However, as good as this system sounds, it is being seen as merely a step down the path to what its creator DINGO believes is truly possible.

DINGO has developed a series of predictive analytics models with its Trakka Predictive Analytics package that provide anomaly detection and failure prediction for miners’ assets.

However, DINGO director of product engineering Colin Donnelly told Australia’s Mining Monthly that predictive analytics was only the first step.

He said prescriptive analytics was the Holy Grail its customers were seeking.

The tried and tested models driving Trakka Predictive Analytics have been built by uniting failure data from equipment with data science and DINGO’s industry expertise.

The company worked with the Queensland University of Technology’s data scientists to develop its predictive analytics package.

Indeed, one of the university’s PhD students has since become a full-time DINGO employee.

Donnelly said the university brought the algorithms and some of the data analytics smarts while DINGO added its real world experience.

He said the product had been about two years in the making.

Interestingly the final six to nine months of that were spent working on the commercial model.

That is one of the challenges facing mining equipment, technology and service companies, particularly those trying to create innovative solutions to their customers’ problems.

The mining companies are often happy to receive those solutions, just not so happy to pay for them until they have been proven.

The DINGO Trakka Predictive Analytics solution uses a proprietary machine learning library to predict the time until asset or component failure with a high degree of accuracy.

This will let customers benefit from remaining useful life models. Among other things they reduce unexpected failures and downtime, cut repair costs because scheduling is optimised, improve equipment availability and reliability, and boost component life through acting on problems earlier.

Before any predictions can be made DINGO domain experts and its data science team have to work with a customer’s historical failure and condition monitoring data to deploy or adapt existing models or create machine learning models to correctly identify failures within the customer’s fleet.

This usually involves collecting, cleansing and validating data to ensure model outputs are as accurate as possible.

The transition to predictive analytics is complete once the data ingestion pipeline is ready and the models are fully “trained” and tested.

DINGO’s models are designed with scalability in mind and can be retrained to work with a broad range of asset and failure mode problems mining operations experience.

The models are optimised through ongoing validation and the input of fresh data and equipment performance information.

“We’re trying to take reactive maintenance and turn it into predictive maintenance,” Donnelly said.

“As the market starts building confidence in it they may be able to start moving from scheduled maintenance to predictive maintenance.”
That way an asset does not have to be pulled out of service just because the manual says a part needs to be replaced. Rather, a decision could be made on that maintenance based on what the data says about the component.

In some ways what DINGO is offering is not too dissimilar to what other predictive analytics models do. It takes the information into its models and spits out a number.

Donnelly said it was knowing what to do with that number that set companies such as DINGO apart.

“The way I look at how we do it slightly differently is that we are already plugged into the customers,” he said.

“We’re already plugged into their enterprise resource system.

“We can turn that number [that the model generates] into the next two steps you have to take.”

Moving to prescriptive analytics will involve taking the experience DINGO’s experts have built up with its customers and “teaching” it to the system.

That way when it spits out the number it will also be able to suggest what has to be done next to rectify the problem or extend the asset’s life.

For more information on DINGO’s industry leading predictive maintenance solutions, email us at info@newsitetest.dingo.com or use our contact form. 

Noel Dyson │ Australia’s Mining Monthly │ May 2019