A fleet of strength

Release date: 4/12/2021

Mining Magazine - Machine learning, artificial intelligence and other newer technologies designed to make functions like fleet management more efficient are more firmly planted in the modern mine now more than ever before.  Mining, though, sometimes is slower to embrace change. Are we ready for these changes? How are technology’s earliest adopters faring? MM asked Dingo for their perspective. 

Dingo product engineering vice president Colin Donnelly says that the mining industry as a whole is hesitant to take the leap on integrating AI into its workflows for maintenance. 

“The mining industry is fractured in its approach to these technologies, and to new initiatives in general,” Donnelly says. “The early adopters have a highly innovative culture and workforce, whereas others ‘wait and see’ and tend to be more risk-averse…[wanting] the early adopters to prove the value before jumping in.” 

Dingo notes it has seen a sizeable push from C-suite executives to trial technologies that contribute to digital transformation, seeking to capitalise on a potential for greater efficiency, productivity and to reduce costs. At the same time, there has been some resistance to technology projects from operational resources.

Donnelly says he feels the reasoning behind the latter is an ever-increasing pressure on operations to increase production volumes and realise a high return on capital. 

“These mandates have made it difficult to get traction at site-level for newer, unproven technologies, due to the perceived risk and lack of bandwidth. Without some easing of these operational pressures, it will be difficult to convince operations teams to try new technologies, even if they promise extremely large benefits.” 

With those that have been considered early adopters of such technology, Dingo has seen many come to them to explore their respective options. 

“We have deployed our mobile Asset Health Manager™ app to help customers digitise inspections previously captured on paper,” Donnelly points out.

“A wide variety of applications such as structural frame inspections, equipment defect inspections, temperature data collection and fluid leak inspections are now being performed digitally. This is significantly improving efficiency and their ability to capture data, as well as allowing them to act much faster on the information collected. Recently, one of our customers did 700 inspections in a month using this technology.” 

Some of the company’s “more mature” clients, as it refers to them, have worked in tandem with Dingo to gain a better understanding of how AI and machine learning can benefit their business. At press time the company was working alongside several in the industry community on the advantages of machine learning, like developing models for remaining useful life (RUL) that can provide accurate predictions on component failures. 

Connelly says that it also has a number of sites across the world that have adopted the capabilities of predictive analytics into their businesses. He adds that many use predictive analytics to proactively identify and address issues on specific equipment fleets of equipment as well as within plant environments.
“Haul trucks have been a major pain point for our customers, and we have developed failure models for final drives, rod bearings, cylinder wear and fuel pumps,” he explains. One example, he adds, is the use of Dingo’s Anomaly Detection models for anomaly detection on data from its onboard systems for a haul truck fleet at a large gold miner.
“The Dingo Asset Health Process integrates predictive analytics into the wider maintenance workflow and then embeds the Dingo work recommendations directly into the ERP/CMMS.  This focus on integration into existing processes and systems helps ensure our customers can use this new information, while creating a seamless experience.” 

Those that have taken on the practice of predictive analytic modelling have developed confidence in the outputs they are getting, though Dingo does not have any sites in its customer portfolio at this time that have implemented full automation that would allow AI to make all of its decisions. Its adoption rate comes down to trust. 

“Dingo has a wide range of customers…[they] tend to be at different stages of maturity in both their maintenance practices and level of trust in new technologies, such as machine learning,” Connelly points out. 

Some customers are choosing to “dip their toes in the water” and test out accuracy and potential benefits for predictive analytics models. Others use machine learning to augment the decision-making being performed by human beings.
“Miners…are starting to delve into prescriptive analytics –the practice of using machine learning to classify data and make recommendations,” he notes, and as such it continues to broaden its work and capabilities with customers across the world.
“Our data science team is…working on a Virtual Condition Intelligence™ product – essentially a virtual assistant for reliability analysts – leveraging the millions of data points and analyst reviews we have in our system. The Virtual Intelligence™ product uses machine learning (AI) to identify faults in the condition monitoring data and then prescribe the best corrective action/s to resolve the issue.”
With machine learning, regardless of its end use, there is always the question of the amount of ‘learning’ that goes into these systems. It obviously is not a one-size-fits-all technology, and as Connelly notes, the differences can be seen in the models and also in the planned applications.
“The amount of learning that goes into the models varies extensively from one model to another,” he points out, adding that the data type also has a significant impact on the amount of training that is or will be needed.
“In our experience, the quality, context and applicability of the data determine the amount of data needed to successfully train a model. Some models need only a few thousand data points and can predict or classify with reasonably high accuracy, whereas other models are not very accurate with millions of data points due to low data quality and/or relevance.” 

Dingo’s new Virtual Condition Intelligence solution, for which the company has not released an unveiling timeline, will involve the constant review of outputs by subject matter experts. This supervised learning, it says, will enhance the models’ overall accuracy.
The company’s popular Trakka predictive analytics models are also in the process of being broadened thanks to significant feedback from its customers. Its goal is to address an even broader range of problems on a wider variety of equipment using the system.
“Dingo is also advancing its AI capabilities in other areas and one of the most interesting is the deployment of AI models to the edge or in the field. Dingo is working on several use cases where AI models are deployed onto devices that can be disconnected from the network and these models can undertake prediction and classification without a round-trip to the cloud,” Donnelly explains. 

“For the inspection app, this could mean that the inspector will get immediate feedback on the data/images they have collected and potentially what the next course of action should be in real-time. We believe there is a significant opportunity to help mining operations improve their knowledge and decision-making in the field, and Dingo is investing heavily in this area.” 

For more information on Dingo's industry leading predictive maintenance solutions, email us at or use our contact form

Donna Schmidt  │Mining Magazine │April 2021