Yechao electric drive technology (Shanghai) co. LTD

National service hotline:400-820-4288

Address (headquarters) : 15 / f, xietong technology building, 4671 caoan road, jiading district, Shanghai

(factory) : 4645 cao 'an road, jiading district, Shanghai

Company email:


©yechao electric drive technology (Shanghai) co. LTD ICP18048149
Smart manufacturing's next frontier: industrial intelligence

Smart manufacturing's next frontier: industrial intelligence

2018/09/17 18:58
Page view
  Industry is divided into process industry and discrete industry. The biggest differences between the two are the degree of automation of production, the availability of data and the complexity of industry, while the most common feature is that each scenario requires different demands, and entering any segment requires deep enough industry knowhow and upstream and downstream resource integration capabilities.
  Intelligence can be understood as digitization and the AI built on it. Starting from the production line automation, the multi-source heterogeneous industrial data is collected, transferred, analyzed and helped to form decision-making and control, and the end-to-end solution forms the typical portrait of the current industry player.
Why industrial intelligence?
1, the blue ocean
  The total GDP of industry, especially manufacturing, is much higher than that of retail, finance and construction. In fact, the effective data volume generated by the industrial field every day is equal to that of Internet companies such as BAT, and the data volume generated by a large-scale factory can even reach billions to billions.
2, barriers,
  Although the industrial scene produces high frequency and large amounts of data every day, the large amount of raw data itself does not have direct significance, and it is possible to generate large delay and occupy large amounts of bandwidth. We not only need to do real-time monitoring and analysis in some scenarios, but also need to collect more data in the cloud to do more and more long-term economic benefits and value analysis, which is the value of cloud computing. And cloud computing + edge computing, which is a finer granularity and more complex architecture than the traditional consumer Internet, also means higher barriers.
3, a turning point
  The Internet has a logic called Copy to China. The large-scale data application and platform architecture has undergone sufficient verification and evolution in the industries such as finance and telecom, and the catalytic effect of made in China 2025 on the policy side constitutes the prerequisite for the establishment of the inflection point.
The player portrait of industrial intelligence
  Users at this stage need not a single product but an end-to-end solution. A qualified industrial intelligence company should be able to construct the overall solution.
  First of all, the user demand is always the first, the technology that does not meet the demand is false proposition. In addition, a good solution starts with a perfect architecture. For the industrial scenario, from the integration of internal and external multi-source data, to the platform architecture of cloud + terminal, the establishment of knowledge base, the selection of appropriate models, and to the reverse decision and control, only complete integration can form a closed loop.
  As a whole, industrial intelligence presents a horizontal (overall architecture)+N vertical (multiple subdivided industries) pattern.
Path selection for industrial intelligence
  For big B customers in the industrial sector, what is needed at this stage is not a single product but an end-to-end solution. This, though the status quo, is in fact the ultimate goal of industrial entrepreneurs. Path selection, however, is important.
  As for the development path, the mainstream of the industry thinks that automation -(digitization)- informationization - intelligence is the reasonable sequence of industrial users' advancement, and the former stage is the necessary condition for the latter stage to start. Therefore, for a long time, domestic enterprises in the field of industrial intelligence only focus on opportunities in the field of automation, and even equate industrial intelligence with robotics or industrial automation. From a large number of practices on the user site, there is a significant sequence in these stages, but at the same time, cross penetration and iteration are carried out.
  To be specific, most customers in the discrete manufacturing industry do not have enough automation, so it is preferred to complete production line automation. Some manufacturers use industrial Ethernet and board card to realize equipment interconnection and connect equipment level data. After MES feedback to the platform level, they realize preliminary material connection without replacing the original industrial control equipment. The user acceptability is high, and the performance increases several times every year, showing a clear trend. This kind of mode can be called system integration with M2M devices as the core.
  Further demands come from the super-large head customers of the discrete manufacturing industry and the vast majority of customers in the process manufacturing industry. Due to the high degree of automation in the production line itself, we have observed that such customers also have a high degree of acceptance of informatization.
  In addition, a group of manufacturers can directly enter from the top-level design, serve users with industrial big data platform or scene AI model at the platform layer, and solve business problems in real time. In turn, in the data acquisition layer, some data imperfect local installation of sensors, installed intelligent detection equipment, even to do a small section of production line integration, and so on. In this type of model, the user acceptance tends to be higher, which means the project premium tends to be higher. We can call this system integration with data application as the core.
  Therefore, we can see three development paths, facing different customers, different scenarios and different development stages, with different path choices:
1. System integration with production line automation as the core;
2. System integration with M2M equipment material connection as the core;
3. System integration with data application as the core.
  Of course, they all end up providing the user with an overall solution, with the core of satisfying the user's needs.
Industrial intelligence: industrial big data
1. First of all, where is the data?
  One type is management data: structured SQL data, such as product attributes, process, production, procurement, order, service and other data, which usually comes from ERP, SCM, PLM and even MES systems of enterprises.
  Another kind is the machine running and IoT data: in the majority with unstructured, streaming data, such as the working condition of equipment (pressure, temperature, vibration, stress, etc.), audio and video, log data such as text, this kind of general data acquisition from equipment PLC, SCADA and some of the sensors, the data quantity is large, high sampling frequency, need to do some pretreatment combined with edge calculation in the local.
  Generally speaking, due to the fragmentation and dispersion scenario, industry itself has a large amount of data, multi-source, heterogeneous and high real-time demand, and with the coming 28 billion gradually access equipment, these features will further strengthen, this is one of the core difficulties of data outside services, and the Internet data not only order of magnitude larger, different structures, the application is also completely different.
2. Secondly, based on these industrial data, what services should the platform layer provide?
  Complete protocol analysis: data collection should first complete the opening of industrial protocols. Taking the application layer protocol as an example, EtherNet/IP and PROFINET had the largest market share, followed by EtherCAT, modbus-tcp and EtherNetPOWERLINK.
  Standardized data integration: the collected data should be unified master data management. The first step is to establish standards. In general, it is important that we first use ISO or other industry standards to develop uniform coding, structure, flow, and attributes to ensure data consistency.
  In the process of project implementation, it is also important to gradually accumulate industry knowledge base, appropriate algorithm components and related mechanism models. This is a key step from data standard to business standardization, and lays a foundation for realizing real product level micro-service.
  Strong PaaS support: the specificity of the industrial data itself leads to the need for the platform to have strong mid-level support. We take the time series database as an example, which is the typical variety of equipment operating condition and sensor data. This kind of data is high in frequency and large in volume. It is processed by traditional relational database, and all values need to be pulled out every time for calculation. The throughput is great and the performance is poor. Therefore, a highly compressed and high-performance sequential database is one of the necessary capabilities of the platform layer.
3. Finally, what applications should we make?
  Equipment level: quality control. In the era of industrial intelligence, if we are able to collect real-time data where appropriate, combined with the mechanism model of the equipment is applicable to, it is possible to use machine learning methods to dig out the link between the product quality and critical data or cause and effect, and it is possible to realize real-time online quality control, and fault early warning, if the frequency of the data can perfect envelope on the formation process, we are likely to achieve maximum efficiency.
  Factory level: planned production scheduling. The ultimate goal of industrial intelligence is to realize mass customization, namely C2M. The goal of this problem is to achieve the optimal production capacity at that time and locally. The constraints come from the production line equipment, personnel, product attributes, supply chain data and so on. Through the learning and training of historical data, it is not difficult to form a better prediction model.
  This model can be dynamically adjusted according to the real-time data of production lines and factories, so as to help enterprises achieve accurate control and maximize economic benefits. In the foreseeable future, as data integrity and reliability become increasingly high and scenarios become increasingly rich, there will be quite a number of priority enterprises on the data application level that will help industrial users reduce costs, improve efficiency and solve real business problems.
Next article: