“Information is power” is often repeated phrase. The underlying assumption is that information, when accurately transformed into knowledge, becomes the true source of power. The challenge lies in converting data into information and subsequently into knowledge. AI is a technology that can facilitate this transformation.
The foundation for adopting AI technology is availability of high-quality data. Mining sites gather vast amounts of data, some of which is structured and ready for AI deployment. One significant data stream originates from process control systems. Data historians collect, organize, and store process data, including key process variables, alarms, shutdown events, equipment running times and failures, production indicators, and utility consumption, etc.
Historian packages can further process data for visual interpretation such as trending curves, performance KPI and descriptive/predictive analysis. This facilitates root cause analysis, process optimizations, and equipment preventive maintenance, ultimately improving availability and performance while reducing operating costs. Some of Historian packages used with control systems are AVEVA PI , Honeywell Uniformance, GE Proficy, Foxboro EcoStruxure
Can we do more with the available data?
Pattern recognition
Ore variability poses significant challenges for concentrator plant recovery and grade control/optimization. Major parameters driving recovery, such as mill discharge particle size, flotation air and reagent flows are well understood. However, the overall patterns of recovery involving not only major, but majority of process parameters (measurements) are not apparent. AI is the technology that can be applied to historical data to uncover these patterns. These patterns are potential insights to recovery and grade optimizations translating to increase in mine site profitability.
Early in my career, I worked on a plant upgrade project. Operators mentioned a long-standing issue with sporadic loss of measurement signals in a part of the plant, with no apparent pattern. I inquired about the set point for heat tracing on instrument impulse lines and the freezing point of the process fluid. We concluded that the difference between these two values was too narrow, due to the nonlinearity (hysteresis) of temperature controls. The remedy was simple: increase the heat tracing set point to compensate for the nonlinear controls and maintain constant fluid viscosity, preventing the loss of measurement signals.
Operators were too close to the problem to see it, I was “new set of eyes”. AI applications can be a “new set of eyes” detecting patterns in data sets and increasing operability.
Data Aggregation
Finding the right information in a haystack of data stored in process systems, operating procedures, user manuals, design documentation, asset management systems is time consuming and potentially critical. Some level of data aggregation pulling relevant information from different sources is already available. Alarm management system contextualizes alarms. Predictive maintenance tools warn of potential failure. Operators would benefit from additional, on demand data aggregation tools triggered by major alarm / shutdown conditions or operator’s request. This aggregated data, once analyzed and presented to operators, would guide them in uncovering the root cause of events, assessing the condition of impacted equipment or devices, identifying potential hazards, following relevant health and safety procedures, checking the status and location of spare parts, and creating service requests, purchase order and event reports. AI tools can transform the information retrieval processes and report generation from manual operations to automated tools speeding up corrective measures / repairs and ultimately increasing plant availability.
I wish I had more data!
AI-facilitated image and sound analysis has paved the way for incorporation of new “field measurements” into a process controls. Traditionally, instrumented predictive maintenance has relied heavily on temperature and vibration monitoring. Equipment sound analysis can be a good addition to the monitoring system. Sound analysis can detect deviations in equipment performance, or it can serve as indicator of process operating conditions.
Similarly, video and image processing and analysis can be utilized to detect changes in equipment condition, monitor material flow, detect fires, detect presence of foreign material and warn/prevent personnel injuries.
Some of these use cases have been already implemented. However, process control systems can be enhanced with addition of these signals as inputs to their control philosophy and a code.
Environmental conditions such as wind, temperature precipitation are not normally inputs to process control system although much of ore crushing and ore transportation happens in outdoor conditions. These environmental conditions have an impact on the plant operation and personnel safety. Incorporating these conditions / data as input signals into process controls would offer additional context, enhancing visibility into how these variables impact plant performance.
AI will change process control systems
Current process control systems primarily use Boolean logic and feedback control loops as their algorithms. With advancements in AI, these systems will not only enhance their data analytical capabilities but also integrate AI-based algorithms to process controls. Edge devices capable of independent AI data processing are being developed. Additionally, the progress of humanoid robots in manufacturing is expected to extend to the mining industry. Over the coming years, these developments will bring significant changes to data collection / analysis and process control systems.
Nermina Harambasic, P. Eng, CDI.D is a project management and advisory expert with a degree in electrical / automation engineering currently focusing on AI technology for the mining industry. She is Founder of O-MOD and the host of “AI for Mining Industry” webinars.
