How does Siemens, the most popular industrial Inte

The market of industrial IOT is growing. How does Siemens, a veteran company, fight

for Siemens, artificial intelligence is not only an exciting topic, but also a highly demanding work content

if the digital twins are compared to the joints of Siemens, which plays the role of organic connection of industrial data, the existence of artificial intelligence for Siemens is undoubtedly to transmit the value and ability of data to a deeper and further distance through the import volume affected by the massive import of precious non-metallic minerals such as diamonds, and extend to those branches and ends that are not easy to be excavated or difficult to find before

take a simple example: does the old workshop need digital upgrading

the first problem we encounter is to master the basic situation of the workshop, including the operation status of the equipment, the type and structure of the system, operation and maintenance and other information

then, realize the preliminary digital twins

associate these data with different forms and sources to build an integrated semantic model, so as to clearly control the overall situation of the workshop

Siemens helped Shuangxing group build a digital factory. The picture comes from the planning and definition stage of Shuangxing. The next question is often: how many benefits can digitalization bring? Is it worth the investment? Is there much room for improvement

"it is difficult for most customers to have a clear grasp of these problems at the beginning. With AI, we can do intelligent analysis based on semantics and data, and then help customers solve these problems in digital upgrading through simulation verification." Said Li Ming, R & D director of product modeling and simulation R & D Department of Siemens China Research Institute

and in her view, AI's potential in the industrial field does not stop at analysis and evaluation, but also shows its ability in fault diagnosis, predictive maintenance and so on.

"if you only know the fault but can't judge the cause, you can't prevent it in advance, which is not really helping the workshop to complete the improvement work."

there is no doubt that the digital upgrading of old factories is a typical example, but it is not an example

Siemens has a large and diverse business and product system. In the past few decades, Siemens has carried out a series of explorations on artificial intelligence technology and applied it to complex image recognition of CT and MRI result analysis, industrial systems such as gas turbines and wind farms, copper price prediction and electric capacity utilization expectation, as well as physical autonomous systems for collaborative, adaptive and flexible production in industry 4.0

at the same time, Siemens' continuous investment in AI related technologies and talents is also obvious to all

in fiscal year 2017, Siemens' investment in related businesses was about 5.2 billion euros, more than 4.7 billion euros in fiscal year 2016. In fiscal year 2018, Siemens' R & D investment plans to increase by 450million euros from the current level

in addition, Siemens has established R & D branches in Beijing, Shanghai, Suzhou, Nanjing, Wuhan, Wuxi, Qingdao and other cities, as well as intelligent manufacturing innovation centers in Qingdao and Chengdu. As of fiscal year 2017, Siemens' global R & D personnel have reached about 40000

recently, Norbert Gaus, senior vice president of Siemens worldwide, was interviewed by machine power, restoring the unique thinking and survival rules of this old industrial enterprise in the tide of artificial intelligence

Norbert Gaus, senior vice president of Siemens worldwide

the following is the interview record. The machine's ability has been sorted out without changing its original intention

from the initial interconnection to the later mobile interconnection, to the later artificial intelligence technology, and now the hot edge computing and blockchain, what are the technical opportunities for Siemens to seize

this involves our technology department

Siemens has a very wide portfolio of products and technologies. Combined with technology categories, the company has defined 14 core technologies

the 14 core technologies defined by Siemens

in general, they are developed through digital means

digitalization is the foundation. Products should be interconnected and intelligent, which is also one of the aspects we want to promote in the development of technology. Let us know what interconnection and intelligence mean for the safety and reliability of the whole life cycle of field equipment

we have many working groups engaged in this research, one of which is to interconnect the equipment of old factories, and the other is to make one or more systems intelligent, which will realize real-time interaction and communication between field devices and systems in the future. It is the technology involved in the field level to make the field equipment optimize independently

the second is our information technology, which is used to help customers and ourselves in product design, manufacturing process design, factory automation and operation service automation

we have various tools to build models and generate data with these models. Mindsphere (Siemens' cloud based open IOT operating system) is used

mindsphere provides interconnection for devices to manage, implement functions or implement models for devices; At the data side and the model side, the product lifecycle management tools are also connected, providing an ecosystem for developing applications using data. These applications come from Siemens, as well as from our partners, customers and suppliers

in the context of this ecosystem, there are some very popular technologies, such as blockchain, IOT, interconnection devices or edge devices, artificial intelligence, simulation and digital twins

the last aspect is information security. The interconnection of millions of devices will produce a large amount of data. Only by ensuring the security of network information can the reliability of devices be realized

for us, these technical directions are not only an exciting topic, but also a highly demanding work content

what success stories can Siemens share about the application of artificial intelligence in the industrial field

in the industrial field, AI is usually applied to maintenance and service

the first case is to predict the maintenance time of industrial equipment

many times, we need to extract data from key components such as trains and turbines to predict failures caused by wear and tear, which involves the calculation of maintenance intervals

for example, once the gas turbine breaks down, it will take us and our customers a long time to repair, and the cost is very high

therefore, we need to dig up and push data in order to arrange maintenance at a convenient time for customers, so as to save a lot of costs and realize system optimization

on some occasions, we will sign maintenance contracts with customers to provide them with regular maintenance services. Our customers don't care about how you achieve maintenance, only whether the equipment can work normally within its life cycle

of course, we can ensure the normal operation of the equipment by adding spare parts and sending more engineers. The customer doesn't care, but we do. What we want is to pay the least cost to complete the task

in this way, the problem is transferred to ourselves. We need to study how to optimize the working conditions by retrieving data, and predict more accurately where and when problems will occur on the spot

another case is in the electricity industry

in the electricity, you may detect a fault, but you can't know where the fault is, so we need to accurately locate the fault

the more accurate the positioning is, the less the maintenance cost will be

traditional positioning methods rely on manual work, and they will complete the positioning task by evaluating the data. Now we use artificial intelligence to train neural networks, which can improve the accuracy of localization by 20%

this can not only save customers a lot of costs, but also has the advantage that the positioning ability does not rely on a back-end supercomputing to locate, but can be achieved as long as the neural network is deployed on site

in addition, in addition to maintenance services, we will also use artificial intelligence technology to optimize operations, such as machine tools. We are now also using AI technology to support product design

it can be said that our AI is widely used, and now it runs through all stages of the product life cycle, covering design and manufacturing, operation, service, maintenance and other aspects

some people think that the breakthrough direction of artificial intelligence in the industrial field lies in the deep integration of statistical models and knowledge and mechanism models in the field. How do you think of this view

I agree very much

industrial data is very different from data in other environments. Compared with the commercial and medical fields, the amount of data in the industrial field is much smaller

we don't want to get a lot of training data, and there are a lot of complex unstructured and unlabeled data in the industrial field, so we will use simulation models

but sometimes, the design of simulation model is not optimal, so it is difficult to complete this process, so we can only rely on the way that traditional domain experts and artificial intelligence experts work together

now, we will not use the training data alone, but will comprehensively use the data of adjustable surfaces in the experimental space of all parties in the process of product design, manufacturing and operation, make them scene and background, so as to build a knowledge map, and then automate the accumulation of domain knowledge, and then combine it with machine learning and neural networks, so as to overcome the bottleneck of insufficient data in the industrial field

of course, when we say insufficient data, we mean insufficient fault data

indeed, in the industrial field, the number of valuable abnormal and fault samples is relatively scarce. How can Siemens solve the asymmetry and imbalance of data samples? How to maximize the value of industrial data

we will solve this problem in two ways

one is to use simulation models to generate data, but simulation and now digital twins may not be able to generate data that can be used for machine learning

another way is also the main channel, which is to use domain knowledge to customize the design of the network according to data needs, so that we no longer need so much data. After all, the steam turbine can't have a million failures. In this case, the amount of data is enough, but we have bigger problems to solve

moreover, we not only look at machine data and sensor data, but also look at service reports, manufacturing reports, etc. to form a context and context. In this way, we can integrate machine data with other data and conduct comprehensive evaluation, which makes up for the lack of machine data

since 2001, Siemens has acquired more than 20 companies related to industrial software. Can we understand that Siemens is becoming a software platform company in the industrial manufacturing industry? Based on this, Siemens compares with Microsoft and other software companies in technical research