Totally automatic - a standardised model for subcontracting that will be replicated around the globe

13 mins read

Why can’t NC programming for subtractive metalcutting be as easy as it is for additive manufacturing? Basically, present a 3D file to a programming system and press ‘go’. Andrew Allcock visited a company that has achieved that, but also has a much larger ambition

The reality of NC programming for complex prismatic parts, as readers know, is that the generation of a set of instructions to drive a CNC machine can take days or even weeks. Even for more simple parts it still consumes hours of skilled, experienced effort using “horrible-looking software, CAM”, so posits Theo Saville, CEO of CloudNC, a UK-headquartered, venture capital-funded software development and CNC machining company that has just come out of ‘stealth mode’.

Whether making a prototype or setting up to make millions of parts using metalcutting CNC machine tools, NC programming is a constant, but the lower the batch size the greater is the proportion of the entire design-to-make cycle that NC programming consumes. “If you are in the low-volume batch-to-prototype space, that CAM programming can easily make up 90% of what we are charging to the customer. So, small batches and prototypes are phenomenally more expensive than they need to be,” Saville states.

A secondary issue is that most machining cycles generated via the traditional programming route are inefficient, he adds. “There are trillions of ways of making even a simple component with one of these CNC machines. However, only a few of those ways is truly fast. Humans are very good at coming up with an acceptable way to make components, but it is beyond our capability to find the optimal path to make a component with the available equipment, because there are too many ways of achieving it.”

In talking to industry bodies, the CEO says that almost everything out of what he suggests is a global £100bn+ component machining annual market is made at a rate that is half as fast as is possible, meaning costs can be halved “with no changes to the equipment whatsoever”.

He continues: “If you could make this process autonomous; if you could automate the programming and optimise the resulting machining program, then you could more than halve the cost of making all those metal components. In some cases, depending on the method of manufacture, you could reduce part cost by a factor of 10. And that is what we have been working on, in secret, for the last four years.”

The developed software is fed a 3D part file having associated tolerance requirements, the starting block size is defined, with the manufacturing equipment parameters – material, machine, workholding, tools – known to the software. After that, the generation of the optimal NC program is achieved in “minutes”. An aerospace part that would have taken a week to program can be processed in just 10 minutes and is “pretty much guaranteed” to work, not requiring the attendance of a person through the first cycle. And, of course, this reduced programming requirement also addresses the skills challenge. There are far more NC programmers retiring than are being trained, Saville advises.

What this means is that for small quantities or prototypes, you can “vastly reduce the cost” and deliver the component to a customer faster. And with a cycle time that is at least twice as fast as what might have been created, the costs also plummet, and that will be achievable in just a few months’ time, he offers.

But he poses the rhetorical question “So why hasn’t this been done before?” Because it is “phenomenally complex”, he answers, adding: “This is probably one of the hardest pieces of technology being built in manufacturing right now.” Part of the solution is the power of cloud computing, but shear horsepower is not the answer, it required many areas of development, including “completely new ways of representing geometry in a computer”, plus venture capital backing to the tune of £11.35m since 2016, with a further round of fund raising scheduled for next year. That support has allowed the operation to attract “the truly best software engineers in the world to solve this problem, which cannot be solved with just a lot of regular software engineers...requiring the development of a whole new computer science”.

CloudNC's Chelmsford factory, located behind a new retail park - see later

USE IT DON’T SELL IT
Yet the company is not going to sell this software; instead it is going to use it to support the creation
of a global network of manufacturing companies that can exploit it and so deliver a service that others simply cannot match. And it already has its first factory operational. A 25,000 ft2 set-up in Chelmsford, opened in January this year (see extended online article) that is already serving customers with parts that are made “faster, better and cheaper, using CloudNC technology core software and everything else we are building around it”. And that last part of the statement is also important.

By controlling the environment within which the software is applied, the variation that the software would have to accommodate were it to be sold to firms having many different types of equipment is eradicated, the CEO underlines, which means that software development can be highly focused. And the standardisation of the manufacturing environment also offers a second benefit – the ability to optimise all other elements of the manufacturing process chain, he adds. One of those is the automation of quoting, which can go from a process taking days or weeks to one that is instant and automatic.

In addition to that benefit of a standardised manufacturing environment, with known manufacturing times, he says: “You can do very interesting things around scheduling your factory that are just not options available to anybody else...You can build all this technology that makes the entire process, from sending a quote out to getting the component to the customer, hyper-efficient because you have unlocked this automation at the core.”
CloudNC’s first factory, while earning some money by producing parts is, in fact, really the focus for “developing a scaleable model, a blueprint of massive automation and efficiency that is designed to be ‘cookie cutter copy and pasted’ across the world”. And that scaleable roll-out of factories will start following the next round of funding in 2020 “very quickly”. He suggests that there could be four factories by the end of next year.

A big enough vision? Big, but not the ultimate destination, Saville reveals. “Our ultimate vision is somewhat bigger. CNC machining was just the obvious starting point for us, because it is a phenomenally valuable problem to solve; really difficult, but we figured it would be just about possible. It turned out to be much, much harder than we thought it would be – we thought we would be where we are today about three years ago. But what we really want to achieve is full automation of the manufacturing of primary components – parts that are made out of blocks of metal or plastic.

“Today, if I want to buy metal components, the process is truly horrible. I am going to have to email a lot of factories, most of which will not respond; those that do will get me a quote in a number of days; if I pay for it, it is more expensive than it needs to be; quality could be super-low; I will not hear anything about my order as it is being produced; it will hopefully turn up on time, but mostly in industry things are late.” His criticism is not levelled at the people in industry but the software and systems they work with and within, it should be said.
The process that CloudNC’s CEO envisages takes in: uploading a 3D model; instant quote; instant design for manufacturability feedback – price and cost drivers being identified within that, allowing for modification; online purchase – it doesn’t matter where it is made, you can forget about it, he says, adding: “It’s like Amazon. I know that the part is going to turn up quickly, at the right price, and I can definitely rely on the part being correct.”

All the clever stuff will happen in the supply chain, using automated factories where the machines already know how they are going to make the part, so the parts will definitely turn up on time. “You’ll be able to receive thousands of parts in a very small amount of time,” he assures. Even if they are spread across several factories, they will be identical because of the standardised nature of their production environment. “If you can automate the CNC machine fully, there’s nothing to stop you distributing manufacture across thousands of machines. That means that where today it could take a few months to get a run of a thousand units, in the future it could take you just days.” (‘Swarm machining’ seems an appropriate way of phrasing it.)

In developing this software and the manufacturing hardware combination, CloudNC is, says Saville, bringing together the best manufacturing culture that is employed in companies such as Toyota and the high growth tech culture found in companies such as Google. “We are trying to mix those together to create a new tech-manufacturing culture that is super-fast and innovative, and able to solve problems like this one.”
And he concludes: “Fifty years down the line, we expect all types of manufacturing will be autonomous. Whether you want one unit or a million, a part or an entire assembly – even a complex assembly like an engine – you will just throw a 3D file into a supply chain system and get it back without any human involvement. What enables this is if every machine knowing what it is capable of producing, with factories of machines at the top of the supply chain able to call on others in the chain, completely automatically.”

Box item

The people behind CloudNC

The board of directors takes in:

[] Theo Saville has a background in 3D printing research, sales and manufacturing research at the Warwick Manufacturing Group. It was while there that Saville came up with the idea for automating NC programming in the same vein as is 3D printing, when he saw what software was used and how long it took.

[] CTO Chris Emery, ex-Google, has a background in high performance computing and graphics processing units.

[] Siraj Khaliq is founder and ex-CTO of the Climate Corporation, which was sold to Monsanto for $930m.

[] Paul McNabb is ex-chief strategy officer, Cisco Systems.

[] Chris Mairs is ex-CTO Metaswitch Networks

[] Simon Murdoch is ex-vice president of Amazon Europe, reporting directly to Geoff Bezos

Apart from venture capital backing, which comes from “some of the top artificial intelligence investors in the world”, the company has received funding from government-backed body Innovate UK, as well received help from the High Value Manufacturing Catapult.

As of August, CloudNC was a circa 75-person operation, including 20+ software engineers attracted from some “phenomenal” companies, but will likely employ hundreds in another year, the CEO reveals.

Extended article from here

An editor calls

Following Theo Saville’s presentation to the press earlier this year at which he outlined CloudNC and its ambition as detailed in the main article, Machinery went along to ‘Factory One’ in Chelmsford to see the business end of the initiative.

CloudNC’s industrial unit is about 1.5 miles out of the town centre on the new Clock Tower industrial estate development. On a walk to the plant, a now-familiar main road shop scene plays out: car dealerships, a Homebase DIY store, car wash, Big Yellow self-storage and then a major retail park, Clock Tower, where some of the usual suspects are located – Aldi, M&S, Furniture Village, DFS, Dunelm and Costa included. Across the road, there’s Topps Tiles, Halfords, Formula One Autocentre and Hot Tub Barn.

Clock Tower retail park - the usual suspects; CloudNC is hidden behind this main area

There was one large factory on the route, global electronics component innovator Teledyne e2v, and out of view behind the large retail operations already named there were some small industrial activities. But the Clock Tower combined retail and industrial development is located on what was previously drinks maker Britvic’s 11-acre site - Chelmsford’s best-loved businesses, so the local press hailed it in a retrospective article about the company that set up its operations in the town in 1955. The factory shut in 2014, following its purchase by fellow UK drinks maker AG Barr of Irn-Bru fame.

Such a post-industrial townscape could be seen any many places, yet within this setting nothing less than the future model of global metal subcontract machining is being created – a disruptive model. It is an interesting juxtaposition and one that is entirely invisible to the shoppers flocking to the retail park next door.

CloudNC’s ambition has already been outlined in the main article, but what does such a company look like on the ground – ‘where’s the beef?’, so to speak. First up, the operation looks like any modern industrial unit from the outside. Sure, the name, CloudNC and not VC Engineering or VC CNC Machining, alludes to a more modern take on matters. In reception there is a tell-tale clue to the sort of company you are entering. On a table is a copy of ‘Start with why’, authored by Simon Sinek, who started a movement to help people become more inspired at work, and in turn inspire their colleagues and customers. More than 28 million have watched his TED Talk of the same name, the third most popular TED video (https://is.gd/oxehep) of all time, according to Amazon reviews.

Leading off reception on the way to the shopfloor and NC programming areas, you pass through a rest/eating/drinking area that also boasts a football table – a piece of equipment that is more associated with software tech company, perhaps. And the people in the facility are definitely well below the traditional engineering firm demographic; the company average employee age is put at 30 or under by CEO Saville (probably between 10-20 years below the average age of the engineering sector; a third of the manufacturing sector’s workforce is over 50)). That’s generation Y, Facebook creator Mark Zuckerberg’s generation. That generation is defined as: digital natives; aspiring to freedom and flexibility; and digital entrepreneurs that work with organisations, not for them.

Not your usual subcontractor rest area - a visual clue to the tech sector/manufacturing culture mix

Some 35 of CloudNC’s workforce are located at Chelmsford and the remainder at the company’s London HQ, where around 25 software engineers beaver away. And this 75-employee-in-total operation, although it will become much larger, boasts its own internal eight-person HR and talent development department that searches for people that fit CloudNC’s culture - the company is “very precious about our culture”. It wants calm, rational, passionate, gifted people that easily take onboard new things, that get along with others and that have a growth mindset, so it selects for personality, attitude and aptitude ahead of specific skillset. In order to find such people, it must interview many and can make the choice to employ those that fit its parameters, “because we can afford it”. People are very definitely the core asset of the company and its most visibly distinguishing feature.

CloudNC's London HQ, where the software writing is centred

On the shopfloor, which is still mostly space but will house up to 20 CNC machines tools, it appears more traditional. There are DMG Mori 5-axis machining centres plus Erowa automated pallet loading; Mazak 5-axis and FANUC 4-axis machinery, as well as a CNC lathe (required to support a full-service operation, but it is prismatic parts where the focus lies).

This facility the company’s R&D operation; it is here where CloudNC’s standardised manufacturing system and associated processes are being created, and its automated NC programming software tested and applied. It is the place where a high-tech digital world meets the more traditional physical world of metalcutting.

The company is going to double its current space in Chelmsford by acquiring an adjacent unit. This is where CloudNC will develop its ‘template’ – the model that it will then go on to ‘cut and paste’ in locations that will be guided by “the availability of suppliers, customers and talent”. Not any old customers either, it should be said. CloudNC wants to work with companies spending £500k and upwards on subcontracting and ones that are “ready and willing to be fully engaged at all times” with CloudNC; high quality customers. With Factory One having started operation in January this year, the company already has 30-40 customers, with representatives from all major sectors of industry – medical, oil and gas, motorsport and automotive aerospace. But aluminium has been the main material so far tackled using its automated NC program creation software. A mild steel test was underway on the day of Machinery’s visit, however, with the software validated for further materials progressively.

Expansion of the successful factory template will follow a so-called Round B of venture capital funding, £20 million and up, to fund the purchase of “a three-figure number” of machine tools to drive what in VC jargon will then be a “growth engine”.Further funding after that will keep growth coming off the back of a successful model, with four-figure machine tool orders posited. Sourcing large volumes of machine tools to populate the cut-and-pasted factories is a potential issue, it is understood, because there could obviously be supply constraints.

Characterising the type of investment that CloudNC represents as it is seen through VC eyes, Saville uses the term “moonshot” – meaning the opportunity here is large and that it will be backed by commensurately large funding to achieve it.

But there is detailed work to be done ahead of this aggressive expansion. In walking round the shopfloor, the CEO explains some of the current issues that are being worked through on the machines. Mostly these relate to software integration between machine and automation, machine communication capabilities and what they can communicate about themselves, plus variations between machining results on different brand machines running the same basic NC program (but post-processed for each machine), the latter due to differing machine controls and machine parameter settings. But they are all fundamentally software issues and so are solvable, he offers. For example, once a machine is configured, any further machines of the same model will be supplied with the same configuration. But there are even some basic mechanical engineering issues, such as laser-based tool setters that get contaminated with swarf, that need solving.

There is detailed work to do still at shopfloor level at Chelmsford

The CEO is clearly frustrated at the lack of machine connectivity and communication capabilities that come as standard. They are not sufficient and he suggests that, as such, machine tools are not “fully functioning…they should work when they hit the ground…they should be fully connected - it’s 2019 not 1999”.

But in terms of cutting, its chosen equipment can achieve 50 micron tolerances easily, but with sub-20 micron tolerances, and certainly single-figure micron requirements, a ‘semi-finish/measure/offset update/finish cut’ sequence is applied to assure the result. The idea is to develop in-machine processes that are fully automatically capable of producing correct parts. “Even if a cycle time is longer, we want a totally reproducible process that doesn’t require people to be involved”.

But on the topic of automation more generally, the jury is still out on this, the CEO explains. “Cells [of machines] run by people that are directed efficiently is a cheaper option than a robot. A person could manage up to 20 machines. People versus robots is still an open question.”

And just where will such software to direct people efficiently come from? It will be developed internally, of course. CloudNC’s CTO, Chris Emery, is currently developing a Factory Operating System (FOS) that will “run a factory how it should be run…ERP is horrible”. FOS, which will employ AI, will understand the factory environment and have a “a probabilistic understanding of what could go wrong” and so will build appropriate slack into any generated schedule to make sure that parts will be delivered on time. And should anything go wrong, FOS will learn, instantly recalculate, update activities and direct people in the necessary manner. In this way, CloudNC factories will have a “much higher machine utilisation [than is the current norm] without disappointing customers”, the CEO explains, adding: “There are very few limits as to how far we can take this – the system can order tools, materials and schedule.” So, software in the manner that Amazon runs its warehouses, Machinery suggested? Yes, a good parallel, he agrees. Amazon is highly vertically integrated, develops unique technology in many areas and also acquires companies that have it. Acquisition is not to the fore in CloudNC's strategy, however.

The people-based shopfloor processes that sit within any so-directed environment are still being fine-tuned via lean/kaizen training. There is now a high-level employee tasked solely with driving a continuous improvement regime. Tool set-up time reduced by a factor of four in just a few days was an early benefit. Machine set-up will be next. “There will be regular kaizen events; it will be never-ending…we will be improving at high speed.” Such things will go towards speeding the journey of parts through the factory, currently up to three weeks but which “should be days”.

If Theo Saville’s initial presentation of CloudNC’s strategy and ambition was impressive, the visit to its R&D lab factory in Chelmsford served only to underline the focus and effort being put into realising it. And it also clearly demonstrated that the company is a new type of subcontract metalcutting operation whose completeness of its original-software-directed, highly efficient standardised manufacturing system’ approach and ‘cut-and-paste' global ambition places it ahead of other machining operations that proffer online instant quoting and fast delivery.