Optimizing Drilling with AI-Enhanced Motor Inspection Reports
We collaborated with a startup to design and implement AI-enhanced motor inspection reports, significantly improving accuracy and consistency in the drilling process.
Understanding Motor Inspection and Reports
Through interviews with SMEs, desk research, and a thorough documentation process, our product team began understanding the problem. We studied drilling motor inspections, current workflows, documentation processes, and identified main pain points and needs.
To lay a solid groundwork for the project, we also examined selected topics from drilling, physics, math, hydraulic motors and laser measurements that were relevant for the ML model.
ML pipelines documentation
The team already had an ML model, but it lacked documentation and failed to consider relevant input details, impacting the recommendations. We meticulously reviewed the datasets and ML workflows, documented the selected features, and integrated new pipelines to include feedback from the app as model input. This process also refined the model by removing unnecessary features and optimizing workflows.
When we work with an ML-Based product, we have special Discovery processes that combine expertise from Design and AI to produce much more user-centered AI Products.
Enhancing manual measurements with data coming from laser-measurements
Predominantly, inspection activities involve manual measurements, with only 5% using laser equipment that offers significantly improved accuracy.
Through and ancillary ML Model we helped to define and implement, we enhanced the manual measurements with additional data from the laser measurements datasets we had (over ten years of science-backed field data on 10,000+ motor runs with 21 million data points).
This contributed to increase accuracy for the 95% of cases in which the measurements are manually taken.
Adding Visual clarity
We designed custom icons representing each of the possible rotor—stator combinations.
This was a proposal from our team, as we detected it could make each report easier to recognize at a glance. It also helped through the whole report generation process, making the input screens clearer for inspectors.
Input reliability improved through an enhanced UX
Data from the inspection needs to be collected with precision, as it's the main input the ML Model will use to generate its recommendations.
We went through this screen several times, refining it and getting to a proposal that was intuitive for inspectors and easy to fill up.
A shared library of UI components
AI-Fit is part of a suite of products, and we find an opportunity to keep consistency and reduce errors by making a shareable Library of UI Components.
We used Storybook to isolate and document each shared UI component, then included the library on both projects. This helped the products to be consistent, adding intuitiveness for users having access to both and reducing UI mantainability.
Our ML experts and backend engineers made a significant contribution. Their initial mission was to deploy an existing ML model, but we went beyond that by documenting the pipelines related to the model and implementing necessary changes to ensure the model produced accurate recommendations based on different types of user input.
Our product team streamlined the process for motor inspectors, adding guiding details to help them input more reliable data. Visual clarity and iterative design were key to developing an effective UI.
The frontend team created a high-performance UI, integrating existing chart libraries with custom components to deliver the necessary user experience.
AI-Fit is a living product that is making its way into the market. You can take a look at the live project by visiting its landing page.