Wen Lisheng: Professor Of Donghua University Explains Why "Digital Transformation" Is Not "Digital Transformation"

I. why the transformation of garment manufacturing industry is called "digital transformation"

Not "data transformation"

Digitization refers to turning many complex and difficult information into binary code that can be processed by computer in a certain way, so that the machine can read and process information, forming a digital twin in the computer. Digitization is also called "biting" because it can be generalized to represent everything with "0, 1" bit values.

And datalization is to analyze the digitized information to make it organized, and provide powerful data support for decision-making through intelligent and multidimensional analysis and query backtracking. The expression of data is digital, so the concept of data is greater than the concept of digital. Therefore, it can be said that digitization brings about digitalization. Digitalization is a direction of digitization process, but digitization can not replace digitalization. Digitalization can be applied to digitalization. Therefore, the transformation of clothing enterprises is called "digital transformation".

2. Why is data the asset of garment intelligent manufacturing

The concept of data assets is evolved from the concepts of information resources and data resources. In the 1970s, people regarded information as human, material, financial and natural resources; In the 1990s, with the digital transformation of the government and enterprises, data resources came into being. Of course, it was only after the number of information became data and then aggregated to a certain scale that it was called data resources; At the beginning of the 21st century, due to the rise and application of big data technology, data assets have been produced. The development of data management, data application and digital economy makes data assets popular. Therefore, a data asset is a data set, including data rights (ownership, use right and exploration right), and it is valuable, measurable, readable and tradable.

On April 9, 2020, the Central Committee of the Communist Party of China and the State Council promulgated the opinions on building a more perfect market-oriented allocation system and mechanism of factors of production, which clearly pointed out that production factors include land, labor force, capital, technology and data. These five elements are also assets, namely land assets, labor assets, capital assets, technical assets and data assets.

In recent years, the annual growth rate of global data is more than 20%. The growth rate of data assets far exceeds that of technology, manpower, capital and land assets. According to the statistics released by international data company IDC, in 2018, China's data production accounted for about 23% of the world's, which was only about 30% lower than that of EMEA (Europe, Middle East and Africa). China is one of the world's top data resource countries and global data center. As can be seen from Figure 1, apjxc represents Asia Pacific and Japan, accounting for 18%.

Figure 1 global data distribution

According to the "data age 2025" released by International Data Corporation (IDC), the global annual data generation will increase from 33zb (1zb = 1.1 trillion bytes) in 2018 to 175zb, equivalent to 491eb (1eb = 1.1529e + 18 bytes) per day. According to IDC forecast, the global data volume will reach 44zb in 2020 and 19000 ZB in 2035. See Figure 2.

Figure 2: Trend of global data increment

As can be seen from the above figure, the trend of global data increment conforms to the new Moore's law. In 2004, the global total data was 30eb, and in 2005 it was 50eb. By 2015, the data has been growing rapidly, reaching an astonishing 7900eb and reaching 35000eb in 2020. This means that the data generated in the past two years is 90% of all data generated by human civilization as a whole.

This means an increase in global data assets. As the owner of capital assets, only by using data assets can the value of data be reflected. The value of data is related to the following factors: high frequency of data application and more aggregation, the higher the value of data; On the contrary, the longer the data time and sharing, the lower the value of data.

For the garment manufacturing industry, when, what kind of data and how to transmit them to people and machines are the keys to optimize the scientificity, real-time and effectiveness of the most reasonable optimization of manufacturing. The essence of fully automatic flow of data is to drive the operation of enterprises with data. Figure 3 illustrates this process.

Figure 3 data flow in enterprise operation (source: Network)

As an asset, data has its unique characteristics: tradable, with rich resources, frequent updates, diverse categories and other information value; It can be used in circulation and has the natural value-added attribute of continuous generation and continuous creation of value; It has the multi-dimensional nature of meeting human survival and development, zero cost unlimited sharing and high use value.

The implementation of Intelligent Manufacturing in China's manufacturing industry needs to be combined with the three-step strategy of enterprise informatization, digitization and intellectualization. Enterprise informatization is the basis of data and knowledge collection, and also the starting stage of intelligent decision-making; In the digital transformation stage of enterprises, it is necessary to collect, clean, analyze and process the data knowledge in various fields of the enterprise to form usable and valuable big data assets; The stage of enterprise intelligence is to make full use of big data assets to realize the intelligence of enterprise business, that is to realize the intelligent manufacturing of enterprises. See Figure 4. Thus, in the process of realizing the three-step strategy of intelligent manufacturing, the value of data is constantly increasing, that is, only with data can there be big data, with big data can there be artificial intelligence, and with artificial intelligence can intelligent manufacturing be realized.

Figure 4 three step strategy of enterprise informatization, digitalization and intellectualization

The application of data assets is the key to the transformation of manufacturing industry to intelligent manufacturing. Figure 5 shows the market size forecast of global big data from 2011 to 2027 (US $100 million). China's big data industry ecological alliance and "white paper on the development of China's big data industry 2020" released by China's big data industry ecological alliance predict that by 2020, the overall market size of China's big data industry will reach 667.02 billion yuan, and will exceed trillion yuan by 2022. According to this data, it is further predicted that the overall market size of China's big data industry will reach about 1756.8 billion yuan by 2025 (Figure 6).

Figure 5 forecast of global big data market scale from 2011 to 2027

Figure 6 China's big data industry scale from 2020 to 2025

It can be seen that the development of big data means that the development of intelligent manufacturing of enterprises is combined with big data technology and artificial intelligence technology to practice.

How to make good use of data assets in garment manufacturing industry

Big data technology includes data acquisition, storage, storage, search, sharing, transmission, analysis and visualization. The following three aspects are the most important aspects:

Big data of manufacturing industry mainly includes product data, operation data, value chain data and external data. Among them, industrial big data mainly comes from machine equipment data, industrial information data and industrial chain related data (see Figure 7), and of course, it also includes massive key value data, document data, interface data, video data, image data, audio data, etc.

The commonly used data acquisition methods are different types of industrial sensors and RFID radio frequency technology.

Figure 7 big data sources of manufacturing industry

The collected data need to be analyzed. Data analysis and processing can be traced back to the standard model "CRISP-DM" put forward by sig of EU in 1999. The process of this standard is as follows: business understanding - clear objectives, analyzing requirements; Data understanding: collection, description, exploration, inspection and data; Data preparation: select, clean, construct, integrate and format data; Build model: select modeling technology, parameter tuning, test plan, build model; Evaluation model: comprehensive evaluation of the model, evaluation results and review process; Results deployment analysis results, implementation and application of the scheme (Fig. 8).

Figure 8 big data analysis standard process CRISP-DM

The significance of big data lies in the application and analysis of data. In the clothing manufacturing industry, it refers to:

(1) processing equipment status data analysis

The output and display of the analyzed data can let the staff know the real-time status and processing technology data in the production process at the first time, so as to make quick, timely and scientific countermeasures.

(2) optimization of manufacturing process data

It is mainly manifested in two aspects: 1) equipment process parameter monitoring, real-time comparison and control of collected equipment process parameters, such as temperature and pressure, with the set standard parameters, so as to realize real-time, dynamic and strict process control of the production process, and ensure the stability of product quality; ② Process improvement and optimization, the main process parameters of the manufacturing process and the qualified rate of finished products are comprehensively analyzed to facilitate the improvement and optimization of the processing technology.

(3) production process traceability

Through the process data of product processing and manufacturing, the product manufacturing history can be traced to achieve the purpose of problem recurrence and quality traceability.

There are three problems in data asset security in garment manufacturing industry. First, data assets are complex and data risk cannot be quantified; Second, the risk of data security is high, and there is no dynamic data authority control in the use of data; Third, data security risks are wide. How to protect data security is very important. Here are some data security solutions:

(1) Dell's cyber recovery vault. It has air gap gate isolation mechanism and copy locking mechanism to block the possibility of blackmail software contact, thus greatly reducing the probability of virus infection backup data. After the backup data is stored on the storage device of the production side, the replication link is established with the storage device of the cyber recovery vault area, and the backup data is copied from the production center to the cyber recovery vault area (data isolation storage library) through the internal network and special interface. Cyber recovery vault area is "invisible" to network attackers, blocking the possibility of blackmail software infecting backup data. Once the data synchronization is completed, the air gap gateway can be closed, and the data access path is disabled. At the same time, in order to prevent the backup files from being maliciously deleted, the system can lock the data in the isolated storage library, so as to ensure that the backup data copies can not be encrypted, tampered with and deleted.

(2) Google cloud platform. Enterprises can transition to cloud platform when data protection is guaranteed. Google cloud platform has completed the huge transformation of the platform to openness and cloud computing, which has brought more powerful data assets and better and more secure data analysis capabilities for enterprises. Google cloud platform open infrastructure allows customers to choose the most suitable path to the cloud for their own business. With Google cloud's infrastructure, data and AI machine learning solutions, it's easy to put data into the cloud and analyze it.

(3) dawning big data platform with security management and control system. Dawning big data platform is an intelligent analysis and processing solution for massive data combined with big data technology, which can help enterprise users quickly build an efficient, intelligent and easy-to-use integrated big data system and mine data value. Dawning big data analysis platform adopts the technical architecture of fusion, realizes in-depth storage fusion, computing integration, scheduling integration, multi-source data fusion and business process integration, and constructs an overall system of systematic integration (Fig. 9).

Figure 9 dawning big data platform architecture

(4) innovate Qizhi's "Orion automatic machine learning (automl) platform". The platform mainly helps customers make good use of data assets, improve data decision-making ability, and enable customers to run business. The platform is a series of products and solutions that are in line with the future data intelligence paradigm, with flexible selection and configuration of three-tier structure, facing enterprise customers and focusing on privatization deployment. Orion data intelligent engine mainly includes three product units: ① Orion IRC -- intelligent resource scheduling management, providing computing resource management and data asset map; ② Orion DAC --- intelligent data fusion management, supporting data dynamic fusion and realizing data supply chain; ③ Orion AML --- automatic machine learning, intelligent decision-making based on data.

(5) talking data security island platform. Based on the industry-leading trusted data computing technology, talkingdata "safe island" builds a secure and compliant one-stop platform for multi-party data accommodation, providing a new mode of data industrialization application and value release for different business scenarios. In order to smoothly apply the data services and data capabilities of talkingdata to customers, the security Island solution developed is actually a secure computing platform. Its starting point is to enable customers to realize secure and compliant data value exchange on the platform, isolate privacy issues of both parties, and help customers make up for the lack of data capability, Really unleash the value of data.

4. Application examples of data assets in garment intelligent manufacturing

We talked about the relationship between data and intelligent manufacturing. To realize intelligent manufacturing, it is necessary to have real artificial intelligence, which relies on big data, advanced algorithm model and super large computing power (such as CPU / GPU / TPU) (FIG. 10).

Figure 10. Related technologies of artificial intelligence landing

Intelligent design, intelligent production, intelligent management and integrated optimization constitute the intelligent manufacturing of clothing, and the essence of big data assets is to realize these intelligent manufacturing. Here is an example:

Clothing enterprises rely on big data strategy to broaden the breadth and depth of survey data in the clothing industry. Factors such as clothing market composition, market segmentation characteristics, consumer demand and competitor status can be obtained from big data. Through the systematic information data collection, management and analysis, solutions and suggestions are obtained to ensure the uniqueness of enterprise products in the market positioning, Improve the acceptance of enterprise products in the market. The clothing market will analyze the regional population distribution, consumption level, product cognition, customer consumption habits, work consumption preferences and other factors given by big data to obtain the market positioning and provide data support for enterprises to enter or develop the clothing market.

For example, according to the data collected, about 4.74 million pairs of autumn trousers can be sold in one day during the double-11 period in 2019. In this case, an in-depth investigation has been conducted on the Chinese autumn pants Market with sales of more than 50 million pieces. It is predicted that the investment in autumn pants with prices within 100-500 yuan should be strengthened in 2020 (FIG. 11).

Figure 11. Price distribution of autumn trousers (source: CDA)

Our current fashion design is still using the traditional fashion CAD which has not yet been intelligent. If we use the intelligent CAD, it can replace the designer. For example, in 2019, MIT uses the Gan model in AI technology for fashion design. The so-called Gan is a deep learning model, its full name is "generative offensive networks", and its Chinese name is "generative adversarial network". Gan designs two neural networks, through one generation, one judgment to carry on the game. For example, in the dress design, the researchers collected about 5000 big data of the past dress fashion styles. One is the generation model of GaN and the other is the judgment model of Gan's relative resistance. After a few days of training, we can get the new design dress fashion style. See Figure 12, and the old clothing image big data is on the left, On the right is a new fashion created by AI technology using big data.

Figure 12 design and manufacture of clothing with Gan

Figure 13 application of big data technology in optimization of garment processing parameters

The parameters of sewing machine affect its operation, such as sewing speed, sewing tension, stitch form, stitch size, etc. Parameters need to be set in advance before sewing, and the optimal sewing parameters can be obtained through parameter big data and machine vision technology, machine learning and deep learning technology in artificial intelligence, so as to achieve the best sewing quality (FIG. 13).