Ie AI appeared as the representative of advanced technology by the public attention, the industry also recognized as the future direction of AI technology. But how AI industrial floor? It can bring real value to industrial companies? Integration of AI and industry will face challenges? These are the questions need to think about the moment. After the AI industry with parallel digital industry and general AI AI is essentially no difference, in fact, communications equipment, acquisition and calculation capabilities meet certain requirements, after the integration of these capabilities through software to analyze [ 123], industry specific issues arise automatically resolved. Industrial AI landing with the current implementation of the industrial enterprises digitized and there is not much conflict, many companies directly AI as part of the digital process. It also found that communication with the industry in the near future, when it will talk to explore the data collection and analysis algorithm, and then directly to the conclusion that the product uses artificial intelligence. Currently, the industry is also not clearly defined, or the amount of data required algorithm complexity in order to achieve what counted as artificial intelligence. Focused on the numerical modeling of systems engineering and artificial intelligence algorithms Mathworks to make a clearer classification , into the depth of learning, machine learning and enhanced learning three types of industrial applications belong to enhance learning, which is known to a particular target, relying on self-optimizing algorithm, keep close to the stated objectives through training, other judgment completely ignored. The latter mentioned energy efficiency management and predictive maintenance algorithms basically fall into this category, industrial image AI outlook remains fuzzy. Energy efficiency management, improve production, predictive maintenance, visual identity is the main industrial AI landing scenario, these applications are undoubtedly algorithms and data-intensive areas, mostly in the form of performance-based platforms, industrial automation and medium-sized enterprises in the past two years, without exception, made significant investments in software, cloud, three major aspects of communications technology, Schneider electric AVEVA, EcoStruxure Energy Machine advisor, Siemens Mindsphere, ABB Ability EDCS, PTC’s ThingWorx, etc., are all around, but this major application, the goal is a new method of cost reduction and efficiency sought in the realization of a comprehensive interconnection equipment, try to eliminate productionUncertainties materialize, production of reliable and controllable. A class of applications: energy management, and production improved from the past to the present algorithm AI, industrial equipment and not much change in the core module, or communication network linking ability is to enhance the most part, but the force is still considered far it is not used to “mining”, so a lot of data and AI algorithms are implanted into the cloud, and edge servers. It also created the industry’s hot clouds and edge server market. AI is the industrial energy management applications in the top grade, the cloud in the form of a multi-platform, by uploading collected consumption data, to be compared to the analysis, and then combined with production conditions automatically adjust electrical load, reduce unnecessary energy consumption, help plants save money, such applications can be described as currently the entire industrial applications of AI’s Golden Mile, there will be a considerable number of visual and integrated sensing products to this area, assisted AI decision-making. For users, optimizing the energy efficiency benefits are clear. Specific reasons are as follows: 1. Such multi-application software to upgrade the main, or even add an edge server acquisition equipment, upgrading should be relatively easy. 2. a high return on investment to HARTING “small box” data acquisition devices MICA, for example
, very few will be able to invest each month in exchange for 10% of energy consumption saving real money visible. 3. more to promote bundled services market, manufacturers of electrical products are basically delivered with such services, such as ABB, Schneider Electric’s product completion is very high, switchgear devices are generally bundled in energy management services, not only to enhance value-added products, can quickly Distribution. Of course, precisely because manufacturers layout, small factories will be more vulnerable in such applications, although the technology may difference is not large, but because fewer nodes collection, let AI in some cases does not look so smart, despite large factory also adopting an open architecture, but the practical application will encounter some kind of limit, the monomer product cost can be high, but the device is not into a big system synergies, which is a lot of technology-based companies failed to break out of the field in IIOT The reason, do not go in the market, product iteration or even backward. In addition to direct distribution management, energy management and some pan type applications, such as the Festo cylinder valve and relaxation Anwo digital solutions, to control opening and closing of the valve through the intelligent analysis, to improve the energy efficiency of the cylinder. Production and improve energy efficiency management features are very similar in this regard, Weather data collection, through analysis and comparison, and then adopt a different strategy for sustained productivity improvement, the difference is slightly larger human intervention, such applications require manufacturers to have a deep understanding of the industry’s production target, customization requirements also stronger, so the AI algorithms can be divided into many small variety, versatility in energy management will be weaker than others, such as Rockwell automation industrial thin client management platform Thin manager ?, and Schneider electric Transware? transparent factory kit They fall into this category. Class B Applications: Predictive Maintenance Predictive maintenance has long been proposed, but in the industrial field has been tepid. Compared to the actual energy efficiency gains from better management, based on the same algorithm as the core benefits of predictive maintenance is currently still difficult to assess, which also led to progress in this area slightly slower. Predictive maintenance will inevitably lead to some investment, but the investment is not convincing energy efficiency management so strong. The main problem here is that, AI algorithm must be more excellent than regular maintenance, must demonstrate commitment to the algorithm more economical than regular maintenance, but if the part is in accordance with the design life time regular maintenance or replacement costs is not very sensitive components, predictive maintenance of value will be discounted, and the production and improvement as solutions company must have sufficient grasp of the target industry, or in some of the more critical predictive maintenance project will look quite risky. Currently more successful practice of such applications are expensive for large-scale equipment, these behemoths is always in expensive replacement costs in human life and working environment, real-time understanding of their health maintenance and replacement in the best time to do very meet the needs of these industries. Shale gas development special vehicle energy storage device, the gearbox high-speed rail trains Control Engineering Copyright , bearing wind turbines, which are expensive and not easily replaced equipment is predictive maintenance real market force. To note here is that predictive maintenance is made largely for industrial robotics industry, Fanuc and Cisco co-developed ZDT zero downtime system, AI subsidiary of Yaskawa Electric AI Cube Inc for the development of the manufacturing site predictive maintenance programs have a smash hit, but with the sharp decline in industrial robots and robot parts cost, predictive industrial maintenance robot voice becomes weak, because the domestic robot manufacturers and robot should be four familiesSignificant differences with regard field, so predictive maintenance is also essential not to consider. Class C Market: Industrial Vision and Image Recognition general AI did get rapid development in the field of visual and security fields. According to relevant statistics show that: in 2018 accounted for 22% of the entire visual AI applications, but industrial applications accounted for very little. The reason why the vision applications routed Finally, since the industrial code reading and traceability of the art has been perfect, AI will not be much market, but in defect detection, topography measurement, monitoring and other occasions or electrical hazard will use AI, these optical technology generally relates to the field of structured light, laser, infrared imaging and 3D, but the market is still in a very early infancy. On the current development of industrial vision point of view, everyone seems to be interested in upgrading the image sensor, for AI is not cold. Phenomenon is a good machine vision vendors trying to embed FPGA or ASIC chip vision system to further enhance the handling capacity of the image data, although this is a good sign, but cost-sensitive large-scale industrial vision market could fall further to be observed. Current industrial vision can be classified into basically grab AI-type applications, such as crawling scheme Omron Adept robots, image recognition algorithm in the program is weighted heavily occupied, Omron advantage also lies in product sufficiently comprehensive so that the program can do better optimization, it can also be seen from the acquisition Maisi Ken Omron intends to strengthen this advantage, so the next mergers and acquisitions industry will be more frequent occurrence. Other visual scheme Control Engineering Copyright , including sensor fusion vision applications, such as laser-crawling ISRA combine with the camera, of course, is the core algorithm, the biggest advantage is that do not rely on the cloud or edge server, ATOS and other three scanning application is actually similar, as are the services to specific scenes work, so now only as a potential landing AI shares. Another type of application is the visual aids, such as Advantech recently introduced SKY-642 GPU server, uses Nvidia’s GPU, but this product is not limited to industry, general industry, too, can use, so it can not be specialized industrial application for reference. Node is king industrial data sharing AI era, the ability of the Internet industry enterprises will further enlarge the gap between the companies will further widen because of the amount of data volume, multi-node data who, whoGreat natural advantages, each node not only led to an iterative product, and even affect sales. Just recently, Siemens emphasized his global reach 18 million installed nodes, such numbers meaningful. Similar phone terminal node battle may be staged in the industrial age of the Internet, AI really easy to use direct correspondence acquisition node more than enough. Competition for industrial data did not reach as far node tragic extent of the phone, so the relative consumption industry is still a blue ocean. In addition, this year’s Davos forum proposed another voice, data sharing is to promote industry, let AI accelerate industrial floor, now it seems the idea is not yet a reality, but if there is a better allocation scheme, maybe this species with shared imagination cooperation will be an opportunity to promote the industry forward again. Finally add that the article did not say to the simulation and digital simulation technology is good or bad because the twins are entirely dependent on the relationship between manufacturers and upstream business, the more upstream to the more detailed data, simulation tools is naturally more powerful, and therefore will back to specific application scenarios, the AI needs to learn here will become weaker.