On August 2, 2017 China's big data industry ecological conference was held in Beijing.
Conference sponsored by the China electronic information industry development research institute, China's big data industry ecological union joint in the venture investment co., LTD., China software testing center, sadie, a think-tank, ccid consulting co., LTD., and undertaken by magazine of the software and integrated circuit.
The conference attracted government leaders, industry experts, China and the United States nearly 20 big data companies and venture capital institutions gathered in the capital, to discuss the big data industry development trend, promote technological innovation, help enterprises to grow.
American professor, university of Cincinnati college of the national science foundation (NSF) co-operative, director of the center for li jie, intelligent maintenance system as expert committee, director of the alliance members were invited to attend the meeting, and make decision, entitled: industrial big data first half and the rest of the speech, to share how to better use of the industrial data, solve problems and to avoid thinking into value creation thinking of the problem.

　   Li jie, professor of lectures at the university of Cincinnati and director of the national science foundation's (NSF) intelligent maintenance system (IMS) production and learning center

Hello, everyone. I'm glad to share it with you.
Today my topic is the first half of the industrial big data and the second half of the year.
How did the first half of life develop?
This topic is also a small summary of my personal experience and practice.
First, the first half of my work.
I was in the United States for 37 years, and as early as 1983, the United States began to do automobile automation production lines. I worked in industrial automation and robotics.
Later to enter the United states national science foundation (NSF), and then the United Technologies (United Technologies Research Center UTRC) served as director of Research and development in the pratt &whitney engines, carrier air conditioning, audi development technology, elevator, etc.
Then I went back to school to be a professor, and started the intelligent maintenance system center (IMS).
IMS center was established in 2000, has been now for 17 years, the world has more than 90 corporate partners, now do the biggest project is working with denso "Dan - To - Tsu" can be translated as "no one can and factory".
The goal is to have 1 million parts to make only one quality, so no other factory in the world can reach that level.
The plane also took off a million times to have a crash, and the safety factor was so high that it was supported by technology such as big industrial data analysis, failure prediction and health management.
Now we have cooperation with the big gold air conditioner, huawei, Chinese ship, medium high speed rail, and mazak - the world's largest machine tool factory.

Let's start with the development of industry.
Many countries in the development industry started by raising productivity.
There are five phases, the first phase is full practice, Japanese is Kaizen, and Chinese is called Kaizen.
Every day completes, rectify, clean, clean, does the overall standardization the improvement of the renewal.
The second phase is data, Toyota's first lean and GE "6-sigma", which was done in the 1980s and 1990s.
The third stage is the predictive modeling analysis, when the us was making the transition in 2000 to solve the problem of data layer to information layer.
The fourth stage is called the knowledge layer, and now all we have to do is turn data into systems that can support decisions.
The fifth stage is the highest level, which can produce knowledge and decision-making autonomously, without the need to control it, such as driverless cars.
But unpiloted is not the point, and carefree driving is, in other words, driving without worrying about the road ahead.
Potholes, if any, have a kilometre ahead before the car pass here, through the established relationships with my GPS sensor, next time when I walk the same route, have not yet open to the place, the car will receive a reminder - please pay attention to the pit one kilometers ahead.
My car is Shared with other cars. They are not yet on, and they will be reminded that there is a hole in the front of the car.
This is the value conversion of what we call data, from performance optimization to avoiding risk and worry.

In "industrial big data" the book I talked a lot, the first half from visible to solve the problem of, since productivity to find a big problem, for example, problem to be big enough to let us do investment data, thus to solve big problems, this is my first half done.
I didn't do it for the rest of my life, because solving the problem was not the goal.
The goal is to make the problem disappear and not even make it happen.
This is what I call the problem of recessive, even the problem that the client is not aware of, to find out the value, which is the big value.
The problem has not yet emerged, and it has been avoided, which is of great value.
So let's see how we can do a lot of value.
From 1984 to 1987, I worked on the general motors automated production line, and the Manufacturing Automation Protocol was the equivalent of today's IoT.
In another 87 years gm also invested in the hughes satellite, which was developed in 1992.
Anjstar has been the world's first industrial Internet forerunner - all the sensors in the car are connected directly to the human hand.
When the car was in a car accident, it immediately knew that you would not answer, and he immediately sent an ambulance and the police to the scene, which was realized in 1992.
When I was the research and development director of united technologies (UTRC) in 1998-2000, Otis elevator was our unit.
We can know from the use of the elevator which elevator will be down tomorrow, in 1997.
In 1999, GE medical's magnetic resonance ultrasound was able to quickly diagnose and send people to maintain the problem before the problem occurred or before the problem occurred, reducing the failure rate of equipment and personnel expenses.
In 1990, GE medical magnetic resonance device used in hospitals, there are a lot of problems, the doctor found equipment problems will make a phone call to the repair, then a week on the average, there are one thousand telephone service, only 41% is not can solve in the past;
In 99, can not sent before they know what parts you want to change, even a lot of problems can be through the remote diagnosis, only 25% of the questions have to be sent in the past, thus saves 70% of unnecessary human, it is simply a big problem.
By 2004, the data from the medical side, the product data, had been transferred to the patient's data, and that was the beginning of the real industrial big data.
For example, in molecular medicine, where you can use a developer to find out where the body might be, that's the most valuable part.
For example, airplanes, for all the parts of the engine, can anticipate what's going to happen.
When the IMS center was founded in 2001, our idea was how to turn the big problems of the first half into the great value of the second half.
The data and historical data of the sensor are taken in, and detailed and in-depth analysis is made, which brings out the essence of big data.
To find out the hidden problem, that is, the problem that hasn't happened yet, we can solve and even avoid, and create value.
This is the second half of the industrial big data.
We've done a lot of projects, including John Deere, GE engine, alstom high iron, Goodyear tire, Intel, procter & gamble, etc.
It was the first half of the life to do it, solve the big problem - intelligent maintenance, and make intelligent prediction in the second half of the life, and finally achieve the carefree system and great value.
The worry-free system is needed for any future system, such as the worry-free, worry-free factory.
So what to do?
Let's start with Intel semiconductors.
China is now building 26 8-inch wafers and 12-inch wafers, the next source of economic growth and growth in China.
It takes about $2 billion to $4.5 billion to build a factory.
There are very few opportunities for people in wafers, and people need to do simple things that are truly automated, so the data is the most valuable.
Intel found me in 2000 and wanted to analyze and predict the data.
Because the equipment is expensive, they want to anticipate when the equipment will fail, and the relevance to quality, speed, and so on.
This was the focus of our IMS in 2000, which was to deliver predictive and preventative high end semiconductor equipment.
China's semiconductor is on the rise today, many of its high-end components are nanowires, China is now a dozen nanometers, and the United States is now five nanometers, three nanometers.
The next example is p&g.
Procter & gamble's diaper production line is often down.
If you want this production line to reduce downtime and increase usage, you need to use predictive models to predict.
We worked with procter & gamble's diaper production line at that time to be able to do it without downtime.
To help p&g save $450 million a year on unnecessary waste, this is p&g statistics.
This efficient input output ratio is the transformation from big problem to big value.
From 2005 to 2007, komatsu cooperated with us to predict which parts will break down and optimize construction site.
Now komatsu's projects around the world can be scanned with big data, built on the foundation of the site, and then worked together to assign an indicator to the excavator.
This is great value, not selling excavators, but selling site management services, which are of high value.

This is today's Fried egg model, which leads from big issues to big value orientation.
Egg yolks are a big problem and protein is of great value.
We're starting with big data, but it's definitely not the goal, to be valuable, to play the best part, from problems to data to experience.
Experience can be inherited, but it can't be passed on.
And data can be inherited because it's logical.
Finally, do your best and turn your experience into facts.
The next economic competition is the evidence - based economy, not the social network we are talking about now, or the experience - based economy, which is like the small fortune that we consume, which is experience, which is not the big industrial data.
Industry big data is evidence - when is bad, when to avoid, when no problem, best not to worry.
Industrial big data wants focus, convergence.
And the traditional big data is diverging, who is the customer, the directional push advertisement, these are the opportunity orientation, not the value orientation of precision.
The GE engine saves about 1 percent of its oil money by remote monitoring, and it saves a lot of money in 15 years, which is called protein.
China's case, the ship is egg yolks, for each shipping lines parametric modeling, such as weather, waves, wind form waves with gas-guzzling relation model, according to the result of analysis to optimize routes and speed, can save 5% of the gas.

Let's talk about artificial intelligence.
There are four technologies, DT data processing technology, PT platform interface technology, AT analysis software technology and OT operation technology, which are four different operating levels.
There are three main characteristics of industrial data: separability, low quality and background, which are also the basis of industrial data.
So the DT data processing technology to solve these problems is very important, but use DT technology need some background information, requires knowledge of engine, electronics manufacturing and other fields to do;
AT analysis technology, computer science and artificial intelligence are used in computing.
PT platform technology can be Shared after knowledge generation, and also feedback to operation level OT.
We need to use artificial intelligence to strengthen industrial big data, because the data background in industrial big data is hard to identify.
Fan was moving, such as changes in the wind speed, air humidity is changing, the blade to the winter freezes, these people can't see, also don't know, need to pass the parameters of wind speed and power generation, building cluster, and slowly build up relevance, the classification, segmentation, decomposition, analysis, to share, share the six steps.

The future can be used, which speed?
With precision, with complexity, with uncertainty, with the speed of tools not exactly the same, the tools can't be easily solved.
In addition these tools will combine with the embedded becomes a trend, Intel in cooperation with us, to do some of the embedded technology, put all intelligent software in machine inside, even the cloud on the edge of the end.
In this case, I can set the data collection based on experience in FPGA, including GPU, TPU frame and so on.

Future intelligence must be Shared with each other, and that's what I'm going to talk about - memory management using CPS (information - physical systems).
All physical systems need to be able to establish the relationship between data, which cannot be done manually.
For example, WeChat connects to everyone, one person is smart, but three heads are better than one.
Integrating the source of Resource data, Relationship of Relationship data and the meaning of Reference data is more intelligent, so the foundation of society comes from human intelligence.
In the future, a smart rail transit system can be realized.
Concorde in moving, for example, each of the parts they can feel recession, change, bearing has a problem, can see the real-time dynamic problems, what is chip can work out the problem, can pull each carriage, see which car is the worst.
As a whole, there are more than 2,000 trains, times 8, and a total of more than 20,000 carriages, each of which can achieve these capabilities.
After One Belt And One Road goes out, the middle car can do egg yolks, or you can do protein, which is the future of CPS.
Finally I want to make a conclusion, we walk the first half of the past in the past ten years is the traditional "big issue" into a predictable problems and solve, ten years behind the big problem into the value of the "big".
To turn the worry-free system into a technology that guides the future, it has its own maintainability, immunity and toughness, which is the route of industrial big data over the next decade.
The challenge is high, but I hope that each of us with big data will be able to focus on the industry, which is the basis of any great power and the source of world wealth.
For example, in the hospital emergency room, the CPS can be realized, avoiding accidental death of the patient, saving a life to build a seven-stage floating cartu, and also a merit.
Finally, the ai that CPS brings out is artificial intelligence that transcends machines, because it is more inherited.
"Industrial big data", "from the big data to the intelligent manufacture", the new generation of industrial intelligent CPS, the four books "cloud industrial intelligent" as a reference for everyone, I share with you here today.
Thank you very much!