In this exclusive interview with DailyCADCAM, Dr. Rob Ward, CEO & Co-Founder of DigitalCNC, shares how physics-based machining intelligence is transforming CNC manufacturing by bridging the gap between CAM-generated toolpaths and real machine performance. He discusses machine-specific optimization, accurate cycle time prediction, the role of AI in manufacturing, and how DigitalCNC is helping manufacturers improve productivity, reduce prove-out time, and achieve greater machining efficiency across industries.
1. Congratulations on the remarkable journey of DigitalCNC. Could you begin by sharing your professional background and what inspired you to establish DigitalCNC? DigitalCNC originated from research conducted at the AMRC and the University of Sheffield. How did years of academic research evolve into a commercial software platform?
Thank you. My route into this was not a conventional one. I began my career in the Royal Navy, which is where I learned to trust engineering fundamentals when the stakes are high and there is no room for guesswork. From there I moved into advanced manufacturing research, spending around a decade at the University of Sheffield AMRC and completing an engineering doctorate in 5-axis machining. I specialised in CNC interpolation, the discipline that sits between the G-code and what actually happens on the machine’s feed drives.
What inspired DigitalCNC was a problem I kept seeing on the shop floor while working on demanding aerospace programmes. We would generate a toolpath in CAM that looked perfect on screen, and then the machine would behave quite differently once cutting began: cycle times were wrong, feedrates collapsed in the corners, and engineers spent days proving programmes out on expensive machines. That gap between the digital plan and the physical reality was costing the industry enormous amounts of time and money.
The transition from research to product came from a simple realisation: this behaviour is not random, it is predictable. If you model the machine, its controller and its dynamics from first principles, you can forecast what will actually happen before you cut. Our entire reason for existing is to turn years of that research into something a CAM programmer can use in a few clicks, inside the tools they already know.
That is why the validation matters as much as the science. DigitalCNC already runs natively inside the platforms programmers trust, including Siemens NX, where we hold approved Technology Partner status, alongside CATIA V5 and Mastercam. And the timing could hardly be sharper. As aerospace and defence ramp up production and bring more work back onshore, the pressure to get parts right first time, and to squeeze more from every machine already on the floor, has never been greater. The cost of ignoring the gap between plan and reality is rising, not falling.
2. Many CAM users assume that once a toolpath is generated, the machine will follow it exactly. In reality, this isn’t always the case. Could you explain this “CAM-CNC gap” in simple terms?
It is a very common assumption, and an understandable one. The CAM system produces a toolpath with commanded feedrates, it looks precise, and so people naturally expect the machine to follow it to the letter. In practice, the machine almost never achieves those commanded numbers. Put simply, the toolpath is a promise, and the machine decides whether to keep it.
The reason is that a CNC machine is a physical system with real limits. Its controller has to look ahead, slow down for corners and direction changes, respect acceleration and deceleration limits, and honour the machining tolerance you have set. So the feedrate the machine actually achieves along the path, what we call the achieved feedrate, is usually lower and constantly varying. The path is followed closely, but the timing and the dynamic behaviour are different from what CAM assumed.
A simple way to picture it is a racing car on a circuit. The car has a top speed on the straights, but it cannot carry that speed through the corners: it has to brake on the way in, slow through the apex, then accelerate out again. So the lap time is never simply the length of the track divided by the top speed, it is governed by how much the car has to slow for every bend. A CNC machine does exactly the same thing along a toolpath, braking for corners and changes of direction rather than holding the commanded feedrate throughout. And just as two cars with different brakes, grip and handling will post very different lap times on the same circuit, two machines will run the same toolpath in very different times, because each slows down in its own way.
3. DigitalCNC describes itself as a provider of “Physics-Based Machining Intelligence.” What exactly does this mean, and how is it different from conventional machining simulation?
Conventional machining simulation is largely about geometry and safety. It answers questions like: will the tool collide with the fixture, is the material being removed correctly, does the finished shape match the model. That is genuinely valuable, but it tells you very little about how a specific machine will actually perform when it executes the program.
Physics-based machining intelligence answers the performance question. If simulation tells you whether a part is safe to cut, we tell you how your machine will actually cut it. We use first-principles mathematics and control-systems modelling to predict how a particular machine, with its own controller, drives and kinematics, will really run the toolpath before any metal is cut. Crucially, this is deterministic. It is calculated from the physics, not learned from a statistical model or trained on large datasets, which means the answers are explainable and repeatable, something that matters enormously in regulated industries like aerospace.
What makes the approach distinctive is that, other than a full virtual NC kernel, it is the only software that takes machining tolerance, machining type and the actual toolpath geometry into account together, including complex 5-axis, dynamic and trochoidal toolpaths. Unlike a virtual kernel, though, it needs no specialist controller setup or training, it is effectively instantaneous, and a programmer can run it in around five clicks inside their existing CAM environment.
4. One of the biggest challenges manufacturers face is inaccurate cycle time estimation. How does DigitalCNC predict real machine cycle times more accurately than traditional CAM software?
Traditional CAM tends to estimate cycle time with a fairly simple calculation: distance divided by the commanded feedrate, with some allowances. Because it assumes the machine hits the programmed feedrates, it consistently underestimates real cycle time, sometimes badly, particularly on complex parts with lots of small moves and tight tolerances. Distance divided by commanded feedrate is really a guess dressed up as a number.
DigitalCNC predicts the achieved feedrate along the entire toolpath rather than the commanded one. To do that, it models the machine-specific behaviours that actually govern speed: the controller’s look-ahead, its acceleration and deceleration limits, how it rounds and blends corners, how it responds to the tolerance band you have set, and, for multi-axis work, the kinematics of the machine as it repositions the rotary axes. By resolving all of that from the physics, we can calculate a realistic time for that specific machine, not a generic estimate.
The practical result is cycle time predictions that land within around five percent of reality, which is a step change from what programmers are used to on complex parts. That accuracy flows straight into better quoting, more reliable scheduling and far fewer surprises on the shop floor.
5. DigitalCNC promises machine-specific optimization rather than generic optimization. Why is machine-specific intelligence becoming increasingly important in modern manufacturing?
The core point is that no two machines behave the same way, even when they are running an identical toolpath. A Heidenhain-controlled machine, a Fanuc-controlled one and a Mazak will each accelerate, decelerate and handle corners differently, and their kinematics differ too. Generic optimisation ignores that. At best it leaves performance on the table; at worst it optimises for behaviour the machine does not actually exhibit. No two machines run the same toolpath the same way, so it makes little sense to program them as though they do.
This matters more every year for two reasons. First, manufacturers increasingly run mixed fleets from several builders, so a single generic assumption simply does not fit. Second, as tolerances tighten and materials get more difficult, the margin for error shrinks, and small differences in machine behaviour have real consequences for time and quality.
There is also a bigger picture. In today’s climate, aerospace and defence demand is climbing, and manufacturers are under real pressure to deliver more parts, faster, from the machines they already own, often while bringing production back onshore. When capacity is that tight, software that tells you how each machine will genuinely perform stops being a nice-to-have and becomes part of the critical path. It is one of the quiet levers that decides whether a programme ships on time.
Machine-specific intelligence lets you design the toolpath for the machine that will actually cut it, so you get the best genuine performance out of that asset. It also lets you make smarter decisions about which machine to put a job on in the first place, which is a powerful lever when you are trying to maximise the output of an expensive shop floor.
6. Aerospace manufacturers demand extremely tight tolerances while minimizing prove-out time. How does DigitalCNC help manufacturers reduce first-off failures and expensive machine trials?
Aerospace is an unforgiving environment. You are often machining expensive titanium or nickel superalloy components on high-value machines, with long cycle times and tolerances measured in microns. When a new programme goes to the machine for prove-out, every iteration costs money in machine time, engineering hours and, if something goes wrong, scrapped material that may be worth tens of thousands of pounds.
DigitalCNC attacks that cost by moving the discovery of problems off the machine and into the CAM environment. Because we predict the machine-specific behaviour before cutting, a programmer can see where feedrates will collapse, where the chosen tolerance is forcing the machine to crawl, and where the dynamics are likely to cause trouble, all virtually. They can then correct the program before it ever reaches the spindle.
The effect is fewer first-off failures, less scrap and a meaningful reduction in the number of physical trials needed to sign a part off. The cheapest place to fix a part, after all, is before you have cut a single chip. In an industry where prove-out is often treated as an unavoidable cost of doing business, getting much closer to right first time is genuinely valuable.
7. Your platform claims to reduce cycle time prediction errors to within approximately ±5%. Could you share a real customer success story in short that demonstrates this capability?
The plus or minus five percent figure refers to how closely our predicted cycle time matches what the machine actually delivers, and it holds up on genuinely complex parts, not just simple test pieces. A prediction you can trust is a prediction you can quote against, schedule around and commit to a customer.
A good example is a project we ran with Tier 1 aerospace company in the UK on a complex structural component. The team was weighing up two competing machining strategies and had to commit to one, but they had no reliable way to test them unless through trials. Their CAM system predicted a cycle of around forty seconds, while the machine actually took three minutes, a discrepancy of roughly four and a half times. On machines of that size, proving the options out physically runs to more than five thousand pounds a day, so making that call on guesswork was both expensive and risky.
We ran a kinematic simulation of both strategies on their exact machine, entirely inside CAM. That exposed the precise toolpath segments where the feedrate would collapse, flagged the high-risk zones, and let the two options be compared side by side on realistic cycle times. The prediction landed within about a second of the measured machine output, comfortably inside the plus or minus five percent I mentioned, and the problem areas were corrected before anything was cut. The decision was made in under two hours instead of days, and it consumed no machine time at all, so for the strategy-selection phase the prove-out cost was effectively eliminated. The choice was settled at the desk, not on the machine.
8. Manufacturing companies increasingly operate multiple CNC machine brands. Can DigitalCNC compare machining performance across different machines before production begins?
Yes, and it is one of the capabilities customers find most useful. In effect, you can hold a race between your machines before committing a single one to the job. Because our predictions are built from machine-specific characterisation, we can take the same part and toolpath and run it against several different machine profiles to compare how each would perform, before anything is committed to production.
That means you can see, in advance, the predicted cycle time and behaviour of a job on, say, a particular five-axis machining centre versus another builder’s equivalent, and understand where each will be faster or run into difficulty. It removes a lot of the guesswork from a decision that is normally made on intuition, or on whichever machine happens to be free.
Practically, this supports much better machine selection, more accurate quoting, smarter capacity planning and even capital investment decisions, because you can evaluate how a candidate machine would handle your real workload before you buy it. You are comparing on predicted physical performance, not on spec sheets.
9. Which industries currently benefit the most from DigitalCNC? Beyond aerospace and defense, do you see opportunities in automotive, medical devices, precision engineering, energy, or mold and die manufacturing?
Aerospace and defence are our natural starting point, both because the pain is most acute there and because it is the world I come from, but the underlying problem is universal. Anywhere that cycle time, tolerance and machine-specific behaviour affect cost or quality, the technology applies.
Naturally we fit right across precision engineering, in energy for components like turbine parts, and very much in mould and die work. Moulds and dies involve long cycles and complex three-dimensional surfaces where the interaction between toolpath geometry and machine dynamics is exactly what we model, so the accuracy gains can be substantial. In short, the market is as broad as precision machining itself.
The medical sector is another key area for us, given the combination of tight precision, difficult materials and small, high-mix batches where getting it right first time really counts.
10. Artificial Intelligence is transforming every engineering discipline. How does AI complement DigitalCNC’s physics-based approach, and where do you see the balance between AI and engineering science?
This is a subject I feel strongly about. There is a clean distinction at the heart of it: a learned model infers from what has already been cut, whereas our physics calculates what will happen next. Our prediction engine is deterministic physics, and it will stay that way for now, because in regulated, safety-critical industries you need answers that are explainable, repeatable and certifiable. A model that gives a different answer each time, and cannot tell you why, is a difficult thing to certify a flight-critical part against. In a safety-critical world, you cannot certify a black box.
Where AI adds value is around that engine, not inside it. The AI handles workflows, helps with decision support, and learns an individual programmer’s preferences and style, but the physics remains the source of truth for prediction. The knowledge about how engineers make decisions feeds assistive layers only, never the calculation.
So the balance I aim for is simple: physics for the physics, and AI for the human workflow around it. That way you get the productivity and ease of use that modern AI enables, without giving up the rigour and trust that engineering demands. The two are complementary, and keeping them in their proper roles is, I think, the responsible way to bring AI into manufacturing.
11. What role do real machine data and feedback play in continuously improving DigitalCNC’s predictive capabilities?
I want to be precise here, because there is a common assumption that a tool like ours must be continuously ingesting live data from the machine controller and learning from it. DigitalCNC does not work that way, and deliberately so.
Our engine is grounded in first-principles physics, and validated against machines across a wide range of controllers including Siemens, Heidenhain, Fanuc and Mazak. That characterisation tells us how a given machine is set up to behave, and the physics does the rest. We then validate the predictions against real cutting trials to confirm the accuracy holds up in practice.
The important distinction is that this is validation, not training a statistical model on streams of live data. It is precisely what allows us to be deterministic and explainable. It also means we do not need a machine to be instrumented or connected, and we do not need vast historical datasets before we can give a useful answer. As we characterise more machines and controllers, our coverage broadens and our validation deepens, but the predictive core remains the physics, which is exactly what makes it trustworthy.
It is also a misconception that most machine shops are running the latest machines and controllers. In reality, we are still a long way from high-speed data connectivity on the shop floor. It is expensive, it demands specialist skills, and it lands in a world where budgets are squeezed, experienced people are retiring, and the incoming workforce is greener than ever. In an ideal world, every machine would be connected to integrated CAM and inspection systems, and there are good companies working hard towards exactly that. But we are not there yet, so our philosophy is straightforward: add value today, and support the connected vision for tomorrow.
12. Your software integrates with leading CAM platforms. Could you tell our readers about the current integrations and your future roadmap for supporting additional CAD/CAM ecosystems?
As I mentioned earlier, we meet programmers where they already work, so DigitalCNC runs natively inside the CAM environment rather than asking anyone to adopt yet another separate tool. That principle is why we have invested in deep, approved integrations rather than bolt-ons: Siemens NX, where we hold Technology Partner status, along with CATIA V5 and Mastercam.
On the roadmap there are two directions of travel. The first is depth: broadening the range of controllers and machine types we characterise, and, importantly, extending our capability from three-axis, which is deployed today, into full five-axis, which is committed for later this year, with mill-turn to follow. The second is breadth: continuing to expand the CAM and CAD/CAM ecosystems we integrate with, so that more of the industry can reach the technology inside its existing workflow.
The guiding principle throughout is that adopting DigitalCNC should never mean changing how you already program. It should simply add machine-specific intelligence to the environment you know.
13. How easy is it for an existing CAM programmer to adopt DigitalCNC without significantly changing their current workflow?
Ease of adoption was a design goal from the very beginning, because I have seen how many good technologies fail on the shop floor simply because they ask engineers to change how they work. DigitalCNC is deliberately the opposite.
It runs inside the CAM system the programmer already uses. It requires no specialist CNC controller training, which is unusual for this kind of capability, and a programmer can get a machine-specific prediction in around five clicks. Because it is effectively instantaneous, it fits naturally into the normal rhythm of programming rather than interrupting it.
So the honest answer is that it is very easy. We are adding insight to what people already do well, not asking them to abandon their workflow or retrain. We add intelligence to your process; we do not ask you to change it. A skilled CAM programmer can be getting value from it within minutes, not weeks.
14. What are the biggest misconceptions manufacturers still have about machining optimization and CNC simulation?
There are a few that come up again and again. The first is the belief that the machine will faithfully reproduce the feedrates programmed in CAM. As we have discussed, it rarely does, and until you accept that the CAM-CNC gap exists, you cannot manage it.
The second is assuming that collision checking and material-removal simulation tell you about real machine performance. They are important for safety and geometry, but they say almost nothing about achieved feedrates, dynamics or true cycle time.
A third is the idea that optimisation is generic, that a toolpath tuned in general is tuned for your machine. It is not, because machine behaviour is specific to the controller, drives and kinematics in front of you.
And a fourth, which is becoming more common, is the assumption that predicting machine behaviour requires a connected machine and a data-hungry AI model trained on huge datasets. In reality, a first-principles physics approach is more robust, more explainable, and does not depend on collecting years of data first. You do not need a data lake to predict how a machine behaves; you need the physics. Clearing up these misconceptions usually changes how people think about their whole prove-out process.
15. As machining becomes increasingly automated, what skills should the next generation of CAM programmers and manufacturing engineers focus on developing?
As more of the routine work becomes automated, the value shifts decisively towards engineers who understand why things happen, not just which buttons to press. So my first piece of advice is to invest in the fundamentals: the physics of cutting, machine and controller behaviour, and the real relationship between a toolpath and the machine that runs it. That understanding is what lets you judge whether an automated result is actually sensible.
Alongside that, I would encourage genuine digital and data literacy, and the confidence to work alongside AI tools critically rather than either fearing them or trusting them blindly. Systems thinking matters too, because modern manufacturing is an interconnected digital thread, not a set of isolated steps.
Above all, stay curious and keep learning. The engineers who will thrive are those who combine solid hands-on machining understanding with computational skills, and who see automation and AI as tools that amplify their judgement rather than replace it. Automation keeps changing which buttons you press; it never changes the value of understanding why. The human who understands the fundamentals will always be the one making the best decisions.
16. Finally, what message would you like to share with the readers of DailyCADCAM.com, especially CAD/CAM professionals, manufacturing engineers, and decision-makers across India and around the world?
My main message is one of optimism grounded in engineering reality. Manufacturing is entering a period where predictive, physics-based intelligence and the wider digital thread can transform productivity, and that is a genuinely exciting place to be. But the fundamentals still matter, and the gap between the digital plan and the physical machine is real, costly and, importantly, solvable.
To the readers of DailyCADCAM, and especially to the many talented CAD/CAM professionals, manufacturing engineers and decision-makers across India and around the world, I would say this: do not accept the CAM-CNC gap as an unavoidable cost of doing business. Seek out tools that give you machine-specific, explainable insight before you cut, and back the engineers who understand the science underneath the software. India in particular has a fast-growing, ambitious precision manufacturing base, and technologies like this can help it compete at the very highest level.
The future I believe in is one where deterministic engineering science and intelligent software work together, with skilled people firmly at the centre. My philosophy fits in a single line: trust the physics, empower the people, and close the gap before you cut. If that is a future you want to help build, I would love to hear from you.
We thank Dr. Rob Ward for sharing his valuable insights with the readers of DailyCADCAM. His perspectives on machine-specific machining intelligence, predictive manufacturing, and the future of AI-powered CAM highlight how advanced software and engineering science can work together to improve productivity, reduce costs, and enable smarter manufacturing. We look forward to following DigitalCNC’s continued innovation in the global manufacturing industry.
