Motion Metrics News

The latest news and technology in mining.


July 27th, 2018

How to Prevent Dangerous and Costly Crusher Obstructions with AI

Last week, we shared an incident report about a jammed crusher in Western Australia.  Despite an exhaustive risk management procedure, this event saw an obstructing ground engaging tool (GET) ejected towards the operator in a manner that previous experience indicated was not likely.  This week, we explain why jammed crusher incidents are so dangerous and how tools like ShovelMetrics™ use artificial intelligence (AI) to decrease the likelihood of an obstruction occurring.

 

A Serious Situation

Jammed crusher incidents always present a serious safety issue for any mine due to the tremendous amount of stored kinetic energy.  While there are no comprehensive Canadian statistics, American researchers at the National Institute for Occupational Safety and Health found that incidents involving crushers are the second most common cause of fatalities caused by stationary machinery at mines in the United States.

The GET removal process requires extreme caution.  First, an excavator needs to remove all other material from the crusher.  Then, a welder uses a thermic lance or torch to heat and cut the obstruction to loosen it.  Because heat causes the metal to expand, the lancing process generates additional pressure that can cause the rapid, uncontrolled release of the 100-kilogram GET.

Clearing a crusher obstruction is a dangerous endeavour due to the tremendous amount of stored kinetic energy.

A Better Way to Approach Risk

The best way to manage the risk of an obstructed crusher is to avoid the situation entirely but, failing that, mines should aim to mitigate danger as early in the process as possible.  Thankfully, there is a solution based on artificial intelligence (AI) and deep learning (a subset of AI) that presents mines with an opportunity to prevent crusher obstructions before they happen.

Although oversized rocks and other bits of debris can also obstruct crushers, lost GET components are often the most severe, difficult to remove, and entirely preventable.  Our ShovelMetrics™ Missing Tooth Detection system, currently installed on more than 300 shovels around the world, uses AI to mitigate the risk of a broken GET component jamming a crusher.

Deep learning algorithms are capable of solving classification or prediction problems by making inferences from a dataset without human intervention1 – that is, they are able to learn and recognize complex patterns and relationships in datasets by performing millions of mathematical calculations and correcting their own mistakes.  These strengths make deep learning well-suited to mining applications, where large quantities of data are readily available.

ShovelMetrics™ helps mine personnel identify and locate missing teeth before they travel downstream and jam crushers.

 

Missing Tooth Detection with ShovelMetrics™

In the case of ShovelMetrics™, AI algorithms analyze many video frames to learn what a shovel bucket looks like and recognize when a tooth has gone missing.  Because the system continually collects video frames of the shovel bucket, the algorithm is trained to a very high level of accuracy.

The system uses a rugged camera that overlooks the shovel bucket and deep learning algorithms to continually monitor the status of the bucket teeth.  When a missing tooth is detected, the machine operator is alerted via an in-cab monitor so that the object can be intercepted before it reaches downstream crushers and conveyor belts.  The monitor specifies which tooth is missing and the time frame within which it broke so that the operator can determine which truck the missing tooth is in.

ShovelMetrics™ uses a rugged camera to continually monitor the shovel bucket.  Artificial intelligence-based algorithms detect when a tooth breaks or goes missing.

The AI Advantage

The advantage of using deep learning instead of traditional computer vision techniques to identify teeth is the flexibility that this approach provides.  Traditional computer vision systems require an engineer to hand-code the features that the algorithm should look for and any variations that it might encounter.  In contrast, deep learning systems simply learn by example; they do not require the input of calibration data or additional code to account for variation in different bucket and tooth configurations, lighting conditions, perspectives, etc. – only a wide enough variety of training data to account for all expected use cases.

For our purposes, this means that ShovelMetrics™ Missing Tooth Detection works for all shovel types without the need for special calibration or additional training.  It has been installed at mines and quarries worldwide and typically pays for itself within months.

Because the deep learning algorithms that power ShovelMetrics™ are able to account for different bucket and tooth configurations without additional code or special calibration, the solution can be installed on any shovel or excavator.

The Rise of AI

Despite its recent surge in popularity, early forms of AI have been around since the late 1950s3.  At that time, computational power was prohibitively expensive – analyzing vast amounts of data required brute-force computational power that simply wasn’t available outside of university labs.

Today, hardware is exponentially faster.  Specialized processors can perform millions of computations in parallel at lightning speeds – the latest chips from Intel can run more than 10 trillion computations per second2.   Furthermore, the advent of cloud computing means that these tasks can now be deployed to the cloud at a fraction of the cost.

With the global mining industry facing intense pressure to improve performance, media coverage of AI has focused heavily on its productivity benefits for mine applications.  However, the possibilities for enhancing safety are equally impressive and important.

ShovelMetrics™ notifies the shovel operator of a missing tooth via an in-cab monitor and audible alarm.  The solution also interfaces with MetricsManager™ Pro, our centralized data analysis platform.

Technology for a Safer Mine

Although the industry has made significant strides to improve working conditions, miners are still exposed to significant dangers.  Stricter safety regulations, improvements to personal protective equipment (PPE), and better education are all important steps on the path to reducing workplace injuries and fatalities, but we think incident mitigation makes the biggest leap.  To learn more about our safety-enhancing solutions, please contact us.