A startup founded by a former team from the Institute of Acoustics of the Chinese Academy of Sciences is leveraging acoustic AI technology to tap into a market worth hundreds of billions | Exclusive from 36Kr
Text | Lin Qingqing
Editor | Yuan Silai
36Kr learned that recently, "Dishēng Technology", a developer of industrial acoustic AI monitoring technology, announced the completion of a Series D financing of over 100 million RMB. The Beijing Advanced Manufacturing Fund led the investment, and institutions such as Softbank China Capital, Bank of Communications Capital, Guoxin Guozheng, and Jinpan Capital participated in the follow - on investment. The funds from this round will be used for technology R & D, overseas market expansion, and industrial chain resource integration. So far, "Dishēng Technology" has raised hundreds of millions of RMB in total, and its historical investors include the Beijing CMCC Digital New Economy Industry Fund under China Mobile, Lenovo Capital and Incubation Group, Vertex Ventures China, etc.
"Dishēng Technology" was founded in 2017. Its core team members come from the Institute of Acoustics of the Chinese Academy of Sciences. It focuses on providing fault prediction and health management services for industrial equipment through non - contact acoustic monitoring technology. The company independently develops microphone array hardware, AI voiceprint recognition algorithms, and establishes its own acoustic database for industrial equipment. Its business covers fields such as power, rail transit, and energy, and its customers include leading enterprises such as the State Grid, CRRC, and the MTR.
Its technical principle mainly involves studying the sound propagation mechanism of equipment, collecting the acoustic signals of equipment operation in a non - contact manner, and training and iterating the AI recognition algorithm through a fault sound database of over 100T (including more than 50 power grid scenarios, 37 rail transit scenarios, and more than a hundred industrial manufacturing scenarios) to achieve early warning and maintenance suggestions for operating equipment, replacing traditional manual inspections and contact - type sensor solutions.
Acoustic AI monitoring is regarded as an important path for the intelligent transformation of industrial equipment. Traditional maintenance relies on manual experience or invasive sensors. The former is inefficient and difficult to cover complex scenarios, while the latter requires invasive installation and has high transformation costs and safety hazards. Acoustic monitoring can penetrate physical obstructions to identify internal faults in a non - contact state by capturing the vibration and noise signals of equipment, such as DC bias in transformers, internal looseness, cracks in high - speed train wheel - set bearings, and wheel - set polygons.
Globally, companies such as Denmark's BK and Israel's 3D Signals entered this field earlier. However, restricted by data security, local adaptation, and price factors, the penetration of overseas products in domestic monopoly industries is limited. The China Business Industry Research Institute predicts that the market scale of equipment monitoring based on acoustic technology will exceed 350 billion RMB in recent years, which is a huge blue - ocean market.
However, there are still some pain points in the acquisition and standardized processing of effective data in complex industrial scenarios. The noise interference in the equipment operating environment is large, and the acoustic characteristics are easily affected by temperature, humidity, and mechanical structure differences. Traditional solutions are difficult to achieve accurate identification.
The "Dishēng Technology" team initially started with horizontal projects from the Institute of Acoustics of the Chinese Academy of Sciences and chose industries with strong monopoly and sufficient budgets, such as the power grid and rail transit, to build barriers. "The technical paths of domestic industrial equipment are different from those overseas. For example, the layout of China's UHV power grid and the special structure of high - speed train wheel - sets require the localization and reconstruction of acoustic characteristic models." Ding Dongliang, the founder of "Dishēng Technology", believes that future industry competition will focus on the ability of data closed - loop and the efficiency of scenario migration. The company plans to expand into new scenarios such as automotive NVH testing and ultrasonic monitoring and layout the Southeast Asian market with the help of MTR resources.
Ding Dongliang told 36Kr that based on this, the key breakthrough point in the industry lies in the collaboration between front - end hardware and back - end systems. The far - field directional microphone array independently developed by the team has been deployed in multiple industrial scenarios, paired with an AI noise reduction algorithm to extract effective signals from complex sound fields. The back - end database has been accumulated over a decade and covers more than 170 fault scenarios in fields such as power and rail transit. A hierarchical diagnosis system has been formed through cooperation with the China Electric Power Research Institute and the China Academy of Railway Sciences. For example, in a UHV substation project of the State Grid, the accuracy of its equipment in identifying abnormal voiceprint samples of substation equipment reached over 90%, such as over - loading, DC bias, internal looseness, and abnormal noise from coolers.
Currently, "Dishēng Technology" has launched core products such as a transformer voiceprint monitoring system, a high - speed train running gear abnormal sound detection device, and an acoustic imager.