How AI company IntelliSense.io is transforming the mining industry - from Cambridge
Cambridge may seem an unlikely place to bring about a revolution in the mining industry.
But dig a little deeper and you’ll find, nestled for now within Cambridge Biomedical Campus, an innovative and fast-growing technology company that is changing how mines operate from Chile to Kazakhstan.
IntelliSense.io has created a real-time decision-making platform and artificial intelligence-powered applications that are enabling major mining corporations to improve productivity, save energy and reduce their water usage around the world.
The company was founded by Sam Bose, who brings a family history in mining, an understanding of industrial economics and an appreciation of the Internet of Things to his role of CEO.
“I grew up in a mining town in India and have a family background in mining,” he tells the Cambridge Independent. “You wouldn’t believe it, but they were mining uranium. It was tricky…”
Sam spent time in management consulting, running a practice that specialised in machine-to-machine communications.
“It was like a previous generation of IoT,” he recalls. “I always knew this could change the underlying process of any industry.”
IntelliSense began life in manufacturing, after working with Cambridge-based chip designer Arm.
“I had worked with the Arm management team on IoT side and we effectively started from there, back in 2013-4,” says Sam.
“We were building out a hardware solution that was initially being used to collect sensor data, moving from an analogue to digital format, using the Arm microcontroller.
“Arm took it to some of their customers, who were mostly big semiconductor fabrication companies and that’s how we got into manufacturing initially.
“We deployed the first generation of our technology at one of Arm’s customers, a big semiconductor fabricator in Germany.
“Since then we realised that the value is more in what you do with the data you collect – whether it’s sensor data or any kind of ambient data – and how you merge the different data streams to affect equipment or a process. That’s when we started moving up the value chain to become more of a data analytics software company.”
Mining was an ideal home for the technology. An industry on the potentially rocky road to digitisation, there are huge financial and environmental incentives to achieve even small gains in efficiency.
“From a market perspective, we realised the area where our technology gives most value is when things are variable and change every minute or second,” says Sam.
“If you look at manufacturing, things are pretty static. The variability of the process is very low – the changes happen once every few years.
“We wanted to get into a model where we help to improve efficiency on a continuous basis.
“Mining, or any kind of natural resources industry, always has inherent risk because they don’t know what’s coming out of the ground. The variability in these industries is really high.
“Even a small percentage improvement in variability reduction can have a massive impact on the bottom line.”
With a subscription model that requires no capital expenditure, IntelliSense aims to help people and machines make better decisions.
At the root of it is its Brains platform, which takes in sensor data, applies algorithms to create a ‘data lake’ and uses both machine learning and physical models to deliver fresh insights and recommendations. The decision support software is integrated directly into existing control systems or provided to users.
“We are very active in the base metal operation across the world – copper, iron ore and so on. You have the mining site and the plant side, where they are processing the rocks that are coming in,” explains Sam.
“We are very much a data-focused software platform. What we do is take the sensor data for a specific process that we are targeting, and build a digital twin of that process.
“As the data comes in, we are able to correlate and build relationships, and we can then predict how the process is going to perform in the next hour or the next 10 minutes, depending on the granularity they want.
“Based on that process prediction, we identify the right set point that they should run the equipment.
“We close the loop in terms of decision-making by not only giving the operator information on what is coming through from upstream – what kind of rocks, what kind of ore grade is coming – but also tell them if that is different to what they have seen before and what set point to run the equipment on to maximise the output from that process. So we connect back into the control system.
“They use less power, less water and end up improving the throughput because they improve the amount of rock they can process with the equipment.”
Offering discrete applications for processes from mine to market, IntelliSense enables mining companies to optimise the use of water in their pipeline by predicting the demand in real-time and has created the world’s first technology for accurately predicting variables in the grinding mill.
It can also optimise the thickener circuit at the end of the process, which separates metal from waste rock using large amounts of water, by accurately predicting the physical and geometallurgical properties of the feed material.
“We work with some of the mining majors because they have a network of mines,” says Sam. “If this is successful in Kazakhstan, for example, they will roll it out everywhere, because everywhere they have similar problems.”
Among IntelliSense’s customers is Antofagasta, the Chile-based copper mining group, and AngloAmerican, a global mining major headquartered in London that earned $8.8billion in 2017 from mining everything from diamonds to nickel.
“We work across their operations in South Africa, Brazil and Chile,” says Sam. “And we work in Kazakhstan with some local mining companies, including central Asia’s biggest gold miner.”
It represents significant growth for a company still in its infancy.
“We spent a lot of time trying to build the platform and our first specific application,” notes Sam. “There is a lot of hype with AI but most of the traditional industries don’t understand it. They already collect a lot of sensor data. The problem is, they don’t do anything with it.
“This is why we moved out of our hardware business because we realised the mining industry already has a lot of data.
“They also have data gaps or issues where the sensors are not working, or are not calibrated properly, because this is the real world.
“So we have built in workflow sensors where we effectively combine different data streams to come up with a new data point that doesn’t need a measurement.
“We also clean up the data and calibrate much better than they could have done before.”
IntelliSense deploys what Sam describes as ‘hybrid AI’ – a dual modelling approach for greater accuracy.
“We have built a physical model based on the laws of physics. Then we have the machine learning-based models based on the data and algorithm,” he explains.
“The clever thing we do is try to merge the physical model with the machine-learning model because the machine learning model will only work based on the data it receives.”
This also helps to solve the problem of data gaps, not to mention acts as reassurance for those with long expertise in the industry who might be distrusting of machine-led decisions.
“You can’t allow a billion-dollar operation to rely on a machine-learning model that is telling you to do something when an operator with 30 years’ experience says ‘That is bonkers, we shouldn’t be doing that’,” Sam says.
“We have done a lot of work with the mining operations. They are very concerned about any black box models – they want to understand the algorithms and the calculations you are using. Because we have physics-based models, that is a level of comfort. We put a constraint around the machine learning model. That allows the algorithms to stay within some boundaries.
“It’s all part of change management. That is something we’ve learned. You can’t just go in and tell an operator who has been doing this for 20 years that a machine is going to tell you how to do it better.
“We work with the operators so that they trust the predictions.
“Once they do that, they will see the machine is recommending the right way, and they will allow the machine to run the equipment or process for a fixed period of time, doing all the commissioning.
“We also have a simulator that shows what would have happened if you are not using our application.
“Once they are comfortable, we can go into the final stage, which is closed-loop control where the machine can control the system directly. But that is a journey...”
IntelliSense’s own journey into the likes of Kazakstan and Chile was eased considerably by support from the Department for International Trade and Digital Catapult, the UK agency promoting the adoption of advanced digital technologies.
“We had really good support from the start from DIT,” Sam says. “They got us into Chile, which was incredible, and they got us into Kazakhstan and various other places.”
The company’s ambition is vast.
“We have a whole industry to conquer. Our objective is to transform the mining industry. The industry has a lot of efficiencies it can make,” says Sam. “Humans have been mining for the last 150 years, so a lot of the low hanging fruit has been taken – like copper and gold.
“Now they have to dig in really deep in underground mines. And it can be really harmful for humans so they need automated systems to do a lot of the work. The need for our kind of technology is getting higher. As you go underground, you have to dig more, you need more energy, more people and the risk goes up.
“What most of the mining companies are starting to do is build an integrated control centre. They want to remotely control and run these mines – that’s their vision.
“To do that, you need systems like ours which can run them automatically and adapt with changing conditions. There’s a big safety component where AI can help.”
Unsurprisingly, given the take-up, IntelliSense is recruiting.
“Cambridge, as you can imagine, is not a mining hub but attracts a lot of mining companies because they come looking for technologies,” says Sam.
“We are looking for good Cambridge-based data scientists and software engineers who can help us transform this industry.”
Currently employing close to 40 people at its base in the Clifford Allbutt Building on the Cambridge Biomedical Campus, the company aims to double in size this year, and to move to city centre offices off Hills Road. Having taken some initial angel investment, it now enjoys organic growth.
“We have very good customers and we are a very customer-driven company. We drive ourselves with building new applications and we have an eye to do the entire value chain, from when they are digging in the ground to when the copper goes out the plant.”
It is some pipeline.
“We have some aggressive growth plans,” says Sam. “We are excited about it.”
How IntelliSense was catapulted to Kazakhstan
In Kazakhstan, geological data is a national asset.
With help from the Department for International Trade, IntelliSense was selected to work with the country’s government on a joint venture there. This meant its platform needed a multi-cloud architecture so that it could be configured as a local platform and deployed with a local data centre.
And it needed to analyse three years’ worth of data from six thickeners used in the mining process, which each measured around 800 different metrics every minute.
Enormous computing power and specialist expertise were required so IntelliSense turned to the Machine Intelligence Garage, Digital Catapult’s AI programme that helps businesses access the computation power and expertise they need.
IntelliSense received access to cloud computing resources on AWS, including the latest GPU technology for machine learning.
It allowed the company to develop and train larger models on these huge data sets, reducing the time to market. It also tested the deployment of its entire platform on AWS, and developed a flexible system that can work across different hardware platforms.
The interoperability was essential to fulfil the obligations of the venture in Kazakhstan.
IntelliSense created a world-first solution that takes in 800 variables a minute, using AI to predict how thickeners will operate up to an hour in the future.
It is now testing a framework that will allow it automatically to train models.
International Trade Secretary Liam Fox praises company
Dr Liam Fox, the International Trade Secretary, said: “Companies like IntelliSense.io are transforming the way we extract our valuable natural resources. In an industry characterised by unpredictability, promoting sustainable development in the UK mining industry is vital. Thanks to support from my department and through collaboration with Digital Catapult, the business has been able take the first step in its exporting journey for its AI solutions, reaching markets such as Kazakhstan and Chile. The Department for International Trade will continue to support like-minded businesses through our Export Strategy as we look to drive exports as percentage of GDP to 35 per cent.”