[ad_1]
Table of Contents
DTU is partnering with a world leader to leverage quantum-inspired algorithms in the manufacture of chemicals. The goal is to make the process of bringing new chemicals to market cheaper and faster and to create insights that can be applied to the manufacture of other products.
Taking an idea that works on a small scale in the lab and upscaling it to produce large volumes can be hugely time-consuming and expensive. Because even if a chemical process works without a glitch on a small, experimental scale, it is rarely possible to simply take that idea and use it commercially.
Researchers and students as well as new and well-established companies, come to DTU Pilot Plant to test whether their ideas can be scaled up and become commercially viable in the chemical industry. Image credit: Jørgen True / DTU
Most often, large amounts of costly and time-intensive chemical experiments are needed to tweak ideas before you can achieve the same outcome outside the lab.
“Process systems have a multiscale nature – just like our bodies. If I have a pain in my head for example, the root cause could be a microbe in my gut,” Associate Professor at DTU Seyed Soheil Mansouri explains and continues:
“It’s the same thing within the biomanufacturing industry. When we encounter a problem in the process, the solution to address that problem may be found somewhere else. So, we need an insight into how these different scales–all the way from the molecular or atomic scale to the enterprise-wide scale–interact with and influence each other.”
Chemistry – artistic interpretation. Image credit: Shafin Protic via Pixabay, free license
Less brute force testing in the chemical industry
DTU has a state-of-the-art pilot plant that is used to turn ideas that involve chemical and biological processes into commercial reality. In the pilot plant, users can test and validate and qualify strategies until problems are sorted and the idea is commercially ready.
To avoid unnecessary testing, users take advantage of chemical modelling to narrow down the areas that are likely causing problems so they can focus testing on these areas. “Rather than having to conduct experiment after experiment after experiment and failing and failing and failing using brute force, we think more intelligently,” Seyed Soheil Mansouri explains.
In a new project, researchers from DTU will partner with American AI-experts, Zapata Computing, to explore how utilizing quantum-inspired algorithms (see explanation below) can bring an idea to market cheaper and faster by reducing the search space even further, thus avoiding a lot of the brute force experimentation for the chemical industry.
“Bringing together our knowledge of chemical and biological processes with Zapata’s expertise in advanced computational calculations we aim to develop more advanced algorithms that can tackle the problems that we cannot today,” Seyed Soheil Mansouri says.
Revolutionary approach
Zapata Computing is named after the man who started the Mexican revolution – and listening to CEO Christopher Savoie it quickly becomes evident that his company shares Zapata’s zeal for bringing about radical change:
“Most CEOs of technology companies will say: ‘My technology is going to change the world’. They can be selling a dating app and they will tell you that. And so, it hurts me to hear these words coming out of my mouth, but this is really going to change the world in a very significant way in areas that we care about.”
There is no need for quantum computers when working with quantum-inspired algorithms. However, the algorithms which the partnership aims to develop have the potential of being transferrable to quantum computers as and when they become available.
“Our approach in this new project gives us some advantages here and now, making it possible to do more computation in a shorter amount of time,” Christopher Savoie explains.
In the project, the partners will work backwards in the sense that they take problems that have already been solved using the conventional way of working to test if the quantum approach provides an advantage by making the scale up process quicker, cheaper or reducing the cost of experiments.
“Success is finding a problem where we can successfully demonstrate that we can further accelerate the manufacturability of biomolecules using quantum-inspired algorithms. Because that would prove that we are on the right track,” Seyed Soheil Mansouri says.
About quantum-inspired algorithms
Quantum-inspired algorithms are based on math that was originally pioneered by physicists to simulate complex quantum systems. They often involve highly complex constructions that can effectively represent correlations between variables.
Quantum-inspired algorithms can run on today’s classical hardware and can be adapted to run on quantum hardware in the future. This would speed up the transition from high performance computing to quantum machines.
One widely used application of quantum-inspired algorithms is in high dimensional integration, which the financial industry uses to understand risk in complex financial models. The models typically require extensive so-called Monte Carlo simulations for running through a tremendous number of scenarios, which require a powerful classical computer to gain insights quickly and accurately. By doing so, critical decisions can be made that impact how a financial institution operates.
More recently, quantum-inspired algorithms have been used to solve optimization problems, particularly using quantum-inspired generative models. In a recent project with BMW, Zapata demonstrated the advantages of quantum-inspired generative models by applying them to a manufacturing scheduling problem. In the research, a quantum-inspired generative model generated new solutions that beat or tied the solutions generated by traditional state-of-the-art optimization algorithms in 71% of problem configurations.
This approach performed particularly well in problem configurations with the widest range of possible solutions, suggesting quantum-inspired algorithms can be a powerful tool for more complex optimization problems.
Source: DTU
[ad_2]
Source link
