3 Proven Strategies for AI in Manufacturing
Haven’t deployed AI yet? See the highest ROI use cases for manufacturing.
Organizations in the $3.5 trillion metals and mining industry face constant pressure to optimize production and avoid unexpected costs. A leading manufacturer in this industry was struggling with a critical challenge at one of its key mineral processing facilities: unexpected downtime from unpredictable equipment wear and tear.
The downtime stemmed from industrial jaw crushers, which are used for breaking down large rocks and preparing them for processing. Over time, the jaw crusher's performance would degrade, leading to an increase in the size of crushed rocks. If these rocks exceeded a certain threshold, they would cause blockages and damage downstream equipment, resulting in unplanned downtime and millions of dollars worth of lost productivity.
Due to several variables in the production environment, monitoring wear and tear and predicting when the jaw crushers would require maintenance was difficult. As a result, the team at this leading manufacturer decided they needed a better solution for monitoring jaw crusher performance.
To address this challenge, the company developed and deployed a vision AI solution using Roboflow. By running a purpose-built vision model, the new solution records the dimensions of rocks moving along the production line. With this data, the team can observe trends in rock size over time and identify subtle changes in the performance of jaw crushers, allowing them to proactively schedule maintenance and avoid costly downtime.
The AI model was trained on an initial dataset of 400 sample images, showing rocks in the production environment, which was expanded using data augmentation techniques. Thanks to Roboflow’s dataset optimization and training tools, the company created an accurate AI model and deployed it quickly, without having to wait to collect a larger number of sample images.
“Global manufacturers are finding new ways to generate value and avoid millions of dollars of surprise costs with custom-built visions like this,” explained Jim Duffy, Head of Field Engineering at Roboflow. “With a modern AI platform like Roboflow, it’s possible to create and roll out highly accurate AI systems in days instead of months.”
To run the purpose-built AI system, the organization installed a high-resolution network camera over each production line, along with a lightweight yet powerful edge compute device. This edge AI solution is capable of processing the live feed, evaluating the sizes of hundreds of rocks at any given moment, in real time and without internet connectivity.
Finally, by leveraging Roboflow’s integration with a wide range of industrial systems and standard protocols, the system was configured to record the AI-powered rock measurements to the company’s data analytics platform. This integration was key in enabling the team to track changes over time and identify anomalies in jaw crusher performance.
In the metals and mining industry, where a single jaw crusher costs hundreds of thousands of dollars, vision AI has enabled this leading manufacturer to transform their equipment monitoring and maintenance strategy. By detecting anomalies early and avoiding unplanned downtime, the company is set to save millions by avoiding surprise costs while increasing throughput and total yield.
