Engineering design optimization

New buildings are constructed every day and old ones retrofitted, and they all require foundations. Currently, engineers design foundations based on soil information and experience. These foundations are of high quality, but not of the most optimal cost.

Optimizing engineering structures, such as screw and driven piles, allows to maintain the quality standards, while reducing the amount and type of materials needed. Moreover, optimization algorithms can provide multiple scenarios in most cases, serving more as a complex ML tool for an engineer, rather than replacing the profession. The Optimization algorithms can maintain the high quality as required by the local engineering guidelines due to the customized cost function(s). The cost function essentially encodes the formulas and the process the engineer completes when designing the foundation. This allows hundreds of thousands of possibilities to be run through a simulation in minutes exactly as an engineer would and thus maintain the explainability and result quality as required by the engineering guidelines.

Such approach is applicable to virtually any engineering design, from designing a lug nut to an electronically operated hybrid pump. The encoding approach will hold as long as forces, stresses and material properties can be quantified and analyzed via engineering formulas, and all exceptions accounted for. An example of an engineering problem that lies in a completely different domain would be completions optimization. Completion engineering schedules can be optimized in the field of hydraulic fracturing based on known or modeled rock properties. In completions engineering, fluids with proppant (tiny granules) are pumped down the well at high pressures. These fluids pierce the rock formations around the well in specific locations, creating fracture networks. As the pressures and volumes decrease, fractures do not close fully due to being propped open by the proppant. Modeling the volumes, pressures, fluid composition and proppant characteristics is completed via engineering workflows based on rock properties. These parameters are optimized with a similar methodology to achieve the required fracture network geometry and extent, fit in a tighter schedule and use cheaper and more environmentally friendly materials.