Shulou(Shulou.com)06/02 Report--

By the end of the year, there will be all kinds of mysterious articles such as "Top Ten Technologies in 2019", "these five Technologies will change the World next year", "these eight technologies are going to change, I don't know you won't get a raise next year."

In fact, such predictions are either common sense or simply unreliable, and there are not many insights. But wouldn't it be a little embarrassing if there was a technology that wanted to be on this kind of list?

This is like a high school student who has been admitted to the school's announcement that "these ten people will be admitted to Peking University next year". It is obviously worth being happy. But if you make the list for three years in a row, the family will probably be in a hurry.

There is such a technology, and if you are interested in going through the list of predictions from 2016 to today, you will find that there is a technology on it all the time, which is called Digital Twin.

The so-called digital twin, in a more formal interpretation, refers to the digital twin reached by the simulation model integrated by the physical entity of the product in the information space, and the related technology of using the digital twin to realize the archived management of the product life cycle.

Don't you understand?

It's okay. It doesn't matter. To put it bluntly, digital twins means that there is a machine in reality, and we build exactly the same virtual machine in the computer. In this way, we can beat about the digital world and see when it breaks down, and we can predict when the real machine will be overhauled.

Does that sound reliable? In fact, digital twins do have a wide range of applications in various fields, such as architectural engineering, smart city, aviation design, and is also one of the core technologies explored in the famous German Industry 4.0. Siemens is recognized as the company with the largest investment and the most in-depth exploration of digital twins in the world.

But the question is, why does this sound like a solid technology always live in the prediction that "it must be hot next year"?

This technology is really the "art of killing dragons" in the world of the Internet of things. Is it of no practical use?

The purpose of this article is to answer these questions, otherwise why else would I say so much? But note that we have to go through a logical turn at the end of the first paragraph, and we must keep up.

Why does it always live in "next year"?

Aviation Weekly once predicted that when an airline receives an aircraft in 2035, it will receive a set of digital aircraft at the same time. This set of digital aircraft contains every part, every structure of the real aircraft, and ages with every flight of the real aircraft. In this way, any problems with the aircraft can be sensed in advance in the digital twin system, thus bringing aviation safety to a new level.

However, this idea was subsequently opposed by people in the aviation industry, who believe that relying on virtual mapping to judge aircraft faults is a really unreliable fallacy. The differences in air pressure, air flow and temperature in each flight affect the mechanical structure of the aircraft in varying degrees. This kind of influence cannot be finely reflected in the virtual world, and it is only by relying on the digital system to judge the real situation of the aircraft that it is really possible to bring the aircraft to danger.

This debate reflects an innate problem in the field of digital twins: the idea that digital twins work well in areas such as manufacturing. However, digital twins cannot really copy physical details, but the industrial system must ensure a high degree of precision. Leading to the disappearance of such technologies hovers between looking cool and actually no one is using them.

One concept that must be clarified here is that since 2002, when Dr Michael Grieves, a professor at the University of Michigan, first publicly mentioned the concept of digital twinning, it is not a sequence of technological methods, but a technical goal.

Like AI, artificial intelligence is actually a collection of technologies that simulate human intelligence, thinking, and emotion, not just expert systems or machine learning. The same is true of array twinning, which is a technology cluster composed of various technological solutions aimed at replicating real production systems, rather than actually having a technology called digital twinning.

In the digital twins we discuss in general, we design physical simulation, sensor system, big data, immersion technology, Internet of things data visualization technology and so on. In short, technologies that can participate in the goal of physical production system replication can be counted as components of digital twins.

Precisely because some of these technologies are experiencing rapid development in recent years, digital twins have frequently entered the forecast list, giving people the impression that the technology will come soon.

However, if you really want to copy a production line or a plane completely in the virtual world, you will encounter a number of obstacles, such as:

1. Lack of general platform. Digital twinning involves multi-dimensional technologies, such as design, sensing, virtual reality, data identification, physical virtual and so on. These are difficult to express in a unified platform.

2. There are not many devices that really need twins. Let's recall that most of the machines in the factory do not have virtual backups and up-to-date shadow systems. So the digital twin is likely to be a very small type of technical service with very small commercial space. There are not many industries that are needed, and they are mainly customized.

3. Physical representation is still difficult to carry out digital simulation. Today, the "digital twin" services provided by many cloud platforms of the Internet of things can only provide a data monitoring and 3D model.

4. The calculation power required for a huge production system, industrial system, and even transportation system is astonishing. However, it is doubtful whether such a large amount of computing power can maximize the return.

From this point of view, the real process cycle monitoring and prediction of the digital twin system is still too far away from us. But to some extent, we can call all kinds of existing technologies part of digital twins, and even CAD mapping is in the twin production system.

But if a very strict boundary is defined, the digital twins may still be alive in the laboratory.

But don't be disappointed, this story is still a little similar to what artificial intelligence is like today. For example, if we want Ultron's artificial intelligence, it may have to wait until 8102, but if we want to use artificial intelligence to complete personal face recognition or something, then 2018 will be fine.

The same is true of digital twins. Although the ideal industrial digital twins or even urban digital twins are unreliable, if we regard it as a train of thought and are willing to admit that incomplete digital twins are also valuable, then there may be unexpected discoveries in the industrial service market in the past two years.

For example, relying on the technological trajectory of data mining + IoT cloud + AI, the digital twins of industrial data are showing great potential-and don't have to wait until next year.

The key point of this logic is, why do we have to see a detachable and assembled machine on the screen, as in science fiction movies? We just need to make the machine emit more productivity through the calculation and analysis of the virtual world.

Machine learning + data mining is activating digital twins to some extent

Most of the industrial production equipment, in essence, is to put into production materials, output production results of the data operation. Among them, the number of resources used, the rate of good products, production efficiency, production results, these most critical nodes are also data. If we give up direct visualization and retain the digital twins of physical properties, and just twinning the data system, it does not seem to face a particular technical challenge.

With the rise of the concepts of cloud computing and digital upgrade, more and more cloud service providers and enterprise digital service providers begin to provide data-level system twin services based on data collection, such as Oracle provided similar capabilities a long time ago.

But for enterprises, the more embarrassing place is, after my machine is twins, what can I get? Seeing a lot of numbers running on the screen doesn't seem to mean much to the enterprise. The improvement is nothing more than two points: one is that enterprise managers and engineers can see the global data more accurately; but the enterprise data has been backed up and can be queried if something goes wrong.

These two points are meaningful, of course, but compared with the high service charges, it seems to be worth thinking about.

The arrival of AI, to be exact, the integration of data mining technology and machine learning technology, is activating the new vitality of this crude version of digital twins.

In the traditional sense, all kinds of raw materials, equipment, personnel and quality inspection in a factory are independent production systems, which cooperate with each other and rely more on the experience of workers. The so-called production line is often to retain the maximum production time for the previous process, and then enter the next process.

This is like an intersection in a foggy day. For fear of a collision, we have to wait for the previous car to go a long way before the latter one dares to drive past. Great production efficiency is wasted in it. If we use machine learning technology and the connection of simulated production links in the digital twin system, we can peel off the cognitive fog between production links and direct the traffic in the workshop from the perspective of God.

Of course, the imagination of machine learning plus industrial data is much more than that. The problems of energy consumption, batching, the causes of defective products, and other industrial production problems can be solved by similar capabilities.

Therefore, the imagination brought by AI has reactivated the enterprise application value of data twins. Including 3D simulation machines, with the help of many AI algorithms, it is also possible to predict damage points and maintenance times.

As a result, the story turns out to be like this: today, when we expect the all-physical mimicry of digital twins to be still far away, digital twins may be based on the landing of industrial IoT+AI, creating another value. And this development opportunity is reversely affecting many industrial relations in the digital industry.

The trial version of digital twins also excites industrial IoT.

Finally, let's talk about some of the changes that digital twin technology has brought about in the AI+ industry today. In the B-end intelligent technology market, which focuses on more and more vitality, the value of digital twins is to provide enterprises with a frame of reference for backup, transfer, learning and logical analysis in the virtual world. Digital cloning has been introduced into many intelligent industrial platforms, and the value-added services based on digital twins for enterprises are becoming more and more diverse.

In the industrial IoT system, the more mainstream intelligence goes through such a process: first, based on the data collection and sensing system, the data acquisition system is built in the perfect part; however, the data is uploaded based on IoT cloud, so that the enterprise has a digital foundation; then the digital twin solution is adopted to build an abstract image of industrial equipment in the virtual platform. Then, according to the specific objectives, we use technologies such as deep learning algorithm to complete the intelligent analysis of the data, and give the feasibility analysis of optimizing the production process; finally, based on the analysis results, we carry out the technical optimization and manual optimization of each link.

Of course, this is only a basic logic, when solving specific problems, enterprises have to go through ever-changing special problem handling. In the whole industrial intelligent system, digital twins bring many new possibilities to enterprises. For example:

1. The threshold of industrial IoT has been lowered. Internet solutions for industrial production, such as remote analysis and centralized data monitoring, become possible. An enterprise does not necessarily have to hire powerful data experts or AI scholars, but can find industrial optimization solutions remotely by means of digital twins.

2. Customized production is easier. Through the combination of digital twin technology and AI, it will be easier for enterprises to solve the problems of design and production relations of customized industrial products. Both fine production and fast-flow production have become easier.

3. The "experience" of an enterprise can also be twinned. Today, many enterprises are reluctant to try digital transformation because many production processes in enterprises are abstract experiences that are explored step by step and handed down from hand to ear. Blind digitization may cause workers' maladjustment and reduce production efficiency. Another possibility brought by digital twins is based on machine vision and data absorption devices. Enterprises may twinning things that cannot be realized, such as production experience and process habits, on the data platform. Achieve digital integration of industry experience and industrial entities.

Although at present we can only use the trial version of the plain digital twin technology, probably not even by that name. But in the context of an improvement in the technological system, this technology is indeed likely to bring no small enlightenment to the real economy. Although digital twins like shadow planes may still be waiting for years, it doesn't matter.

Let the prediction be classified as prediction and practical work. Many technologies do not have to wait until they are fully mature before they can be applied. In this atypical case, we may be able to obtain a more typical technical possibility.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.