Siemens invited me to Munich, along with a number of other analysts, to its Innovation Conference. This is their showcase for new and upcoming technology. Day 1 included a visit to the Siemens R&D facility in Munich. Day 2 showcased technology that is market-ready.
Siemens apparent reluctance to send new developments out into the market is admirable from a quality and reliability point-of-view, but risky in a market which demands the latest innovations now. I get the impression that Siemens is not happy with making a solution work, it wants all of its solutions to work well. Indeed, the comment was made that Siemens had been working on the Digital Twin concept for 10-15 years, but expect it to take another 10-15 years to perfect it.
Siemens is not just a company that provides software to improve how manufacturers perform, it is also a manufacturer in its own right, with more than 270 factories worldwide. That is one of the key strengths it possesses – an in-depth knowledge of the challenges facing manufacturers. For customers, that means that solutions are beta-tested by manufacturing divisions within Siemens well before general release.
For infrastructure customers, the same thing applies. Siemens has its own Real Estate division that runs buildings in 2,000 locations. In much the same way as for manufacturing, building technology solutions are beta-tested by internal teams.
Siemens has a good story to tell about digitalisation. The fleshed-out Digital Twin concept with Product, Production and Performance Twins (matched by Design, Build and Operate Infrastructure Digital Twins) is a neat way to bring together Siemens’ traditional strengths in design and production control alongside the newer ‘Industry 4.0’ technologies that are emerging. I hope that the reluctance to send new developments out into the market doesn’t allow its competition to overtake and leave it behind.
Disclosure: Siemens paid for flights, airport transfers and the hotel stay in Munich. Siemens is also a customer of Cambashi.
Notes from the event
Day 1 at Siemens R&D
The afternoon at the R&D centre was split between expert briefings and hands-on demonstrations of technology.
Most of the technologies are based around MindSphere cloud, which is described as an open IoT operating system. Users can access native Siemens-built applications, but also can develop their own applications, plus access applications written by a rapidly growing ecosystem of third parties.
Every device will be connected and intelligent, but need to cope with longer cycle times – for example you might change your mobile phone every 1-2 years, but you won’t change your factory machines (or trains) for about 10 years or more.
MindSphere provides a single harmonized data layer that brings together software tools and physical devices. It is being used by early adopter users (as well as internal Siemens users), but is still very much in its infancy in terms of maturity.
Digital Twin is one of Siemens’ key technologies. Like MindSphere, it is already being used – indeed some of the underlying technologies, like product simulation, are well-established. But the packaging up into the virtual product, linked to the real world by MindSphere, is still a relatively new concept.
The digital twin is a digital representation of a physical asset. Siemens defines three different types:
Digital product twin – for design and simulation
Digital production twin – for process simulation
Digital performance twin – for continuous simulation during the lifetime of the product in its current configuration.
Knowledge from the engineering models are used to improve the accuracy of the digital performance twin – rather than requiring the system to learn everything from scratch.
Generative design modules can learn from the performance twin, enabling an optimised product design for how the product is actually used in practice. In this way, engineers are able to increase the quality of their designs.
The digital performance twin runs an online simulation during operation of the product. The simulation is fed by real data from the device, enabling the digital twin to be able to predict potential issues. For example, it can help an operator identify why a motor is going wrong, or predict the lifetime of a product. This continuous simulation helps to verify design decisions, or improve future designs. However, in order to run this continuous simulation, much smaller models are required – going from 1 million degrees of freedom to just 100.
Benefits of the Digital Twin
Some of the expected benefits of using digital twins are impressive. Siemens claims that you can accurately predict the temperature of a part inside a motor that is impossible to measure by conventional means. One of the research demonstrations showed this capability, alongside an Augmented Reality user interface that actually allowed you to modify the motor speed.
The simulation is one of the keys to the Siemens Machine Learning strategy. The reasoning is that, although industrial systems are still producing a vast amount of data, the variety and even the sheer volume is not the same as for consumer data. This means that teaching machine learning systems takes much longer in the industrial setting. Siemens intends to run multiple simulations to generate enough data to train its machine learning systems in instances where there is not enough real data available. In parallel, Siemens is also proposing to calibrate its simulation models using results from those same machine learning systems. How it intends to avoid simulations essentially verifying themselves is unclear.
The Siemens research team introduced its concept for using artificial intelligence to enhance its offerings.
The overriding model takes inputs from the environment using a range of ‘Perception’ tools:
- Sensor processing
- Image processing
- Speech recognition
- Text processing
The ‘Cognition’ tools then make sense of those inputs. Reasoning allows the system to draw conclusions from the raw data; learning allows the system to continually adapt and improve; creativity enables the system to generate hypotheses.
The ‘Decision’ tools enable decision-making, even in uncertainty.
The tools combine to determine a set of actions to take, then a feedback loop allows the system to learn from its mistakes.
The system brings together lots of data siloes, using Artificial Intelligence and Machine Learning to use different data signals in parallel.
The challenge for Machine Learning systems in industrial settings, is that, despite the wealth of data, there simply isn’t as much as for consumer applications. This means that machine learning needs to be supplemented with human knowledge.
Industry Knowledge Graph
The Industry Knowledge Graph is one of the ways that Siemens uses to cope whenever there is a lack of data for its artificial intelligence to use for its predictions. Essentially it’s a way of pre-loading knowledge into the system, so it doesn’t have to learn everything from scratch. It’s this semantic knowledge that means that we automatically know that a lion is dangerous without having been attacked by one ourselves. Similarly, an experienced servicing technician will notice an issue and know what needs fixing based on his prior knowledge of the system.
The combination of these ‘knowledge graphs’ and artificial intelligence algorithms are what Siemens describes as ‘Augmented Intelligence’. A typical workflow would be as follows:
- Understand – for example, conducting a tender search, which is often time-consuming and detailed – ideal for AI enhancement;
- Offer – configure a solution, work out the optimal layout – this may need to take into account lots of constraints;
- Produce – know how best to manufacture the product as configured;
- Supply – rank suitable vendors;
- Maintain – planning machine upgrades;
- Repair – learn from past product failures and propose repair activities
The capabilities of the artificial intelligence solution were demonstrated alongside factory automation systems. The ‘playful’ demonstration was of two PLCs learning how to play Pong. The learning mode consisted of them observing two humans playing. From this, they derived the rules and objectives. Note that the inputs for the PLCs were the pixels displayed on the screen, so the AI systems had to process a considerable amount of data in real time to make its decisions.
The second demonstration was of a pair of robots deciding how to assemble electronic components on the fly. Each robot used its camera to identify the part required, then decided what actions were needed to assemble it correctly. It didn’t matter whether the part was the right way round, on its side, or even in a specific place. The robots (or at least the AI driving them) had to work out how to ensure each part was correctly oriented, sometimes even using a combination of the two robots to achieve the result. Each build was performed slightly differently, but the final result was always the same. It was noticeable how slowly the system worked compared to standard production systems. However, this is very much an early prototype, so I would expect similar systems to be running much faster in a few years’ time, when the technology is ready for the market.
A number of other technologies were on display at the research centre. We were advised that these were very much early prototypes that were at least three years from general release. Having said that, some of the demonstrations seemed familiar.
The Virtual Reality headset, allowing service personnel to ‘see’ inside a large gas turbine was impressive, but sounded very similar to solutions that GE have been talking about.
More impressive was the Augmented Reality solutions that simulated the inside conditions of a working motor, but then also allowed the user to control the speed of the motor to control the temperature. Again, these are already being trialled by other vendors as ways to augment service personnel, although those solutions do not typically allow control via the AR system.
Siemens Innovation Awards
In the evening, analysts were invited to the Innovation Awards. This is an internal Siemens event, honouring staff that have made a significant contribution to innovation. Additionally, there was another award, which is sponsored by Siemens that rewards innovation that has made a significant contribution to the health of the German economy. This was slightly surreal, as the entire ceremony was conducted in English (the official language of Siemens)…
But the purpose of inviting analysts was clear – Siemens wanted to highlight the depth and breadth of its innovation. This is not just in design and automation software and hardware, but the full range of medical devices, trains and gas turbines. This range is important to remember, as it provides Siemens with an in-built store of live testers. For example, its Digital Twin capability will have been used in anger by its own manufacturing facilities.
Innovation day itself
Joe Kaeser, CEO, talking about digitalisation as a growth driver
Joe Kaeser gave the opening keynote speech. The theme was around the “need to create an innovative environment” within Siemens that encourages new product development.
One of the keys to innovation is partnerships with customers and the value chain (and even competitors). It’s about working together to add value. There is even a Siemens-funded venture capital program (next 47) to encourage entrepreneurs (including Siemens employees) and pursue their own passions on a path that could ultimately lead to commercialization of their ideas.
Something worth remembering is that Siemens is a very large company – it employs more people than Microsoft, Apple, Facebook and Google combined. Size can have a number of effects on R&D – larger companies often have bigger budgets and are more able to cope with setbacks. However, the bureaucracy within large companies can hamstring R&D efforts, allowing smaller, more nimble companies to make significant breakthroughs. Kaeser is keen to strike a balance. He wants to focus Siemens innovation on what will give results and create value. That innovation is not only in products and services, but also business models and processes, with the intention of making Siemens more innovative than ever.
Roland Busch, CTO
The full cycle of research, from initial University research through applied research and into product development, is being shortened by digitalization. It’s all about speed and scale. According to Busch, Siemens research depends on the following qualities:
- Deep domain know-how;
- Strong installed base / connected fleet (depending on product);
- Powerful ecosystem;
- Digitalization portfolio of MindSphere, software, services and security;
- Internal proof points.
Busch positioned MindSphere as connecting the virtual world to the real world, not just in the manufacturing industry, but also in construction.
Co-creation and co-location with customers is one of the key strengths for Siemens. There are now more than 900 MindSphere software developers, in 20 MindSphere Application Centers across 50 locations worldwide.
There were some examples of customers that are using MindSphere applications to improve their business:
- Algonquin College in Canada saving $3.7m annually;
- Amtrak gaining 99% availability;
- Deutsche Bahn predicting >80% of failures;
- A US Utility getting $1m savings through advanced metering;
- Siemens electronics manufacturing reducing lead time from 25 days to <7.
Jan Mrosik, CEO Digital Factory Division
Jan Mrosik summed up the problems that Siemens is trying to solve. Customers are increasingly requiring a combination of the following:
- Speed – reduced time to market;
- Flexibility – down to lot size of 1;
- Improved quality;
- Improved cost / efficiency.
Often these require new business models and improved security.
The solution – Digital Twin
Siemens is digitalising the entire value chain:
Design > Production Planning > Production Engineering > Production Execution > Product Performance and Services
This chain is underpinned by Teamcenter, which also integrates supplier management and logistics.
There is also a Digital Twin to support all 5 steps:
- Digital Product Twin
- enables capture, design, simulation and verification of models
- Digital Production Twin (Virtual Production)
- enables simulation of factory design and layout
- simulates production and automation engineering
- determines the code to run live production systems
- Digital Performance Twin
- Simulates how the product will perform in real life
- Provides feedback into Digital Twins focused on earlier lifecycle phases
- Can control or update products with embedded software
The Digital Twins form a self-learning loop to improve products and production.
Some customers were present to show how they are using the solutions.
Bausch and Ströbel (makers of pharmaceutical packaging machinery) are using the Digital Twin to reduce engineering time. They use virtual commissioning to create programs that direct the movement of the machine, instead of building full-size wooden models. The virtual model can simulate use by a variety of different people – for example different strengths or sizes.
Heller is using digital technology to change its business model from selling machine tools to selling “machine time”. This new pay-per-use model is supported by an application written on MindSphere that displays machine data in a dashboard. It is able to analyse maintenance situations and can contact customers to avoid uncontrolled failure.
Matthias Rebellius – CEO Building Technology Division
The division’s mission was succinctly stated as “creating perfect places with the power of data”. Smart buildings are a potential source of competitive advantage.
The opportunity to optimise building use is huge:
- 40% of energy use is from buildings;
- 80% of building lifecycle costs are from operations;
- 30% of real estate is flexible office space.
Operators want lower maintenance costs, but with high reliability and high operating efficiency. The Siemens solution addresses these 3 main areas of importance for operators:
- Comfortable and safe
- Users want optimized room conditions – smart sensors;
- They want good air quality;
- They want to be safe and secure.
- Energy and asset efficiency
- High asset availability;
- Reduced maintenance costs;
- Predictive / preventative maintenance – reduce reactive services and increase asset lifetime;
- Optimised energy footprint (reduced CO2 emissions).
- Space and user efficiency
- High occupancy;
- Asset tracking;
- Indoor navigation;
- Intelligent evacuation.
Infrastructure Digital Twins
Much as in Manufacturing, Siemens has a suite of Infrastructure Digital Twin technologies that cover the full scope of Design, Build and Operate. This is in partnership with Bentley Systems. The cover infrastructure simulation, construction simulation and asset performance simulation. The Building Technology division focuses on the Infrastructure Performance Digital Twin. Alongside the MindSphere applications, facilities managers can assess the performance of their assets and identify where things are not performing as they should.
The example shown was of an office building that was using power at unexpected times of the day and over the weekend. That flagged the system to identify the specific asset at fault – and air conditioning unit that was heating and cooling at the same time.
Siemens Real Estate is using the solution (as you might expect) and claims to be making €20m savings per year as a result.
Siemens has a very advanced vision of the Digital Twin, which very neatly brings together its technologies for design, engineering, production and product performance. It also recognises the need to have an open platform that 3rd-party developers and customers can run their own applications on.
The Digital Twin is still very early in its lifecycle and it remains to be seen what features customers will find most useful – and be willing to pay for. But Siemens is in a strong position to make the most of that demand, as one of the early leaders.