From two weeks to two seconds AI is revolutionising mining by boosting safety, cutting downtime and delivering smarter, faster operational decisions

Every hour of downtime costs a mine tens of thousands of dollars, and Professor Amir Gandomi told the NSW Resources Regulator’s Mechanical Engineering Safety Seminar how artificial intelligence is now cutting those losses by predicting failures and optimising operations in seconds.

For mining operators, unplanned equipment failures can cost up to $20,000 per hour in downtime. Gandomi pointed to studies showing that AI can reduce downtime by 50 percent, cut failures by 70 percent, and lower costs by 30 percent, with payback periods as short as three months.

“The real strength of AI is in early detection and prevention,” Gandomi explained. “By fusing data from multiple sources — sensors, images, logs, even text messages — we can predict unsafe conditions or structural weaknesses before they become critical.”

The approach is particularly relevant in mining, where mechanical failures or undetected hazards can carry serious safety consequences. AI’s ability to learn from large, heterogeneous datasets gives it a clear advantage over traditional methods of monitoring and inspection.

Scheduling Breakthroughs: From Days to Seconds

One of the strongest examples came from mining itself. A company was struggling with a complex scheduling problem that took up to 14 days to compute. That was unacceptable for daily operations, where optimisation was required overnight.

“We designed an algorithm that solved the problem with 2,000 variables in less than two seconds,” Gandomi said. “Scaling further, we proved it could solve problems with up to a billion variables in six days. What was once impossible to compute became routine.”

For mine planners, this shift in computational efficiency could be transformative — allowing daily optimisation of schedules, better resource allocation, and more agile responses to operational changes.

AI That Engineers Can Trust

A recurring barrier to AI adoption in engineering-heavy sectors like mining is the “black box” problem. Engineers need transparency and interpretability, not just outputs. Gandomi emphasised techniques such as genetic programming, which generate explicit equations rather than opaque code.

“Engineers like to see equations,” he noted. “Something they can interpret, use in ANSYS or Abaqus, or take a derivative from. Genetic programming gives you that.”

By producing models that can be validated and embedded directly into engineering software, AI becomes a tool that complements — rather than replaces — traditional expertise. This interpretability is likely to reassure practitioners who are cautious about relying on algorithms they cannot interrogate.

Optimisation Across the Mining Value Chain

Beyond scheduling, AI has broad applications across mining operations:

• Mine design: Gandomi and his collaborators have applied AI to open pit mine design problems, finding optimised layouts that account for cost, time, and risk.

• Supply chain: AI-driven optimisation has been used to streamline mining supply chains, reducing bottlenecks and improving throughput.

• Drilling shafts: One case involved optimising drilling shaft designs by embedding engineering constraints directly into the AI system. The only way the company achieved the right outcome was by using AI to explore options beyond the reach of traditional methods.

“Real-world engineering is multi-objective,” Gandomi said. “You’re balancing cost, time, reliability, and risk. AI-based optimisation allows you to consider all of these simultaneously.”

When to Use AI — and When Not To

Gandomi was careful to temper enthusiasm with practical advice. Not every problem requires AI.

“If you can solve your problem with mathematics, derivatives, or known equations — do that,” he advised. “AI should be reserved for complex, nonlinear, data-rich problems where traditional methods fall short.”

This pragmatic approach will resonate with mining engineers who are wary of adopting technology for its own sake. The message: use AI where it delivers clear value, and integrate it with existing engineering knowledge wherever possible.

Embedding Domain Knowledge into AI

A critical factor in success is combining AI’s computational power with engineers’ domain expertise. Even simple pieces of knowledge — such as knowing that one variable must always be greater than another — can dramatically shrink the search space and speed optimisation.

“By embedding engineering principles, rules of thumb, or constraints into the algorithm, you can boost performance significantly,” Gandomi said.

This interactive process of problem formulation, where engineers and data scientists work together to define objectives and constraints, is where the greatest value emerges. It ensures that AI is not operating in isolation but grounded in the realities of mining operations.

Learning from Other Industries

To underscore AI’s maturity, Gandomi referenced cross-industry successes. NASA used genetic programming to design an antenna that outperformed human designs. Boeing and BMW have long applied AI to aerospace and automotive engineering. Even during the COVID-19 pandemic, AI-based optimisation was used to model and balance health interventions against economic costs.

“These examples show AI is not experimental,” he said. “It’s proven in high-stakes environments. Mining can leverage those same capabilities.”

Data Explosion and the Mining Challenge

The challenge ahead is managing the sheer volume of data being generated. In 2025, humanity creates millions of queries, posts, and hours of video every minute. Mining operations, with their IoT sensors, geospatial data, and production logs, are part of this deluge.

“The first step is divide and conquer — break large problems into smaller ones and solve them efficiently,” Gandomi said. His team has pioneered approaches that tackle high-dimensional problems by reducing them into manageable dimensions, ensuring scalability for real-world datasets.

Practical Takeaways for Mining Professionals

Mining practitioners left Gandomi’s talk with clear, actionable insights:

• Start with predictive monitoring: AI tools can deliver immediate gains in downtime reduction and safety risk management.

• Optimise complex schedules: Replace slow, impractical computations with AI algorithms that deliver results in seconds.

• Insist on interpretability: Use AI methods that generate equations or models you can validate.

• Embed your expertise: Combine engineering principles with AI’s power for faster, more reliable outcomes.

• Focus on ROI: Look for applications with clear cost savings and quick payback periods.

Conclusion

For mining professionals, AI is no longer about futuristic possibilities. It is about solving practical, high-stakes problems with measurable returns. From predictive safety to near-instant scheduling, the applications outlined by Professor Amir Gandomi demonstrate that AI is already reshaping how mechanical engineering and safety challenges are tackled in the sector.

The challenge now is less about whether AI can deliver value, and more about how quickly mining companies can integrate it into their operations — safely, effectively, and with engineers at the centre of the process.