Computational techniques for simulating and optimizing combustion - QC-185

Preferred Disciplines: Statistics, Computer Science, and/or Mechanical Engineering (Masters or PhD)
Project length: 8-12 months (2 units)
Approx. start date: June 2019
Location: Montreal, QC
No. of Positions: 3
Preferences: None
Company: ThermoAI

About Company:

ThermoAI develops sophisticated IoT technology and state of the art machine learning systems to optimize industrial combustion. Using historical data, we build a "digital twin" of each facility, allowing our algorithms to run millions of simulations to test performance under a wide variety of conditions. Our goal is to boost thermal efficiency for power plants and manufacturers, saving them millions on fuel and reducing greenhouse gas emissions.

Summary of Project:

Power plants typically peak at 33% efficiency on average. Various factors impact performance, including fuel composition, equipment, and weather. We combine causal graphs with flexible regression techniques to simulate and optimize these messy processes with the goal of boosting thermal efficiency and reducing greenhouse gas emissions.
We are seeking motivated researchers to help improve our analytics pipeline, developing faster regression ensembles and reliable optimization algorithms.

Research Objectives/Sub-Objectives:

  1. Review the latest research in combustion modelling.
  2. Analyze propietary data gathered from energy facilities around the world.
  3. Experiment with a variety of regression techniques for simulating industrial combustion, with a focus on differentiable methods.
  4. Test the relative performance of various optimization methods, such as gradient descent and Newton-Raphson techniques, for minimizing emissions and maximizing thermal efficiency.

Methodology:

    Our approach is completely driven by empirical results. No previous expertise in engineering is required. We make no assumptions about the target system, and are willing to test all new ideas against the data to determine viability. Initial experiments will be conducted in silico and trialed on testbed facilities. If results are promising, then we will implement designs in  industrial boilers worldwide.

    Expertise and Skills Needed:

    • Advanced knowledge of Python, R, and/or C++.
    • Strong command of machine learning approaches to regression, with a focus on ensemble methods.
    • Deep learning experience, including familiarity with at least one framework language such as TensorFlow, PyTorch, Keras, or Caffe.
    • Experience in LSTM models a big plus.

    For more info or to apply to this applied research position, please

    1. Check your eligibility and find more information about open projects.
    2. Interested students need to get the approval from their supervisor and send their CV along with a link to their supervisor’s university webpage by applying through the webform or directly to Gabriel Garcia-Curiel
    Program: