CASE STUDY
Quantum Utility Unlocked for the Life Sciences Industry
Access world-class algorithms, build, tune and run hybrid quantum models and unlock business value today
The Problem: Chemical Mixers Design
Efficiently designing optimal chemical mixers and reactors is critical for improving manufacturing processes across many industries like pharmaceuticals, agriculture, and more. Better mixer designs directly enable higher product quality, increased yields, reduced costs, and decreased environmental impacts. However, identifying the ideal mixer shape to maximize mixing performance is incredibly challenging. Testing physical prototype after prototype is slow and costly.
The Objective
The goal is to optimize the geometry of a Y-shaped fluid mixer to maximize mixing quality. This is evaluated computationally by a coefficient of variation (CoV) metric. However, minimizing the CoV by tweaking geometry parameters is challenging and new optimization methods are required to make this process faster, cheaper, and more effective so that chemical innovation can be accelerated.
Our Solution
Using our proprietary TetraOpt algorithm, a leading specialty chemicals company achieved a 2.3x lower cost function versus standard methods for optimizing the geometry of a chemical mixer, enabling significantly improved mixing efficiency, by leveraging parallelization and more global search. This demonstrates the potential of quantum-inspired algorithms to accelerate and enhance chemical engineering design processes.
Performance Metrics
- Lower the cost function versus standard Bayesian methods for optimizing the geometry of a chemical mixer
- Increasing mixing efficiency and quality by finding the design that requires the least energy consumption
- Minimize the time and computational effort required to identify optimal mixer designs
45%
Improvement in solution quality versus the customer’s existing approach to optimizing mixer design
2.3X
Lower CoV (coefficient of variation) versus existing Bayesian optimization methods
35%
Decrease in runtime and computational effort required versus the existing approach