Man Proposes, Cloud Disposes

Factory machines need modern control systems to work efficiently. Yet, the software used locally is often outdated. A research project at the University of Stuttgart is now working on ways to update or completely replace control systems from a remote location, which would have many benefits.

Modern production plant is multi-talented, a fact, which is exemplified by a highly flexible furniture factory. A conveyor belt transports wooden boards to a place where they are sawn down to size and milled to the desired shape. The machinery involved performs a plethora of discrete and fully automated tasks cleanly and with a high degree of precision, whereby complex control systems ensure that each machine always “knows” exactly which parts to process and how. Finally, the factory-made, customised sideboards, shelving units and chests of drawers are sent out to customers. Control systems are the brains of any production facility whether it makes furniture, machine components or car accessories. The electronic component groups are placed either directly in the individual machines or in switch cabinets connected by cables and positioned within the production hall but some distance away from the production line. These control units run complex software programmes, which control the production process in accordance with the firm’s specifications.

Factories today depend on such computer programmes to fulfil their orders in an efficient manner, yet the programmes in question are frequently antiquated. This is because, unlike computers that run the Microsoft Windows operating system, they are not automatically updated via the Internet every few months. On the contrary, for many companies, updating production plant represents an organisational challenge. As Professor Alexander Verl explains: “it is not unusual to see a member of staff walking around the production facility with a USB stick in his hand from which he copies the required update to each machine individually.” Professor Verl heads up the Institute for Control Engineering of Machine Tools and Manufacturing Units at the University of Stuttgart. His research concerns the viability of off-loading industrial control technology to the Cloud. In this context, the term Cloud, simply refers to a central server located at a service provider’s facility that may be hundreds of kilometres away, which the factories can access via the Internet. One of the benefits of this type of virtual plant control system is that production software could be updated or even completely replaced in a matter of minutes.

This would open up new possibilities for production. Until now, automatic production processes have been organised hierarchically, i.e., update decisions are taken at the upper management levels, whilst those lower down in the hierarchy implement the plan until, in the final step, the machine in question is equipped with new control software. There are advantages to this system, but it does lack flexibility, which is becoming increasingly important in the modern production industry. Production plant attached to the Cloud (Internet) could be updated automatically as soon as new software became available.

Today, any advances in modern machine tool manufacturing will be entirely reliant on software. The actuators and mechanical systems are already fully developed and offer little scope for further improvement. “Most hardware is pretty standard”, says Verl. This is because machine tool manufacturers procure their components from an ever diminishing pool of suppliers. “Therefore”, Verl continues, “without bespoke software, it is virtually impossible to gain a competitive edge.” Because, the quality of a piece of manufacturing plant is determined by the software that controls it. If, for example, a fully automated milling mechanism is used to shape a wooden board, then it needs to do so as precisely and as rapidly as possible. The more up-to-date the software, the more efficiently even an older machine can perform its allotted tasks. In future, it may be possible to download updated control software directly from the Cloud to the machine by a simple mouse click, and use it immediately.

The situation is similar with respect to special functions, such as collision monitoring. The associated software ensures that mechanical arms used to wield and apply a given tool do not cause damage by following collision-prone trajectories. The programme knows the length, breadth and height of the machine as well as the action fields of action of both the arm and the tool. Based on these parameters, the software can calculate the movements for the subsequent task in advance and ensures that the workpiece is processed right from the initial step in such a way that collisions are avoided.

Whether a collision avoidance system is required or not depends on what is being fabricated and how. In some cases, it is not required and can be switched off, which saves resources, as the additional process costs both computing power and sometimes time. “Therefore, the collision monitoring service could be paid for as a value-added service and could be toggled in from the Cloud whenever it is really needed”, says Felix Kretschmer, an expert in networked (distributed) production who works under Professor Verl. Kretschmer coordinates the PICASSO project, which stands for Projekt Industrielle CloudbASierte SteuerungsplattfOrm für eine Produktion mit cyberphysischen Systemen ([the] Project for an Industrial Cloud-based Control Platform for Production Using Cyber-Physical Systems). The project involves collaborative research between technology companies and the University of Stuttgart into how to increase industrial production efficiency through the flexible provision of control system technology. The project is funded by the German Federal Minister for Research and is supported by the Karlsruhe Institute of Technology (KIT).

Many of the functions performed by production plant can be off-loaded to the Cloud, but not all of them. Sensors that determine the dimensions of a given workpiece must, of course, remain within the machine. “But”, says Verl, “everything that is calculated on the basis of these dimensions can be done in the Cloud.” “The processes need to happen extremely rapidly, usually in a matter of milliseconds, but, anyone who regularly accesses the Internet knows how diffi cult that can be. Yet, whilst Internet latency can be a minor annoyance for the offi ce worker, the consequences of unduly long transmission speeds in a production scenario could be very serious. As Kretschmer explains: “the revolving hourglass familiar from our PC screens could result in a total production plant outage.” Or it could cause costly production errors. If a milling machine (or router) suddenly stops whilst cutting its way through a piece of timber, the result is a black area of burnt wood. The workpiece is damaged and processing cannot continue. Therefore the quality of the Internet connection is crucial.

The transmission speed depends on the network technology used. It needs to ensure that a production facility’s data signals are prioritised over other data traffi c using the same network. Additionally, data from the Cloud needs to arrive simultaneously at the factory. “Our research is about fi nding ways to do this”, says Verl.

So-called time-sensitive networks offer one solution. The way they work can be explained using the example of home loudspeakers connected via a WLAN. In this case too it is important that signals reach each of the speakers at the same time. That was fi rst made possible using technology developed by a company called Sonos. As Verl explains: “it was they who succeeded in creating a prioritisation protocol for use in home WLANs. Attempts are currently ongoing to create a similar method of transmitting production data packets, whereby one of the most signifi cant challenges is to unify the various available standards, which differ signifi cantly from one another. However, there will always be a limit to the rate at which data can be transmitted, the limiting factor being the speed of light, which is the speed at which electronic signals travel. In the course of the PICASSO project, for example, the time taken for an electronic signal to travel between a Google data centre based in Northern Europe to Stuttgart and back was measured. The lower limit was around 20 milliseconds. However, for certain production tasks a single millisecond represents the critical threshold. “What this means”, says Verl as he describes the challenge, “is that the control system for an application based in Stuttgart cannot be located further than 150 kilometres away, which would be somewhere like Frankfurt am Main.” Yet, he does not see this as an insurmountable obstacle. The Cloud could be distributed across a number of data centres, each of which would be rapidly accessible by factories with a certain region. It would also be possible for companies to maintain their own Cloud systems located as close to the centre of their various facilities as possible.

Security presents another challenge for Cloud-based control systems, and that too is being tackled by researcher in the PICASSO project. Hackers have demonstrated on multiple occasions that it is easy to gain access to industrial plant via IT systems. For example, cyber criminals succeeded in shutting down a blast furnace at a German steelworks in 2014. The Israeli secret service used the Stuxnet computer worm to sabotage Iran’s nuclear power programme a few years ago. If the connection between production plant and control systems runs via public networks in future, hackers could easily fi nd a way in. So, the IT security demands are high. The data must only be accessible by authorised people and systems. Unauthorised manipulation must be impossible, yet, at the same time, the data must always be available whenever it is needed.

The data must be encrypted as a matter of course, but there are still debates surrounding the question of whether that should be the case throughout the entire connection between the Cloud server and the production plant. This is because the various parties involved have completely different interests. The factory operators want absolute control over access to their systems, which means that they need to know the exact origin of data being transmitted to their machines. For them, therefore, end-to-end encryption is out of the question: as a minimum, they want to be able to check the origin of the data just before they are fed into the factory systems. However, the Cloud operator is not interested in this: they want to protect their proprietary know-how, which is encoded within the transmitted data packets, and which nobody but them should have access to. Therefore, a reconciliation of interests between the sender and receiver is an important prerequisite for effective monitoring of the data stream.

Modern encryption techniques are extremely secure. As Verl points out: “even government agencies struggle to crack modern encryption codes.” These days, the most common way in which such codes are cracked is simply by spying on someone who holds the decryption key, which can sometimes be done with very simple methods, such as breaking into business premises and stealing the relevant passwords. However, new IT applications are not the only route in for hackers. Older systems provide points of attack.

Verl knows that: “some factories are still using operating systems that are 20 years old.” In such cases, a single USB interface could be all it takes to access and manipulate a system. However, the plant manufacturers are just glad if their machines work as promised and that they are able to sell them. They have no interest in continuously updating them, which is precisely what could leave them open to cyber attacks. For Verl that is an important argument: “systems attached to the Cloud can be kept continuously up-to-date with the newest technical developments, particularly in terms of security.”

Heimo Fischer