Celanese's Vision for an Autonomous, Self-Optimizing Plant Powered by Generative AI
Celanese is a specialty materials and chemical products manufacturer based in Irving, Texas, with about 50 plants and 13,000 employees worldwide. I last spoke, Ibrahim Al-Syed, the company’s director of digital manufacturing, at the Advisory Group Forum in Orlando in February and followed up with an article, published on April 24, about his company’s digitalization efforts. Since that time, the company's digital data platform provider, Cognite, announced that Celanese is also leveraging generative artificial intelligence (AI) tools as part of its digital transformation. The following is an edited version of my follow-up discussion with Al-Syed.
How are the company’s digitalization efforts progressing since we last spoke?
One of the key things we've been planning to do in 2023 is scaling the (Cognite) platform, bringing all the data together, putting the right context, the right meaning to it, getting it contextualized and modeling it. As part of that investment, we’re using artificial intelligence and generative AI capabilities. But our artificial intelligence journey or generative artificial intelligence is only as good as our underlying data. So, the biggest effort for us has been to standardize the data on common data models, bring it all together, contextualize it and then start leveraging AI capabilities on top of that.
We were committed this year to scaling that to nearly 30 plants. That is in progress. We've gone through the first wave of about six or seven plants already. We have all the data ingested in the platform. We're hoping to complete the rest of our sites by December. And while we're doing that, we've already started working on leveraging large language models like AI to create a game-changer technology that will allow us to develop this AI-powered mission-control system where you could ask questions like, "What is the production status of this facility in Frankfurt? What's the production status of our Clear Lake (Texas) facility?"
In our previous conversation you talked about the goal to create an autonomous, self-optimizing plant. What does that mean?
You have to look at it in two parts. The first part involves taking the industrial data that you have in your manufacturing facility and accessing it remotely. And then the second piece of that is ensuring the computer or the engine or the artificial intelligence, whatever technology you're using, can interpret the data, understand it, develop an insight, and then also create the actions, and then go back and execute them. That's our vision for a remote, autonomous self-optimizing facility. For example, can I predict a failure of a compressor? The first thing we need to know is whether we have the right sensors and IOTs or devices that can give us the insight and then apply artificial intelligence to make those predictions.
But we don't want to just stop at predicting it. We also want the autonomous piece of it. So, if I know a machine is going to fail, what is the system going to do about it? Then you want to generate the insights. So, do I have a backup compressor or a backup pump for the action? Can I initiate an action to switch the process out to start up that backup compressor? Can I create a work package and mobilize maintenance technicians before the failure happens and plan the job in a way to prevent a production disruption? All of these things, if it can happen more autonomously, more automatically will be the game-changer.
You mentioned you've rolled this out to about six, seven plants so far. What's the greatest challenge you face in such an ambitious project like this?
One of the things we have to recognize is we are a company that has been around for a very long time. And we have different legacies in our facilities, which means some plants are 30 years old, some plants are 40 years old. The hardest part of this is, how do I take different businesses, different plants, different ages and different data types and data structures and bring it all together on scale so that I can standardize it and connect it in a way that it can define and drive insights at scale? The only way as a company we can transform is if we can standardize things, and we can do it at scale. And the biggest challenge for us is how do we address the data problem across our multiple facilities, which has a legacy that needs to be addressed and reconciled with?
And this is where a mix of technology and some effort from our part is foundational in getting there. That's one challenge. The second challenge we also have is: Our journey is around people, people-centric transformation. And it's important that whatever we're doing in terms of the transformation across our facilities is that we put emphasis around change management. It's going to be as good as people use it, and they'll value only come from people using stuff. And that can be a challenge, especially if you're a company that is across 27-plus countries with different cultures, with different experiences, different dynamics of people, and doing that at scale can be a challenging experience. So those are some of the key parts of our journey that we're working through.
How do you address that people issue? Do you have a formal training program in place?
Yes, we are trying to communicate globally at scale to explain what is happening, why it's happening, why it's important. The second piece for us is, as you develop your transformation strategy, is it has to be by the people, with the people, for the people. It cannot be an effort that corporate-wise we're pushing things down to our manufacturing facilities. We very much like the concept of product owners. Product owners at the facilities drive their use cases and solutions, so that they own it, they drive it, they implement it at the site. We facilitate that process, and we provide the resources to help do that.
But the good thing about that is we are trying to give people the skills, the tools and the ability to build solutions and build use cases while we are trying to maintain the governance around making sure that everything that is being developed is scalable, is standardized on the data model, so that as we roll up on an enterprise wide strategy of leveraging AI or generative AI technologies and start getting insights into this remote capabilities, it's all flushed out in a standardized way.
The last thing we want is people feeling that we're pushing things down to them, and that balance is quite important.
You have to make sure that whatever you're architecting actually is intuitive and works and addresses the needs of the people. For example, you have this phone, right? I don't need a user manual or training for this. It just works, and I am married to it. I can't live without it. So we have to find the balance of making the right solutions for the people and keeping that in mind. Also, we have developed what we call a Digital Manufacturing Academy that is now available globally for all our users. And that academy is really around giving people the ability to upskill, have more data literacy, more digital literacy skills, and even give people the opportunity to start learning how to code, if they need to.
I wanted to talk a little more about generative AI. Can you provide some real-world examples of how that works in your operations?
We always had this vision around enabling an industrial Google or search capabilities. With generative AI, you can actually search asking in a copilot or a chatbot something like, “Hey, can you tell me what preventative maintenance work orders are due today?” And it can understand the context of that question, that prompt, and then give you the answer with that context in mind.
Generative AI can actually interpret what you're asking in a language model and give you answers. The second thing that's a game-changer for us is the ability to generate content, whether it's through summarization or just providing insights. So, for example, we're looking at use cases where we could say, “My pump has failed. What should I do now? I need help in troubleshooting.” And you have this copilot experience where the generative AI can tell you it had a bearing failure, for example, or whatever failure mode, and here's the maintenance history of what was done on this pump. Here's all the data that you need from documents and previous pressure and temperature readings. And, by the way, do you want me to create a root-cause analysis report for you? So, all your data is on your fingertips. And not only that, generative AI will write the root-cause analysis for you and create that content and summarize all that information for you. In the past, this would've taken a week or two for a team to figure all that out. Another possibility is using generative AI to create documents like piping and instrumentation diagrams, data sheets.
What's next for Celanese and your digital transformation?
I think what's next for us is that we recently completed last year the acquisition of our materials, mobility and materials business from DuPont. And next year we will be onboarding those facilities to our journey once the integration is complete. So that's the next foundational investment for us in terms of making sure that all of Celanese is on one common platform, one data foundation, one knowledge graph that we can build out and then really power any and every use case that we have across our facilities.