Unscramble the context of industrial big data

Unscramble the context of industrial big data

Industrial big data is different from big data and has its own unique characteristics. This paper focuses on the definition and category, source, characteristics, technology and application fields of industrial big data, and the problems it faces, and comprehensively analyzes all aspects of industrial big data, so that you can understand the context of industrial big data in an article!

Industrial Big Data refers to all kinds of data and related technologies and applications generated in the industrial field around the typical intelligent manufacturing model, from customer demand to sales, to order, planning, research and development, design, technology, manufacturing, procurement, supply, inventory, delivery and delivery, after-sales service, operation and maintenance, scrapping or recycling and remanufacturing. It takes product data as the core and greatly expands the scope of traditional industrial data.

-Big industrial data sources-

What we call industrial big data is not exactly the same as the data flowing in enterprise information software. From the perspective of industry, there are three main sources. The first category is business data related to enterprise operation, which comes from the field of enterprise information, including enterprise resource planning (ERP), product life cycle management (PLM), supply chain management (SCM), customer relationship management (CRM) and environmental management system (EMS).

The second category is the interconnection data of machinery and equipment, which mainly refers to the operation data of equipment, materials and product processing, such as working conditions and environmental parameters, which are transmitted in real time through MES system. At present, with the extensive application of smart devices, the amount of such data is growing fastest.

The third category is the external data of enterprises, including the use and operation data of products sold by industrial enterprises, as well as a large number of customers, suppliers, Internet and other data status.

-The characteristics of industrial big data-

The author interviewed Professor Li Jie, a well-known expert in the field of industrial big data and director of the Intelligent Maintenance System (IMS) Center of the National Science Foundation (NSF), on topics such as the characteristics of industrial big data and data-driven industrial value creation. He said: The biggest difference between industrial big data and Internet big data is that industrial big data is very purposeful, while Internet big data is more of a related mining and more divergent analysis.

In addition, there are differences in data characteristics and problems between them. Different from Internet big data, the core of analysis technology of industrial big data should solve the "3B" problem:

1) Under the surface-hidden, that is, we need to know the meaning behind it.

The most important difference between big data in industrial environment and big data on the Internet is the extraction of data features. Industrial big data focuses on the physical meaning behind features and the mechanism logic of association between features, while Internet big data tends to rely only on statistical tools to mine the association between attributes.

2) fragmentation-fragmentation, that is, it is necessary to avoid discontinuity and pay attention to timeliness.

Compared with the large amount of data on the Internet, industrial big data pays more attention to the integrity of data, that is, it requires the use of application-oriented samples as comprehensively as possible to cover all kinds of changes in industrial processes and ensure the comprehensiveness of information that can be extracted from data to reflect the real state of objects. Therefore, on the one hand, industrial big data needs to overcome the difficulties brought by data fragmentation in back-end analysis methods, and transform these data into useful information through feature extraction. On the other hand, it is necessary to start from the front-end design of data collection, formulate data standards based on value requirements, and then build a unified data environment in the platform of data and information circulation.

3) Poor quality-low quality, that is, data quality needs to be improved to meet low fault tolerance.

On the other hand, the source of data fragmentation defects also shows concern about data quality, that is, the quantity of data can not guarantee the quality of data, which may lead to low usability of data, because low-quality data may directly affect the analysis process, resulting in unusable results, but big data on the Internet is different. It can only mine and correlate the data itself, regardless of the meaning of the data itself, that is, what results are mined. Most typically, after the data mining of supermarket shopping habits, beer shelves can be placed opposite to diaper shelves, regardless of the logical relationship between them;

In other words, compared with big data on the Internet, it is usually not required how accurate the pushed results are. The fault tolerance rate of industrial big data for prediction and analysis results is much lower than that of Internet big data. When making predictions and decisions, Internet big data only considers whether the association between two attributes is statistically significant. When the sample size is large enough, the noise and differences between individuals can be ignored, so the accuracy of the prediction results will be greatly reduced. For example, when I think it is 70% meaningful to recommend a class A movie to a user, even if the user doesn't really like this kind of movie, it won't cause too serious consequences. However, in the industrial environment, if the analysis results are only given by statistical significance, even a mistake may have serious consequences.

—— Industrial Big Data Technology: Algorithms and Models—

The massive accumulation of industrial data does not mean direct commercial benefits, and there is a very critical channel-industrial big data technology. In recent years, many big data experts and industry experts are also arguing: whether the amount of data is more important or the big data algorithm is more important, both sides hold their own words. For example, Googole thinks that the amount of data is very important, even bluntly: more data is not as good as algorithms. This view is similar to "the more information, the closer to the truth" in our conscious cognition.

For example, in signal and noise (author NateSilver), one point of this book is that "more data means more noise. The signal is the truth, but the noise makes us farther and farther away from the truth. " Therefore, people need to establish effective algorithms and models to identify and identify what is the truth.

We don't discuss whether the data volume is more important or the algorithm model is more important here, but the effective use of industrial big data is definitely inseparable from the analysis technology of industrial big data.

—— Industrial big data application field (scenario)—

1.R&D design: it is mainly used to improve the R&D innovation ability, efficiency and quality of R&D personnel and support collaborative design, which is embodied in: (1), R&D design based on model and simulation; (2) Design based on product life cycle; (3) Design of integrating consumer feedback.

Second, the application in complex production process optimization: (1), industrial Internet of Things production line; (2), production quality control; (3), production planning and scheduling;

Third, the application in product demand forecasting

Fourthly, the application in industrial supply chain optimization.

—— The main problems in the development of industrial big data applications—

The White Paper on Industrial Big Data 20 17 pointed out that the research and application of industrial big data, product big data is the core, material big data is the means of realization, and integration is the foundation (business model, business and value-driven, key extraction and application). However, in practice, there are difficulties in these three aspects to varying degrees.

Cover of White Paper on Industrial Big Data Version 20 17

1, product big data: product big data is the root and core of industrial big data, but the industrial manufacturing field covers a wide range, with a large number of industries and products, and it is still growing. How to standardize the definition and classification of product big data and establish standardized, identifiable, traceable and locatable product big data will be the premise for the smooth application of industrial big data.

2.IOT access equipment: IOT data is a necessary means to realize the smooth flow of industrial big data. However, in practical industrial applications, industrial software and high-end IOT devices are not self-controllable, and high-end IOT access devices are not open to reading and writing, forming an island of device information and poor data circulation. Breaking through this constraint is the key to realizing industrial big data.

3. Information integration: The difficulty of integration lies in business driving, getting through key points and links, controlling product sources and equipment, and continuously optimizing.