Dingo says its new Trakka Predictive Analytics solution uses artificial intelligence and machine learning to predict impending equipment failures with confidence, allowing customers to proactively perform corrective maintenance actions to minimise downtime and optimise asset life.

The release comes around five months since the company laid the groundwork for the new solution with an announcement that it would introduce practical machine learning models built using real customer data and targeted at specific industry problems from January.

The new Trakka solution includes a series of sophisticated predictive analytics models to provide anomaly detection and failure prediction for asset intensive industries, the company said. These models are built by uniting failure data from actual equipment, “Dingo’s industry expertise and data science to address common component-specific failure modes, such as final drive gear teeth wear”.

Powered by a proprietary machine-learning library, the Trakka Predictive Analytics solution can, Dingo says, predict the time until asset/component failure with a high degree of accuracy. The company said its customers will reap the benefits of these remaining useful life (RUL) models (pictured) as they:

• Reduce unexpected failures and downtime;
• Reduce repair cost as scheduling is optimised;
• Reduce loss of wasted potential in capital;
• Reduce unnecessary maintenance activities;
• Reduce personnel and process risk by creating a safer and more controlled environment;
• Improve component life by acting earlier;
• Improve confidence in planning component replacements;
• Improve equipment availability and reliability;
• Improve budgeting and the bottom line, and;
• Improve business related processes such as procurement, logistics and management.

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

“This process typically involves data collecting, cleansing and validation to ensure model outputs are as accurate as possible. The transition to online predictive analytics is complete once the data ingestion pipeline is ready and the models are fully trained and tested.”

The predictive models are designed with scalability in mind, Dingo said, meaning they can be easily re-trained to work with a broad range of asset and failure mode problems experienced by real mining operations, making them highly reusable.

“The models are continuously optimised through ongoing validation and the input of new data and equipment performance information,” Dingo said.

And, the platform connects a broad range of systems and software to provide data surrounding asset health, including enterprise resource planning & enterprise asset management systems, computerised maintenance management systems, fleet management systems and all forms of condition monitoring data, including oil analysis, visual inspections, sensor data, vibration and thermography.