PdM for Mineral Processing Plants
Mineral processing plants are equipped with a high number of fixed assets that perform essential transformations in the ore feed. Equipment downtime directly results in production losses and high costs. Although most of the processing units are well instrumented and connected, only a few miners are effectively using this data to improve their maintenance performance.
In 2018, the Barrick Gold Cortez Mine, in Nevada, deployed a PdM project in the gold refining process applying ML and equipment sensors in order to identify potential issues before they escalated into failure. The algorithm generates a healthy asset score using collected sensor data and tracks relevant
FIGURE 7.8
DT capabilities in all phases of asset life.
variations that might indicate undesired failures. After implementation, the investment paid off in a short period since several major failures were detected and avoided. For example, one of the fault predictions alone saved $600,000 [17,18].
At Newcrest Mining's Lihir operation in Papua New Guinea, ML algorithms were applied to reduce unplanned breakdowns in its semi-autogenous grinding mills. The project was developed in partnership with the technological solutions company Petra Data Science, which also provides digital twin solutions for mining operations. Taking data going back one year, the team focused on the critical signals or indicators associated with overload events. This big data included noise, power, speed, energy consumption, and control variables, most of them recorded at five-second intervals. The model was able to find patterns and anomalies in the data and use them to predict the probability of an overload event one hour in advance, enabling operators and engineers to proactively take the required actions to avoid it [19].
In a partnership between the industrial AI software company Uptake and the world's largest copper producer, Codelco, an AI solution is implemented to monitor the health processing plant (and mining) equipment through an enterprise-wide Asset Performance Management solution across all of the mining sites. The project's approach will generate industrial data science original equipment manufacturer-agnostic insights, predictions, and prescriptions for any asset [20,15].
Belt transport systems play an important role in the mine's ore transport system. In order to meet operations requirements, an online fault detection system is essential. In [21], multivariate analytical models based on data fusion and artificial intelligence techniques were implemented to avoid belt conveyor failures, supporting planned repairs, and reduce repair costs and production losses associated with breakdowns.
In [15], a data science team developed an anomaly detection algorithm that was able to determine the indicators of failure in a Ball Mill drive gearbox and pinion bearings. Although there was a large amount of condition monitoring data available (vibration, pressure, and temperature sensors), only a small number of failure records were available. The challenge was to isolate the different failure mechanisms (no failure classifications provided) and develop a predictive model that would allow component repair/replacement to be scheduled during quarterly plant shutdowns. The developed model was tested against previous failure data to determine the appropriate confidence level of the model. Back testing results show 85% accuracy in predictions, correctly identifying past failure events separately to repair events.
