While water utility Northumbrian Water measures water quality after it leaves the wastewater treatment plant, its ability to track quality as the water flows to customers then declines. The utility is accountable for water quality across the network and can be fined when quality targets fall below minimum levels. As a result, it conducts 180,000 samples per year by hand. The utility was looking to gain a more immediate knowledge of water quality issues, as well as reducing the risk of fines.
Meanwhile, Babcock International manages a fleet of Warrior armoured vehicles for the UK MOD. During the overhaul process, the stripped-down, bare aluminium hull is scrutinised by hand for damage, flaws and wear in critical dimensions, using ultrasound. The paper-based records produced from those inspections are difficult to use for further engineering actions.
Both projects are two of 14 that took part in a government-funded programme, the Made Smarter Technology Accelerator, a project of the Digital Catapult industry R&D organisation. In phase one, the Accelerator worked with large organisations to refine the problem, launched a competitive tender and provided up to £20,000 funding to find a prototype solution. Four projects were later selected for phase two development, which involves funding of £100,000 to reach MVP, minimum viable product, stage. In addition to the projects profiled below, two others reached this stage: a system of machine learning for manufactured product inspections from Machine Intelligence and BAE Systems, and an optimised project planning and resource tool to be developed by Total Control Pro, ElegantIT and New Intelligence with Safran Landing Systems. The scheme ended in December 2021.
THE ITERATION PROJECT
Northumbrian Water’s vision is to be able to install an automated sensing system upstream of each of its 1,200 district metered areas. “What we’re effectively trying to do is a laboratory experiment in a remote, rugged environment,” observes Roger Singleton, MD of engineering consultancy Riscon Solutions, which worked together with the utility and Internet of Things firm Inventia. A suite of sensors monitor different indicators of pollution: conductivity, turbidity, residual chlorine and pH. Because a number of the sensors use light-based or pressure-dependent methods, Riscon had to design a special manifold where the sensors sit to present the water in a way that would not skew the results.
Another complication is the regulatory prohibition that sampled water be returned to the line. To take a sample, a valve opens, allowing about a litre of water into the sampling reservoir, and then closes again. With a designed sampling frequency of 15 minutes, that amounts to a usage of 80 litres a day.
The sensors, all transducer-based, output data using the Modbus protocol. All four readings, plus a timestamp, are packaged up on an MBIOT shield and transmitted to the cloud, via a buried coaxial antenna, using telecoms carrier Vodafone. The battery-powered prototype unit is about the size of a pet carrier.
A dashboard shows the values of a number of recorded parameters, including temperature and pressure, as both instantaneous and time-series data over the past week. There are alert levels that can be set to whatever the company wants, and a reporting function that can spit out raw data. A statistical analysis report provides more information.
Says Singleton: “This is a great and quick-hit method of giving information to people monitoring.”
He continues: “A common method for risk modelling for drinking and wastewater is called water safety planning. Getting data feedback to inform you when the risks are valid has been very labour-intensive, especially in water quality. There is an opportunity with systems like this of feeding back more live data. I’d also argue that for others monitoring things like pump maintenance data could feed into risk models - we should be bringing data together to make it more holistic.”
Outside of the project, the partners are collaborating on one tweak of the system, by measuring another water quality parameter. “One of the big business case elements about putting sensors in the network is to try to remove manual sampling and lab testing from operators: not totally, but minimise it. The problem with the current element is that nothing replaces what the regulator wants, which is a measurement of bacteria in water,” adds Singleton. In the water quality field, it has been established that microbiological pollution load can be measured by the absorption of ultraviolet light at 254nm, so plans are afoot to integrate a sensor with that function into the device.
‘COPLOT’ INSPECTION SYSTEM
Instead of recording readings by hand, technicians inspecting a Warrior body using an ultrasound probe now have software to do that for them, thanks to the Coplot project, and the work of software developer Jetsoft. The first step involved installing hardware – a Faro positioning arm – to which to mount the Sonatest probe. The next step was to make the software, says Jetsoft director Tom Martin. He says: “It takes in data as a waveform and position, and then creates a 3D map of that component and what you’re seeing inside it.”
As aluminium is a homogeneous material, any reflections before the back wall of the part indicate the presence of a flaw. The software outputs the signal as a heat map. Although the colours are customisable, in this case close reflections show up as red, far reflections (the back wall) as green, and a gradation in between. Doing so highlights potential issues immediately.
Although it would be possible to stitch the data together to get an overall composite view of an entire vehicle or component being looked at, the more important benefit is traceability, observes Martin. Previously, Babcock relied on an inspector’s word, and competence, that the entire surface of the body had been checked.
A further refinement in the project, due later this year, is to rearrange the positioning standard, he explains. “Because the part is larger than the scanning range of the arm, the next stage of development is to move the arm to scan the next area, as we want to stitch the readings together. To do this, there is a three-point correction system, now in development, that will be with reference to the part, not the arm.” That will require mapping the data to a 3D model of the part. Also planned is recording inspection ‘metadata’ such as equipment used and annotations.
But generating data was only part of the goal. The second aspect of the project was to develop a platform to digitise the process, and others, to bring them together in one system. Then the data could be used again, for other purposes. Martin says: “One of the USPs of our software is to be able to create a form, link it to database fields, and capture quality data into a database that is used as a source for improvement.”