points of the potential role of AI in the water industry, focusing on the process engineering part of the industry.
AI in Process Engineering
An article from 2021 set out expectations from the authors with regards to the future of process engineering (Thon et al., 2021; link, open access). The article describes the current state of AI in process engineering. To understand this we need to look into the basics of what AI can do for engineers, and what it can’t (yet) do.
AI basics
Artificial Intelligence is a difficult concept to grasp. In it’s most basic form, AI is a computer-based system that can perceive information, synthesize information, and infer information. This means as much that it is a system that can identify patterns beyond what regular computer software can do, and use these patterns to identify new logic. The theoretical applications are endless!
Process Engineering and AI
Modern processing plants generate huge amounts of data, a lot of which is used for plant and process control, as well as modelling and calculations for improvement. Some of this data, however, is merely generated as direct process data (such as temperatures and flow rates), but these data points are not used for anything other than process control.
AI has the ability to quickly process huge amounts of data, opening up opportunities to analyse previously unused data. This could lead to newly discovered patterns in the plant, increasing operating efficiency and improving safety and profitability.
Some more examples of areas with potential improvements by AI include real-time optimisation, design of energy recovery networks, and large scale equation-based process modelling.
Is Water Treatment Process Engineering?
Finding yourself on this website probably means you know the answer to this: yes, water treatment is an incredibly broad (but also niche) part of process engineering. This means that some of the AI applications mentioned above could also hold for the water industry. The main problem with the water industry is that its systems are generally incredibly large and standardised, not to mention that the industry is very slow in terms of innovation (looking at you, sewage treatment plants). The important question is then, what are the overlapping features of water treatment and process engineering? To name a few:
- Process control: flow rate and temperature are incredibly important parameters to ensure consistent water quality.
- Waste and byproduct generation: wastewater treatment especially generates a lot of sludge which can be measured. Process optimisation is partially focused on reducing the water content of the sludge.
- Chemical usage: input chemicals that are used for treatment form an important part of the plant’s overall cost.
These areas are where optimisation happens or needs to happen, and this is where general process engineering principles can be used.
Use of AI for Water Treatment
Water treatment is a broad and diverse branch of process engineering, with several key automated and manually operated processes. AI has the potential to fill the gap between manual operations, for which a ‘feel’ for the system is often needed for fine control, and fully automated control systems, for which large datasets and calculations are used.
Specifically, the following AI applications for water treatment systems have been described as being successful to some degree:
| Variable | Inputs | Outputs |
| Estimation of adsorption capacity of heavy metals by pumice, resins, and activated carbon during operation | Concentration (inlet and outlet), mass of sorbent materials, contact time, pH and other water quality inputs | Adsorption capacity and removal efficiency |
| Calculation of the adsorption capacity of organic compounds, pharmaceuticals, drugs, pesticides, etc. | Water quality inputs, temperature, concentrations (in/outlet) | Adsorption capacity and removal efficiency |
| Removal of dyes such as methyl orange and methylene blue | pH and other water quality indicators, sorbent dosage, metal ion concentration (used for the dyeing process). | Adsorption capacity and removal efficiency |
| Hybrid AI techniques, in which multiple types are combined to determine multiple inputs and outputs. | Multiple inputs and outputs relating to the variables listed above | -variable- |
It can be seen that different inputs and outputs can be used for different processes, but industry experts will quickly realise that many of these variables listed as inputs are standard operating parameters. The reason for using AI in this case is not that humans would not be able to calculate the adsorption capacity of a carbon, for example, but that rapid processing of multiple variables is much easier using a regression classification or a pattern recognition system than it is using a standard model.
Purely from a scientific point of view, AI can be immensely helpful, as was shown in 2022 by a team of researchers from the University of California. The team discovered a new filter design to more efficienty remove boric acid from tap water using a membrane with aquaporin proteins. They tested thousands of layouts of the membrane, coated with different layers, to calculate the optimal water flow while removing as much boric acid as possible. This learning can be applied to other systems too, made easier by the already trained model.
Limitations of AI in water treatment
Currently, AI is not widespread in the water treatment industry, due to a number of issues. As is the case with regular modelling, AI needs a lot of input data to train and verify its results. This means that uptake is slower, especially in an industry tradionally seen as slow to pick up new technologies.
Future of AI in water treatment
In essence, the use of AI enables process engineers and operators to assess very large datasets in a very short time. This will be useful for many applications within the water industry, such as reducing OPEX and predicting emergencies. Autodesk estimates that a 20-30% OPEX saving is possible using diverse AI techniques, for example by identifying leaks based on user data (which is currently very difficult), and by reducing the cost of chemical additives to the treatment process by predicting water quality more effectively.