In April 2018, Taipei – Artificial Intelligence (AI) was undoubtedly the hottest topic in the technology industry in the last year or two. In addition to the huge investment from technology giants, financial and other service providers also introduced artificial intelligence. Have a strong interest. Manufacturing industry's attention to AI technology is not to mention, and the actual introduction of action has already begun with related key technologies gradually in place.
In addition to advancing smart manufacturing, Advantech spares no efforts to provide corresponding advanced solutions for various industries, and gradually introduces artificial intelligence elements in its own production online. For example, the state detection/diagnosis of machine equipment, the use of raw materials/energy, and the quality control process of products have gradually been introduced into artificial intelligence. Arm's silicon intellectual property (IP) and SoC and Xilinx's Field Programmable Gate Array (FPGA) technology are the two effective assistants for Advantech's promotion of AI in production lines.
Advantech IoT.SENSE interviewed Advantech Chief Technical Officer Yang Ruixiang and Chief Engineer Lin Dongjie to discuss Advantech's solutions in the wave of innovations such as AI, IoT, and intelligent manufacturing. The following is an excerpt of the interview:
AI Evolution Speeds Amazing Business Application Values <br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br> Artificial intelligence is not a new topic in the academic research field The evolutionary speed is too astounding, and it has been able to create enormous commercial value. It is no longer just a subject of academic research.
In addition to targeting specific areas, AI technology is also pursuing greater versatility. Deepmind's latest chess program has taken Go (Go) and called it Alpha. Because this program also knows about other games such as Japanese chess, and defeats other world-class dedicated chess programs one after another. This is undoubtedly an important milestone in the development of artificial intelligence.
It is not surprising that artificial intelligence has become the most highly-respected technology issue at the moment of rapid evolution and the existence of huge commercial value. However, discussions have gone into discussions. There are still many details to overcome in order to introduce artificial intelligence in all walks of life and achieve industrial intelligence.
Artificial Intelligence Adds Smart Manufacturing Momentum <br> In the case of manufacturing, regardless of the final manufactured product, manufacturing is always inseparable from the four elements of "man, machine, material, and law." People refer to employees, while machines refer to various tools and machines. Materials refer to various raw materials and energy, and rules are process methods. Since the industrial revolution, no matter how the manufacturing products evolved, these four elements cannot be separated. How to optimally manage these four elements is the task that manufacturers face every day.
Yang Ruixiang analyzed that the introduction of artificial intelligence, the most important four KPIs, is to demonstrate the optimization and improvement of artificial material methods. In terms of people, how to transform the experience of the master teacher into quantifiable parameters, and then copy and diffuse human experience, is an important goal for the introduction of AI.
However, to realize the above four optimizations, the most important thing is the degree of understanding of the AI ​​and the quality of the collected data sets. First of all, the manufacturer must have a correct understanding of AI, know what the AI ​​is suitable for dealing with, and what restrictions apply. Second, the training results of the AI ​​inference model, in addition to the design of the model itself, the quality of training data is also very important. If the model is trained with data of questionable quality, the results of AI inferences will fall short of reality.
Finally, organizational culture has to be adjusted. Prior to the introduction of AI, all decision-makers on the production line were humans, relying on past experience; after importing AI, although the final decision-making power was still human, it was no longer based solely on subjective feelings or experience. It is a relatively objective statistical science. The trust relationship between people and machines will take some time to improve. Of course, AI itself must continue to evolve, improving the reliability and accuracy of its predictions.
Different processors each have their own strengths. Arm architecture is suitable for inference operations. Yang Ruixiang further explained that artificial intelligence can be divided into two parts: training and inference. For production field applications, most of the inferred applications are executed using already trained models. Model training is not performed directly at the edge because model training requires powerful computing performance and large data sets, and is more suitable for data centers or On the cloud.
Because the inference is less demanding on computing performance, there are many ready-made processor solutions available on the market. For example, x86 CPUs, GPUs, and Arm-based SoC processors can perform related computing tasks. The only difference is that Whether the cost, power consumption and heat dissipation can meet the specifications of field devices.
From a technical point of view, the GPU is currently the most suitable processor architecture for model training. It is certainly more than sufficient to infer its execution model. However, the heat dissipation problem of the GPU's cost and power consumption follows this category. Processors have the biggest limitation on the application of edge nodes or field devices. The x86 CPU also has a very powerful computational efficiency, but because its architectural design goal is to satisfy a variety of computation/control applications, the efficiency of the AI ​​algorithm is not as good as that of the GPU.
Yang Ruixiang analyzed that this issue is related to the nature of AI. AI usually handles large amounts of data with only a few instructions, or even a single instruction. Deep learning and Convolutional Neural Network (CNN), for example, are matrix operations in terms of mathematics and are very similar to drawing operations. Therefore, GPU naturally has an inherent advantage in this regard. The x86 CPU is longer than the Multiple Instruction (Multiple Instruction, Multiple Data, MIMD) computing environment, but when the data volume is too large, it must rely on frequency, or multi-core and multi-core. Thread architecture to deal with.
ARM processor with Reduced Instruction Set (RISC), the innate feature is between the GPU and the x86 CPU. In addition, the Single Instruction (Multiple Data, SIMD) performance of the Arm processor has been continuously enhanced in recent years. , so it is more handy when performing AI calculations. Although Arm's processor is currently used for model training, it still cannot be compared with GPU in terms of efficiency. However, when performing inference tasks, it is the most balanced solution for power consumption, cost, and performance.
Yang Ruixiang revealed that in recent years, Advantech has worked closely with Onmook to gain a certain understanding of the product development blueprint of Onm. The future security plan will introduce more specialized, more efficient processor cores and peripheral IPs for AI computing needs. This will be a great boost to the popularity of edge computing and AI applications. Advantech will also continue to maintain a close partnership with Andromeda.
Edge computing progress rapid artificial intelligence stationed in the manufacturing site <br> Closely grasp the four elements of the artificial material method, Advantech has begun to use the Arm architecture SoC and Xilinx FPGA module as the hardware basis, gradually introduce artificial intelligence in its own production line.
Lin Dongjie, Advantech's chief director, said that at present, Advantech's online introduction of AI in production has entered the stage of using AI to assist in interpretation of the original data. In the era of industrial Internet of Things, not only do individual machines that produce on-line machines generate large amounts of data, but the infrastructure of the plant area also generates considerable amounts of data and data. To use artificial data to interpret these data and analyze the meaning behind it is a time-consuming and limited benefit approach.
Finally, since the environment in which Advantech is located is typically a small number of diverse and order-based production types, it is very different from the general consumer product specifications and mass production. Therefore, the management of production lines is also relatively complicated. This is also one of the pain points that Advantech hopes to solve when it introduces artificial intelligence.
Lin Dongjie stated that due to technical limitations, the ultimate goal of comprehensive interpretation of raw data by the system cannot be achieved at present, but this is the direction of future efforts of Advantech.
More specifically, the future of Advantech's smart manufacturing hopes to achieve three major goals: First, the modernization of production equipment, hope that all machine equipment can support Industry 4.0; Second, to achieve data acquisition and software interface, mainly The data is interfaced with systems such as Manufacturing Execution System (MES) and Product Lifecycle Management (PLM). Third, machine vision and deep learning are further expanded in the quality assurance process.
Regarding the first point, Lin Dongjie does not say that the upgrading and transformation of existing machines is usually not cheap, especially if they need the original factory to provide support or authorization and cannot change their own hands. However, under certain circumstances, the existing machine can already obtain enough parameter data through the data acquisition module developed by Advantech.
As for the expansion of machine vision and deep learning, Advantech is currently working with the Academia Sinica to develop a machine vision system that can detect a variety of different products. In fact, Advantech has been using optical automatic inspection (AOI) for a long time, but the existing AOI system is only applicable to the detection of subtle components on the main board and the circuit board. It is not suitable for detecting the finished end product or larger zeros. part.
On the other hand, the small number of features of Advantech's products has also made the machine vision solutions currently available in the market to be used in Advantech's production lines, and has encountered considerable difficulties. Currently, most of the machine vision solutions on the market are designed for the detection requirements of a large number of products, but Advantech's needs are machine vision detection solutions that can automatically adapt to various product types. Therefore, Advantech decided to cooperate with the Academia Sinica to develop a customized deep learning algorithm so that the machine vision system can more intelligently adapt to different types of products.
FPGA Module Accelerates Machine Vision Algorithm <br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br> Through the FPGA module, Advantech is free to decide which image recognition steps need hardware acceleration to improve the performance of the visual inspection system.
Yang Ruixiang pointed out that in addition to the CPU and GPU, the use of dedicated hardware acceleration chips to enhance the performance of AI systems is theoretically a viable path. However, at present, AI algorithm is still rapidly evolving. If ASIC is adopted, it may not be able to catch up with the pace of technological development. FPGA is a tradeoff between performance and flexibility. The structure of the computing unit can be customized to meet the acceleration of specific algorithms, and because of programmability, when the algorithm needs to be modified or updated, there is no need to reopen a chip. , Just modify the design program code.
Therefore, at this stage, FPGA is one of the ideal schemes for realizing acceleration of AI algorithms. Advantech also has a fairly mature FPGA application development team and will continue to invest in this technology in the future.
About Advantech
Advantech was established in 1983 with the vision of “Smart Earth Pusherâ€, focusing on the industrial Internet of Things and smart city industry, providing software and hardware integration solutions. Advantech's business is distributed in 26 countries and employs nearly 8,000 employees. It provides customers with localized responsive services with a strong technical service and marketing network. At the same time, Advantech has assembled the power of industrial partners to actively create a cooperative and win-win IoT ecosystem and accelerate the application of Internet of Things. (company website:)
About Advantech Embedded IoT Group
As a global leader in embedded platforms, Advantech's IoT embedded platform business group not only provides embedded platform and design services, but also integrates software and hardware integrated IoT solutions to help customers deploy IoT applications. Advantech IoT solutions include the M2.COM awareness platform, gateway, Edge Intelligence Server (EIS) and WISE-PaaS IoT software platform. (URL: http://)
About Embedded IOT Online
Advantech Embedded Online is an official online consultancy sales channel established by the Advantech IoT Embedded Platform Business Group in Shenzhen. In order to serve customers sincerely, Advantech's embedded online department is equipped with a well-developed product and technical support team to make unremitting efforts to provide customers with highly-adapted and cost-effective embedded product solutions; and to open embedded hotlines (400-001). -9088), Advantech Online Mall (buy.advantech.com.cn), Jingdong Embedded Flagship Store, Alibaba Embedded Flagship, and many other ports, providing customers with one-stop service from product selection, procurement to after-sales service In order to solve the diversified needs of customers for embedded products, to achieve rapid response and convenient communication.
In addition to advancing smart manufacturing, Advantech spares no efforts to provide corresponding advanced solutions for various industries, and gradually introduces artificial intelligence elements in its own production online. For example, the state detection/diagnosis of machine equipment, the use of raw materials/energy, and the quality control process of products have gradually been introduced into artificial intelligence. Arm's silicon intellectual property (IP) and SoC and Xilinx's Field Programmable Gate Array (FPGA) technology are the two effective assistants for Advantech's promotion of AI in production lines.
Advantech IoT.SENSE interviewed Advantech Chief Technical Officer Yang Ruixiang and Chief Engineer Lin Dongjie to discuss Advantech's solutions in the wave of innovations such as AI, IoT, and intelligent manufacturing. The following is an excerpt of the interview:
AI Evolution Speeds Amazing Business Application Values <br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br> Artificial intelligence is not a new topic in the academic research field The evolutionary speed is too astounding, and it has been able to create enormous commercial value. It is no longer just a subject of academic research.
In addition to targeting specific areas, AI technology is also pursuing greater versatility. Deepmind's latest chess program has taken Go (Go) and called it Alpha. Because this program also knows about other games such as Japanese chess, and defeats other world-class dedicated chess programs one after another. This is undoubtedly an important milestone in the development of artificial intelligence.
It is not surprising that artificial intelligence has become the most highly-respected technology issue at the moment of rapid evolution and the existence of huge commercial value. However, discussions have gone into discussions. There are still many details to overcome in order to introduce artificial intelligence in all walks of life and achieve industrial intelligence.
Artificial Intelligence Adds Smart Manufacturing Momentum <br> In the case of manufacturing, regardless of the final manufactured product, manufacturing is always inseparable from the four elements of "man, machine, material, and law." People refer to employees, while machines refer to various tools and machines. Materials refer to various raw materials and energy, and rules are process methods. Since the industrial revolution, no matter how the manufacturing products evolved, these four elements cannot be separated. How to optimally manage these four elements is the task that manufacturers face every day.
Yang Ruixiang analyzed that the introduction of artificial intelligence, the most important four KPIs, is to demonstrate the optimization and improvement of artificial material methods. In terms of people, how to transform the experience of the master teacher into quantifiable parameters, and then copy and diffuse human experience, is an important goal for the introduction of AI.
However, to realize the above four optimizations, the most important thing is the degree of understanding of the AI ​​and the quality of the collected data sets. First of all, the manufacturer must have a correct understanding of AI, know what the AI ​​is suitable for dealing with, and what restrictions apply. Second, the training results of the AI ​​inference model, in addition to the design of the model itself, the quality of training data is also very important. If the model is trained with data of questionable quality, the results of AI inferences will fall short of reality.
Finally, organizational culture has to be adjusted. Prior to the introduction of AI, all decision-makers on the production line were humans, relying on past experience; after importing AI, although the final decision-making power was still human, it was no longer based solely on subjective feelings or experience. It is a relatively objective statistical science. The trust relationship between people and machines will take some time to improve. Of course, AI itself must continue to evolve, improving the reliability and accuracy of its predictions.
Different processors each have their own strengths. Arm architecture is suitable for inference operations. Yang Ruixiang further explained that artificial intelligence can be divided into two parts: training and inference. For production field applications, most of the inferred applications are executed using already trained models. Model training is not performed directly at the edge because model training requires powerful computing performance and large data sets, and is more suitable for data centers or On the cloud.
Because the inference is less demanding on computing performance, there are many ready-made processor solutions available on the market. For example, x86 CPUs, GPUs, and Arm-based SoC processors can perform related computing tasks. The only difference is that Whether the cost, power consumption and heat dissipation can meet the specifications of field devices.
From a technical point of view, the GPU is currently the most suitable processor architecture for model training. It is certainly more than sufficient to infer its execution model. However, the heat dissipation problem of the GPU's cost and power consumption follows this category. Processors have the biggest limitation on the application of edge nodes or field devices. The x86 CPU also has a very powerful computational efficiency, but because its architectural design goal is to satisfy a variety of computation/control applications, the efficiency of the AI ​​algorithm is not as good as that of the GPU.
Yang Ruixiang analyzed that this issue is related to the nature of AI. AI usually handles large amounts of data with only a few instructions, or even a single instruction. Deep learning and Convolutional Neural Network (CNN), for example, are matrix operations in terms of mathematics and are very similar to drawing operations. Therefore, GPU naturally has an inherent advantage in this regard. The x86 CPU is longer than the Multiple Instruction (Multiple Instruction, Multiple Data, MIMD) computing environment, but when the data volume is too large, it must rely on frequency, or multi-core and multi-core. Thread architecture to deal with.
ARM processor with Reduced Instruction Set (RISC), the innate feature is between the GPU and the x86 CPU. In addition, the Single Instruction (Multiple Data, SIMD) performance of the Arm processor has been continuously enhanced in recent years. , so it is more handy when performing AI calculations. Although Arm's processor is currently used for model training, it still cannot be compared with GPU in terms of efficiency. However, when performing inference tasks, it is the most balanced solution for power consumption, cost, and performance.
Yang Ruixiang revealed that in recent years, Advantech has worked closely with Onmook to gain a certain understanding of the product development blueprint of Onm. The future security plan will introduce more specialized, more efficient processor cores and peripheral IPs for AI computing needs. This will be a great boost to the popularity of edge computing and AI applications. Advantech will also continue to maintain a close partnership with Andromeda.
Edge computing progress rapid artificial intelligence stationed in the manufacturing site <br> Closely grasp the four elements of the artificial material method, Advantech has begun to use the Arm architecture SoC and Xilinx FPGA module as the hardware basis, gradually introduce artificial intelligence in its own production line.
Lin Dongjie, Advantech's chief director, said that at present, Advantech's online introduction of AI in production has entered the stage of using AI to assist in interpretation of the original data. In the era of industrial Internet of Things, not only do individual machines that produce on-line machines generate large amounts of data, but the infrastructure of the plant area also generates considerable amounts of data and data. To use artificial data to interpret these data and analyze the meaning behind it is a time-consuming and limited benefit approach.
Finally, since the environment in which Advantech is located is typically a small number of diverse and order-based production types, it is very different from the general consumer product specifications and mass production. Therefore, the management of production lines is also relatively complicated. This is also one of the pain points that Advantech hopes to solve when it introduces artificial intelligence.
Lin Dongjie stated that due to technical limitations, the ultimate goal of comprehensive interpretation of raw data by the system cannot be achieved at present, but this is the direction of future efforts of Advantech.
More specifically, the future of Advantech's smart manufacturing hopes to achieve three major goals: First, the modernization of production equipment, hope that all machine equipment can support Industry 4.0; Second, to achieve data acquisition and software interface, mainly The data is interfaced with systems such as Manufacturing Execution System (MES) and Product Lifecycle Management (PLM). Third, machine vision and deep learning are further expanded in the quality assurance process.
Regarding the first point, Lin Dongjie does not say that the upgrading and transformation of existing machines is usually not cheap, especially if they need the original factory to provide support or authorization and cannot change their own hands. However, under certain circumstances, the existing machine can already obtain enough parameter data through the data acquisition module developed by Advantech.
As for the expansion of machine vision and deep learning, Advantech is currently working with the Academia Sinica to develop a machine vision system that can detect a variety of different products. In fact, Advantech has been using optical automatic inspection (AOI) for a long time, but the existing AOI system is only applicable to the detection of subtle components on the main board and the circuit board. It is not suitable for detecting the finished end product or larger zeros. part.
On the other hand, the small number of features of Advantech's products has also made the machine vision solutions currently available in the market to be used in Advantech's production lines, and has encountered considerable difficulties. Currently, most of the machine vision solutions on the market are designed for the detection requirements of a large number of products, but Advantech's needs are machine vision detection solutions that can automatically adapt to various product types. Therefore, Advantech decided to cooperate with the Academia Sinica to develop a customized deep learning algorithm so that the machine vision system can more intelligently adapt to different types of products.
FPGA Module Accelerates Machine Vision Algorithm <br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br><br> Through the FPGA module, Advantech is free to decide which image recognition steps need hardware acceleration to improve the performance of the visual inspection system.
Yang Ruixiang pointed out that in addition to the CPU and GPU, the use of dedicated hardware acceleration chips to enhance the performance of AI systems is theoretically a viable path. However, at present, AI algorithm is still rapidly evolving. If ASIC is adopted, it may not be able to catch up with the pace of technological development. FPGA is a tradeoff between performance and flexibility. The structure of the computing unit can be customized to meet the acceleration of specific algorithms, and because of programmability, when the algorithm needs to be modified or updated, there is no need to reopen a chip. , Just modify the design program code.
Therefore, at this stage, FPGA is one of the ideal schemes for realizing acceleration of AI algorithms. Advantech also has a fairly mature FPGA application development team and will continue to invest in this technology in the future.
Advantech Embedded DTOS FPGA Capability
It is reported that at present, Advantech has already applied cases to complete cooperation. For more details, please contact Advantech's service hotline “400-001-9088â€. About Advantech
Advantech was established in 1983 with the vision of “Smart Earth Pusherâ€, focusing on the industrial Internet of Things and smart city industry, providing software and hardware integration solutions. Advantech's business is distributed in 26 countries and employs nearly 8,000 employees. It provides customers with localized responsive services with a strong technical service and marketing network. At the same time, Advantech has assembled the power of industrial partners to actively create a cooperative and win-win IoT ecosystem and accelerate the application of Internet of Things. (company website:)
About Advantech Embedded IoT Group
As a global leader in embedded platforms, Advantech's IoT embedded platform business group not only provides embedded platform and design services, but also integrates software and hardware integrated IoT solutions to help customers deploy IoT applications. Advantech IoT solutions include the M2.COM awareness platform, gateway, Edge Intelligence Server (EIS) and WISE-PaaS IoT software platform. (URL: http://)
About Embedded IOT Online
Advantech Embedded Online is an official online consultancy sales channel established by the Advantech IoT Embedded Platform Business Group in Shenzhen. In order to serve customers sincerely, Advantech's embedded online department is equipped with a well-developed product and technical support team to make unremitting efforts to provide customers with highly-adapted and cost-effective embedded product solutions; and to open embedded hotlines (400-001). -9088), Advantech Online Mall (buy.advantech.com.cn), Jingdong Embedded Flagship Store, Alibaba Embedded Flagship, and many other ports, providing customers with one-stop service from product selection, procurement to after-sales service In order to solve the diversified needs of customers for embedded products, to achieve rapid response and convenient communication.
Car Phone Holder,Mobile Holder For Cars,Mobile Phone Holder For Dashboard,Mobile Phone Holder For Car Dashboard
Ningbo Luke Automotive Supplies Ltd. , https://www.car-phone-holder.com