From cloud computing to big data, from artificial intelligence to industrial Internet, these advanced technologies have brought unprecedented opportunities and challenges to industrial manufacturing. With the development of technology, many industrial equipment needs to be data interaction and real-time control, which also puts forward higher requirements for the computing power of control equipment.
As an enterprise with strong technical advantages in the fields of PC control, dual-domain operating system development, motion control algorithms and artificial intelligence algorithms, Zhongke Times is committed to providing efficient, stable and reliable industrial automation and intelligent solutions for industrial scenarios. Song Jianwei, an expert in embedded software and operating system of Zhongke Times, said in an interview with Gongkong that since its establishment, Zhongke Times has always been committed to the research and development of industrial intelligence products integrating computing and control based on PC architecture, which not only provides users with a variety of functional options and expansion possibilities, but also improves PC control performance to a new height and realizes the efficient use of system resources.
Pictured: Song Jianwei, an expert in embedded software and operating systems at Zhongke Times, was interviewed by Gongkong.
In the interview, Song Jianwei shared in detail the efforts of Zhongke Times in the innovation and breakthrough of industrial intelligence products. He pointed out that Zhongke Times has always adhered to innovation in the research and development of industrial intelligence products, and constantly sought breakthroughs. This kind of innovation and breakthrough is not only reflected in the hardware design of the product, but also in the optimization of software and the expansion of functions.
Subvert the traditional control architecture, and release the industrial potential with the integration of computing and control
In the traditional split architecture, the control system is divided into two main parts, the upper computer and the lower computer, which are mainly composed of industrial computers and a variety of PLC controllers, which are not only huge in size, but also difficult to improve the real-time performance of the system with the computing power of the industrial computer, which cannot meet the processing of a large amount of real-time data, and the deployment of multiple controllers causes the geometric increase of system complexity. Compared with the traditional combination of industrial computer and PLC, the industrial intelligence machine products of the era of Zhongke combine edge computing and control, and can have functions such as motion control, logic control, machine vision, configuration display and other functions at the same time through a set of controllers with a set of software and a set of programs, which can make industrial control equipment better adapt to the complex and changeable production environment, and have higher real-time and computing power scalability.
Figure: Main products.
Traditional architectures are difficult to achieve effective breakthroughs when dealing with the real-time and stability requirements involved in a large number of servo motors. When the number of axes exceeds 32, the complexity of traditional architectures is doubled. By simplifying the architecture, the machine can easily cope with large scenarios with 256 axes. Song Jianwei added. In addition, the traditional communication between upper and lower computers is replaced by cross-domain shared memory of industrial intelligence machines, and the data throughput is 2-3 orders of magnitude faster than that of traditional Ethernet communication. Compared with the traditional architecture, due to the lack of excessive communication configuration and line maintenance, and the integration of AI programming and IEC standard process programming, the application of the machine can also reduce the deployment cost, reduce the personnel requirements and the corresponding cost.
In terms of software, Zhongke Times has launched MetaOS and Metahypervisor virtualization management software based on Xenomai's dual-domain operating system, which can achieve efficient resource management and accurate task scheduling. Through software definition, Zhongke Times has realized the expansion and contraction of hardware computing power, and through the isolation of the real-time domain and non-real-time domain of the operating system, as well as the configuration of metahypervisor, the computing power is reasonably allocated and flexibly scaled according to the customer's on-site needs, so as to truly apply the computing power to the key points of the customer's on-site industrial needs through software definition. In addition, seamlessly integrating the Codesys PLC platform, Zhongke Times has also launched Metafacture Mearte software, which not only supports IEC61131-3 programming, compilation, bottoming, running and other functions, but also supports Ethercat, Profinet, EthernetIP, DeviceNet, CanOpen, Modbus, Common industrial communication such as OPC also integrates hundreds of self-developed visual algorithm function blocks of the Chinese Academy of Sciences, such as shape recognition, color recognition, measurement, text recognition, etc.
With the help of the Xenomai framework, under the condition of full load of CPU and GPU pressure in the non-real-time domain, the industrial intelligence machine achieves the stable output capability of 1 millisecond DC synchronization of EtherCAT 128 axes, and the high real-time performance of jitter is less than 30 microseconds. In addition, the industrial intelligence machine products also have powerful computing power. This is mainly due to the close cooperation with Intel, Haiguang, NVIDIA, Cambrian and other hardware, and the powerful computing power of these partners has endowed the machine with excellent performance in artificial intelligence applications, thus bringing users a new computerized automation experience.
With its stable performance, the TechSmart using the Xenomai framework has played a key role in many industries. For example, the control and production decision-making system provided by Zhongke Times for leading polysilicon enterprises has achieved stable operation for 15 consecutive months. Through the deployment of the polysilicon reduction furnace intelligent control system MESPV, the functions of reduction furnace management, batch management, curve viewing, control parameters and issuance can be realized, real-time judgment of the reaction in the furnace, identification of the atomization status in the furnace, and at the same time reduce the unit consumption of electricity, which can save 26,000-52,000 kWh of electricity per batch. The outstanding performance of the in-robot is due to the joint efforts of the excellent R&D team and professional service team of the Chinese Academy of Sciences, as well as the adoption of the open-source Xenomai framework and open-source Linux foundation that have been extensively verified and recognized by IT and OT, which ensure the stability and reliability of the in-robot products in various equipment, field and other environments.
Figure: Zhongke Times provides control and production decision-making systems for leading polysilicon companies.
Dual-domain operating system to create an efficient and stable system environment
The launch of the InnoMachine not only effectively solves the integration problem between OT and IT, but also has the ability to solve key problems such as large computing power demand, mutual interference, high cost and insufficient reliability. However, in a complex industrial environment, where different control requirements need to be handled by a single machine at the same time, how can resources be efficiently allocated and managed, and the isolation between real-time and non-real-time tasks can be ensured
During the development process, the main difficulty focused on how to achieve the separation of real-time and non-real-time domains on a single device. Song Jianwei said. The real-time domain is mainly responsible for tasks that require real-time response, such as motion control algorithms and communication with motor servo, and the execution of operation control interpolation algorithms, electronic cams and other algorithms also requires a large amount of computing resources. The non-real-time domain is mainly used to process tasks such as image recognition, stream processing, and artificial intelligence inference, which do not require high real-time performance, but have great requirements for data throughput and computing resources.
Based on Xenomai + MetaHypervisor, MetaOS realizes the complete isolation of real-time domain and non-real-time domain, and desktop resources can be allocated through Metahypervisor, even if the desktop system is updated, restarted or blue screen, the process control tasks in the real-time domain will not be affected. Just like the smart twin function of a mobile phone, the use of metahypervisor virtualized resource management can flexibly divide physical resources to handle real-time and non-real-time work resources separately. Depending on the nature of the work, time-sensitive tasks can be placed in the real-time domain for processingTasks that are less time-sensitive can be processed in a non-real-time domain. In this way, real-time and non-real-time tasks can avoid interfering with each other.
Continue to optimize the software and deepen the exploration of industry demand
The combination of state-of-the-art software, reliable hardware, and powerful computing power is opening a new chapter in software-defined. At present, the company is continuously optimizing its dual-domain operating system, MetaOS, and is committed to providing more efficient and stable system solutions to meet the evolving needs of industrial control. Song Jianwei said that Zhongke Times has two major R&D centers in Beijing and Shanghai. The Beijing team is mainly responsible for improving the operating system, including virtualization-related management software and real-time domain running software. The Shanghai team is mainly responsible for working closely with vertical industry customers to complete customized projects, and summarize and refine standardized products suitable for various vertical industries based on the results of customized development.
In the future, for some occasions with very high security requirements, Zhongke Times will introduce its own runtime and adopt programming languages that meet memory safety requirements, such as rust, to ensure that its security performance reaches the same level as the Linux kernel. In addition to using Rust to improve security, Zhongke Times also plans to add support for artificial intelligence chip manufacturers on the basis of the traditional IEC61131-3 programming language. By integrating into the self-compiler of the Zhongke era. Users don't need to care whether their application is running on Haiguang CPU or NVIDIA's inference chip, they only need to pay attention to their own process, which can greatly simplify the programming process.
At present, industrial equipment with industrial intelligence as the core is playing an important role in more and more emerging industries. Zhongke Times has accelerated the layout of emerging industries such as semiconductors, photovoltaics, and new energy. In these industries, there is a large demand for EtherCAT axes, as well as high standards such as real-time, image recognition, and AI inference, while the traditional combination of industrial computers and PLCs can no longer meet the growing demand for performance and intelligence. Song Jianwei said: "Compared with traditional industries, emerging industries have more room for development and higher technology acceptance, which will provide a good development environment for technological innovation and market expansion of industrial intelligence. "In the future, Zhongke Times will continue to strengthen technological innovation and product research and development capabilities to improve the adaptability and performance of industrial intelligence in complex industrial environments, focus on solving the "bottleneck" problem of core technologies in key fields, and help industrial development to a new height.