Edit | Shi Yaqiong
Image credit | pexels @ Kateryna Babaieva
Numerous industrial Internet reports show that among the application cases at home and abroad, the industrial Internet platform (IIoT) is the most widely used in the machinery and energy industry. According to PTC, the world's leading provider of IIoT software platforms, "These industries can lead the IoT trend because they have complex manufacturing and operation processes and high-asset-worth devices, so they can get high benefits from IoT solutions and digital transformation." ”
36Kr previously briefly reported a company transformed by Tsinghua University's technological achievements - Beijing Huakong Zhijia Technology Co., Ltd. (hereinafter referred to as "Huakong Zhijia"), they have high-end equipment customers, with subtle feature detection, low resources, non-collaborative modeling as the core technology entry point, empower large-scale industrial equipment health and safety monitoring, provide predictive maintenance programs and equipment health management, these days we met with the company's CEO Liu Deguang, specifically learned about the company's technology and business situation.
Large/high-end equipment is highly valuable, has a long life cycle, and places a high value on asset performance management, creating significant value for every 1% reduction in equipment wear and tear or downtime due to maintenance. Existing maintenance strategies are mostly based on preventive maintenance with regular inspections or post-maintenance. Since regular testing is based solely on past experience, it may not be possible to maintain it in a timely manner, or it may cause unnecessary maintenance. It not only consumes a lot of manpower and material resources, but also is inefficient.
36Kr has reported on a number of science and technology innovation companies that have entered the predictive maintenance of industrial equipment with acoustic technology, such as Shuo Orange and Zhensheng, the former receiving Pre-A round financing in July 2019, the latter receiving Series A financing in the same month, and Yaosheng Technology from the Institute of Acoustics of the Chinese Academy of Sciences, which was established in 2019.
The biggest difference between "Huakong Zhijia" and the above projects is that the former extracts multi-domain features, including vibration, sound, image, and multi-sensor, in addition to the use of multi-domain features and subtle feature extraction, modeling in a deep neural network mode that integrates physical mechanisms, configurable intelligent computing on embedded edge chips and deep calculation and management in the cloud, which can realize earlier detection of faults, find that equipment may have "minor diseases", timely early warning and timely maintenance, and break the model of relying only on regular maintenance to achieve maintenance based on equipment health monitoring.
Image credit | Huakong Zhijia
The advantages of the "Huakong Zhijia" scheme mainly lie in two points, one is the extraction of subtle features, the collection of lossless broadband sound signals, and the use of sound and subtle features contained in the ultra-raw frequency band to detect earlier fault vibrations. The second is the use of low-resource deep learning modeling methods, industrial equipment will not cooperate with the detection system, the failure rate of large equipment is also very low, through the deep neural network to extract vibration + sound fusion broadband data in the characterization of the early subtle features of the fault, the scheme to obtain a small number of fault samples can achieve deep learning modeling, easier to adapt to industrial equipment health management.
At the same time, edge computing and cloud intelligent processing model can improve the accuracy and efficiency of fault detection, flexibility and reduce the cost of solution implementation.
In both aspects, the team's technical and scientific research capabilities are very demanding, "Huakong Zhijia" has a completely independent core algorithm in the fields of voiceprint recognition, microphone array, machine voiceprint, subtle feature extraction, speech and audio recognition, and has obtained 12 sets of invention patents transferred by Tsinghua University. In recent years, it has continuously won the top three in the nist evaluation of the most authoritative voiceprint recognition technology evaluation agency in the world, and the participating units include MIT in the United States, Berkeley, California and so on.
The company provides a complete set of solutions from edge computing hardware systems, intelligent terminals, low-resource deep learning modeling, configurable intelligent algorithms to cloud platforms, "Huacon Zhijia" told 36Kr that the company itself has low-cost chip development and development capabilities, and for hardware that meets the requirements that cannot be purchased on the market, such as microphone arrays and vibration sensors, the company has invested in research and development and has relevant patents, and entrusted factories to mass-produce.
In terms of business progress, "Huakong Zhijia" has cooperated with leading customers in the smart city, energy, petrochemical, new materials and steel industries with leading customers in the state grid, Beijing Energy Group and other industries, and has developed the first set of coal mill operation monitoring and fault diagnosis system based on vibration/sound pattern recognition technology at home and abroad.
In terms of profit model, "Huakong Zhijia" currently provides a complete set of solutions, and the unit price of customers is in the million level. At the same time, the company relies on the high scientific research strength of voice and audio signal processing, low-resource modeling of neural network structures that integrate physical mechanisms, etc., to carry out intelligent speech and audio processing and urban credit technology business, the former mainly provides full-stack algorithms and products, empowering companies lacking intelligent voice underlying technology, the latter mainly serves government departments, which is conducive to maintaining the company's hematopoietic capacity. According to CEO Liu Deguang, the company's current performance growth is relatively fast, and it is expected to achieve a flat income this year.
In terms of development planning, Liu Deguang told 36Kr that in the construction of the industrial Internet platform, after waiting for the data to accumulate enough, it will further tap the value of the data, on the one hand, to provide docking financial and insurance services for industrial manufacturers, on the other hand, it will gradually expand its business to the equipment design and manufacturing process of equipment manufacturers.
In terms of team members, the company currently has more than 40 people, the core technology and team were born from the speech laboratory of the Department of Electronic Engineering of Tsinghua University, and the main members of the founding management team graduated from Tsinghua university and have rich entrepreneurial and management experience. Professor Liu Jia of the Department of Electronics of Tsinghua University, the chief scientist of the company, is a well-known scholar at home and abroad, focusing on speech research for more than 30 years; CEO Liu Deguang is an alumnus of Tsinghua University, engaged in venture capital and entrepreneurship for more than 20 years, with rich management and marketing experience.
In May 2018, Huakong Zhijia received more than 20 million in cash investments, invested by Tsinghua Holdings, Huakong Cornerstone Fund and Tsinghua Alumni Management. Currently the company has financing needs.
At present, there are not many solution providers of large/high-end equipment prediction and health management (PHM), according to "Huakong Zhijia", the reason is that there is no mature safety prediction monitoring scheme for large-scale high-end equipment in the market, and the projects received by the company are more like scientific research projects entrusted by manufacturers and users, which requires a company's scientific research strength and brand endorsement. Backed by Tsinghua University, "Huakong Zhijia" has a very strong multidisciplinary scientific research strength as a support, and has a great advantage in obtaining customers.
<h3>Read more</h3>
Who are the players in the nearly trillion industrial Internet track? Here's an industry map
In-depth research | Dismantle the trillion industrial Internet track and find the next GE