When the 14th International Comminution Symposium, Comminution ’25, gathered in Cape Town last year the mood was unmistakably one of transition. The fundamentals of crushing and grinding had, of course, not changed. Rock still has to be broken, energy is still required to break it, and the perennial challenges of wear, throughput, liberation and classification remained firmly on the agenda.
But there was a noticeable shift in the questions being asked. How can we make comminution more selective rather than simply finer? How can sensors tell us what is happening inside a mill before the operator can see the consequences? Can artificial intelligence replace some of the expensive and time-consuming conventional characterisation work? And perhaps most importantly, how do we design a circuit around the whole value chain rather than optimising individual pieces of equipment?
As we look ahead to Comminution ’27, these questions are likely to become even more prominent. So what might have changed between the presentations at Comminution ’25 and Comminution ’27?
Artificial intelligence was already impossible to ignore at Comminution ’25. The opening keynote examined how rapidly developing AI techniques were accelerating comminution modelling, while other presentations explored machine learning for mill optimisation, AI-assisted calibration, soft sensors and model-predictive control. One particularly practical example was work using machine learning to predict standard comminution parameters from Geopyörä breakage-test data.
At Comminution ’25, AI was often presented as a promising addition to existing modelling and control methods. By Comminution ’27, we may see much more emphasis on AI embedded in the workflow. That could mean AI-assisted ore characterisation, rapid geometallurgical variability assessment, real-time soft sensors, predictive control and automated optimisation.
Another strong thread from Comminution ’25 was the rapid development of indirect ways of seeing what is happening inside equipment. Acoustic monitoring was one example. Research presented at the conference investigated how changes in AG/SAG mill feed size produced distinctive acoustic responses. Similarly, the conference included work on instrumentation and analytics for cone crushers, including the integration of noise, vibration, high-frequency monitoring and vision systems for fault detection and condition monitoring. Since the conference, the same direction has continued into fine grinding. This is significant, as for decades operators have had to infer what is happening inside a mill from measurements taken around it: power draw, pressure, density, flowrate, product size and so on. The next generation of systems may make the equipment itself an increasingly sophisticated sensor. At Comminution ’27, expect more work combining acoustic, vibration, power, pressure, vision and process data and turning those signals into information that operators can actually use.
Perhaps the most important evolution since Comminution ’25 is occurring at flowsheet level. There were already strong signs of this in Cape Town. One study presented a framework for integrated optimisation of crushing, milling and cyclone classification, explicitly considering throughput and product quality together. Another proposed a “Parallel Flow Sheet” incorporating HPGR, stirred milling and classification.
At the same time, HPGR was being considered not simply as a more energy-efficient piece of equipment, but as part of fundamentally different circuit architectures. This is a crucial distinction. For many years, comminution innovation could be discussed in terms of better crushers, better mills, better liners or better grinding media. Increasingly, the question is becoming: what is the best sequence of breakage, classification and separation operations for the ore?
That opens the door to circuits that might look very different from the conventional crushing–SAG–ball mill–flotation arrangement. HPGR, stirred mills, dry classification, coarse particle recovery and sensor-based sorting can all become parts of the same optimisation problem. Comminution ’27 could therefore be considerably more about flowsheet architecture than equipment optimisation in isolation.
“How fine do we really need to grind?” This may ultimately be one of the biggest questions of the next decade. The conventional response to poor liberation has often been to grind finer. But finer grinding comes with a price: energy, media consumption, wear, water consumption and, potentially, the generation of fines that are difficult to recover. Comminution ’25 highlighted this tension repeatedly. There was discussion of the balance between improved liberation and fines generation, dry grinding, coarse gangue rejection, HPGR-based circuits, sensor-based sorting and alternative comminution technologies. The implication is that the optimum comminution product may not be the finest product. It may be the product that gives the best combination of liberation, particle size, shape and downstream recoverability.
That moves comminution much closer to the heart of the entire mineral-processing value proposition. This leads naturally to another likely feature of Comminution ’27: increasing interest in selectivity.
Traditional comminution is largely indiscriminate. The objective is to reduce particle size, while liberation is a consequence of the breakage process. But mineralogical information increasingly allows us to ask whether rocks can be broken preferentially along mineral boundaries, or whether valuable minerals can be liberated without unnecessarily grinding barren gangue. Comminution ’25 contained several examples of this philosophy, from high-voltage pulse power and HPGR to GRolls and mineralogical investigations of breakage.
The subsequent literature is moving in the same direction. Work on microstructure and breakage is providing a more detailed understanding of how mineral texture affects fragmentation, while research into slow-compression breakage is examining the transition between single-particle and confined-particle-bed breakage. The big prize is not simply lower specific energy. It is lower specific energy while producing a more useful particle population. That means asking what happens to the particles after they leave the crusher or mill.
One of the most interesting developments at Comminution ’25 was the growing recognition that the grinding environment can affect downstream mineral processing, including subsequent collector adsorption and flotation behaviour. This is another sign that the old boundaries between unit operations are becoming less useful. The comminution engineer cannot simply specify a P80 and hand the product to the flotation engineer. Particle shape, surface chemistry, mineral exposure, fines generation and fracture characteristics can all influence what happens next. By 2027, we may therefore see more papers evaluating comminution performance using downstream recovery and grade, rather than energy consumption and particle size alone. That would be a welcome development. After all, the purpose of grinding is not to make small particles, it is to make valuable minerals recoverable.
Energy efficiency has been a central theme of comminution research for decades, but the sustainability discussion is becoming broader. Water availability is increasingly important, particularly in major mining regions such as Australia, Chile and Peru. Comminution ’25 included several examples of dry grinding, dry classification and flowsheets designed to reduce water consumption. Dry VRM technology, for example, was presented as offering substantial energy savings relative to traditional mill systems, while dry HPGR-based flowsheets were being combined with classification and beneficiation concepts. This suggests that Comminution ’27 may increasingly evaluate technologies against several simultaneous metrics. In other words, specific energy will remain important, but it will no longer be enough.
Comminution ’25 also contained a healthy dose of scepticism. Grant Ballantyne's keynote asked how metallurgists can distinguish the “fine wine from the snake oil” when assessing new comminution and classification technologies. A lower energy process may have higher wear; improved liberation may not translate into improved recovery; laboratory results may not survive scale-up. That may be an even more important conversation in 2027.
The industry has no shortage of exciting technologies. High-voltage pulse power, novel compression devices, advanced stirred mills, AI, sensor-based sorting, digital twins, novel liners, new grinding media and radically different circuit configurations all offer intriguing possibilities. So, what might Comminution ’27 look like?
At Comminution ’25 we saw the beginnings of a move from understanding comminution towards controlling comminution. We saw AI beginning to enter the mainstream, sensors beginning to reveal what happens inside opaque equipment, and alternative technologies challenging the dominance of conventional SAG and ball milling. Two years later, the test will be whether those ideas have survived contact with operating plants.
That is what could make Comminution ’27 particularly interesting. Not simply what is the latest technology? But which of the technologies we were excited about in 2025 have actually delivered? And perhaps, most importantly, what new ideas will have emerged in the meantime that none of us saw coming? One thing seems certain: if Comminution ’25 was about finding new ways to break rock, Comminution ’27 may be increasingly about knowing why, where, when and how much to break it and making the entire mineral-processing circuit respond accordingly.
.jpg)
