AspenTech enters hybrid modelling era
CEO Antonio Pietri discusses his lighthouse pilot for hybrid modelling capabilities
Aspen Technology (AspenTech), founded 38 years ago, was part of the first generation of entrepreneurial software firms that identified an opportunity in the spiraling volume of data being created in the industrial process industries. As digitalisation progressed from its beginnings in the 1970s and accelerated through the 1980s and 1990s, it developed applications from advanced process control (APC) to real-time management for refining. It is dedicated to supporting firms with their digital transformation and optimising assets.
Antonio Pietri has been the president and CEO of AspenTech since 2013. From a background in chemical engineering, he has worked at ABB Simcon before taking a position in 1993 at Setpoint, which was acquired by AspenTech in 1996. Before his current role he was executive vice president of worldwide field operations, where he led global sales, sales operations, professional services as well as customer support and training.
The digitalisation trend has become established over many years. What areas are at the cutting edge?
Pietri: We spent our first 34 years of AspenTech focusing on optimising the engineering and design of assets and optimising operations. Our customers create about $50bn in value every year through the installed base of our products. In the last three to four years, we moved into maintenance, or asset optimisation. We saw an opportunity because technology was not being applied to improve reliability. We acquired companies that brought us machine-learning, multivariate analytics and enterprise utilisation, cloud deployment capabilities, Edge sensor units and internet of things (IoT) capabilities. This provides the ability to accurately predict equipment failure, detect process degradation and analyse reliability, among other capabilities. We have been taking that to market over the last three to four years.
We are now doing a ‘lighthouse pilot’ [a test project that feeds back into its own development] for hybrid modelling capabilities, to apply to very complex-to-model processes. Hybrid models allow you to embed machine-learning-derived data models as an object in the first principle model. We have embarked on a complete transformation of our products by embedding AI. We estimate that the $50bn per year our customers currently create could become $200-250bn by leveraging these new capabilities and the wider deployment of our technology.
In what areas can AI add the most value to customers?
Pietri: Certain processes, so-called nonlinear processes, are very difficult to model from first principles—as reactions and reactors can be a ‘black box’. A lot of [conventional] modelling uses biases and guesses but AI captures how reactions happen—changes in temperature, pressure and volume flows—and a data-derived model can accurately represent dynamics that are hard to explain with physics and chemistry.
We can take a data-derived model and embed it as an object in a first principle model. This creates a more accurate representation of the process, leads to better control of that process and better predicts the quality of the product coming out of that reaction.
“AI is about the ability to derive deeper insights from data”
One of the challenges of deep-learning APC has been [understanding] the capabilities in our customers' organisations. One of the goals of AI is to automate knowledge work—a reduction in the expertise required to implement a technology democratises its use. If it is easy to deploy and maintain a solution, the technology can be self-sustaining.
Ultimately, AI is all about value creation—how can we make the industry safer, more sustainable and more profitable? If you can predict that a critical piece of equipment will fail in 30 days, you can prepare for a planned shutdown of the plant. An unplanned shutdown, resulting from an emergency, could lead to gases being released or perhaps even an accident.
AI algorithms are typically based on a single company’s assets. Could a more collaborative approach be adopted, so that companies can benefit from data collected from similar installations?
Pietri: At the moment, we have very specific use cases with data coming out of a single piece of equipment and its surroundings. Eventually, I envision more deployments on the same use case, so we will see patterns across different types of customers using the same equipment. We could then aggregate data and learn from it.
The AI sector is becoming increasingly crowded. How can customers differentiate between solutions?
Pietri: Customers are bombarded with ideas and opinions, so it is difficult to sift through all the noise. They are very pragmatic about what to pursue and they ask, ‘what are the highest value use cases that are simplest to capture?’ Over time, the market will sift winners from the losers.
The technology is not the barrier to entry. Google AI is open source so anyone can get their hands on it. You can hire data scientists, identify a use case, found a company, and off you go. The barrier is how you package that technology. Being able to scale a technology is important—to scale AI you need engineers rather than data scientists.
There is a lot of hype around digitalisation. Is it possible that there will be a boom and bust?
Pietri: Venture capitalists in the US have funded about 2,600 AI companies across all sorts of industries. We have learned over the last few years that there is a set of three ‘success’ criteria to scale AI implementations — data wrangling/management, the AI algorithms but wrapped with ease-of-use and workflows so that it can be scaled by engineers, and domain expertise to properly guide the AI technology. To achieve real-time AI, you need to do all this in real-time.
There is still money being invested in new companies, but investors are being much more careful. Certainly, there are a lot of companies with valuations that are not justified by their business model or revenue. I do not know if there will be a ‘dotcom-like’ situation or a soft landing. But there is no doubt that there will be a levelling out.
As some end users do not yet necessarily understand all the complexities within the technologies, they tend to gravitate towards companies they trust. Incumbents have an advantage as they have relationships, domain expertise and understand their customers.
When the dotcom bubble popped in 2001, a number of companies went bankrupt and everyone thought “e-commerce” was finished. But over the last 15 years new business models and companies have emerged—Amazon, Facebook, Google and Uber—which have completely disrupted well-established industries.
Similarly, in 20 years’ time, the companies that had not pivoted today will likely have fallen behind and lost their competitive advantage, especially because industry is also being challenged from a societal license-to-operate perspective over climate change, plastics and waste, and this challenge will only speed up over time.
How will digitalisation and AI feed into the energy transition?
They will allow companies to remain competitive and sustain their advantages while they figure out new business models. This is already happening—oil companies are transitioning to become gas/renewables and petrochemical companies and chemical companies realise we are moving to a circular economy where plastics need to be recycled, naturally decompose or reversed to their original monomers.
AI is about the ability to derive deeper insights from data. It will be used to research new types of compounds and products at a more granular level. It will create new types of process technologies and new efficiencies. Eventually, we can start thinking about autonomous plants that operate with very little human involvement. In 50 years, it may take just four people to run a refinery—and that refinery will be feeding an ethylene cracker to produce plastics rather than producing fuels for transportation.
It is not about what is going to disappear, rather what purpose these assets will serve in the future. This industry has always been at the forefront of technology because it operates very complex and dangerous assets. Increasingly, technologies ought to be driving sustainability, mitigating emissions and improving safety.