Recently, the research team led by Associate Professor Chen Peng has achieved new advances in intelligent fault diagnosis for core deep underground engineering equipment. Relevant research findings have been published in Mechanical Systems and Signal Processing, an internationally authoritative journal in mechanical fault diagnosis and signal processing (JCR Q1, CAS Engineering & Technology Q1 Top Journal, IF=8.9).

Multi-cylinder drilling pumps are core power equipment for major projects such as deep underground drilling and oil & gas development, undertaking critical tasks including circulating and conveying high-pressure drilling fluid. Their operating conditions directly determine drilling efficiency, operational safety and service life of equipment. Nevertheless, long-term operation under complex working conditions exposes multi-cylinder drilling pumps to combined impacts of high-pressure pulsation, mechanical shock and multi-source coupled vibration. As a result, fault signals are characterized by prominent non-stationarity, severe aliasing and weak feature information. Particularly under coordinated operation of multiple cylinders, cylinders and components share similar mechanical mechanisms, leading to highly overlapping vibration responses in the frequency spectrum. This typical homotypic multi-source aliasing issue makes early fault features hard to identify and poses great challenges to intelligent diagnosis and health monitoring.

To address the bottlenecks of difficult separation of fault signals, extraction of weak features and insufficient stability across variable pressure conditions for deep underground engineering equipment under complex operating environments, the team proposed the Homotypic Multi-source Joint Representation with Dynamic Hierarchical Feature Tracing (HMR-DHFT) method. Targeting the classic homotypic multi-source aliasing problem of multi-cylinder drilling pumps, the method establishes a joint feature representation and dynamic hierarchical tracing mechanism based on the coupling mechanism of multi-source signals. It can gradually capture in-depth fault information amid complex background interference, enabling stable identification and effective characterization of key abnormal patterns.

Experimental results demonstrate that the proposed method achieves outstanding performance in various health state recognition tasks, outperforming comparative mainstream approaches with remarkable fault discrimination capability. It maintains stable and reliable identification accuracy under different discharge pressures, presenting strong cross-working-condition adaptability and robustness. Furthermore, the method can accurately distinguish varying wear degrees and fault evolution stages, which proves its capacity not only for fault category classification but also for quantifying fault severity. In addition, feature visualization analysis verifies that the proposed method realizes clearer feature separation between different fault types against complex signal backgrounds and progressively amplifies critical fault information during hierarchical feature extraction. The method boasts distinct advantages in handling multi-source aliasing, fine-grained state identification and adaptation to complex working conditions, offering a novel technical route for condition monitoring, fault early warning and intelligent operation & maintenance of deep underground engineering equipment.

This research was supported by the National Natural Science Foundation of China (Grant Nos. 52105111, 52305085) and the Guangdong Basic and Applied Basic Research Foundation (Grant No. 2025A1515012256).

Peng Chen, Yazheng Wang, Yuhao Wu, Yaqiang Jin, Changbo He*, Ge Xin. Homotypic multi-source joint representation with dynamic hierarchical feature tracing in hybrid Walsh–frequency domain for prior knowledge-constrained fault diagnosis of multi-cylinder hydraulic pumps. Mechanical Systems and Signal Processing, 2026, Volume 255, Article ID 114450.DOI: 10.1016/j.ymssp.2026.114450Paper Link: https://doi.org/10.1016/j.ymssp.2026.114450

Chen Peng, Doctor of Engineering, Associate Professor and Master’s Supervisor. He is selected into the Distinguished Talent Program of Shantou University and recognized as a High-Level Talent of Shantou City. He serves as a peer reviewer for the National Natural Science Foundation of China, a member of IEEE, and committee member of the Signal Processing Branch, Dynamic Testing Committee and Rotor Dynamics Branch of the Chinese Society for Vibration Engineering. He is also a dissertation reviewer for the Academic Degrees & Graduate Education Development Center of the Ministry of Education and a reviewer for national undergraduate graduation thesis random inspections.

He obtained his Doctor of Engineering degree from the University of Electronic Science and Technology of China in 2020, supervised by Professor Ming J. Zuo, Fellow of the Canadian Academy of Engineering and IEEE Fellow. He completed joint doctoral training at KU Leuven, Belgium from 2019 to 2020, and was a visiting research scholar at the University of Pretoria, South Africa, working with Professor P. Stephan Heyns, Member of the South African Academy of Sciences in 2018.

In recent years, Associate Professor Chen Peng has focused on intelligent acoustics and multi-modal signal processing, machine vision and advanced sensing, trustworthy multi-modal AI and collaborative computing, embodied intelligence of industrial servo robots, as well as detection and diagnosis of new energy lithium batteries and key equipment. His research features distinctive applications in complex systems including airborne power systems for aerospace, intelligent transportation (high-speed trains), wind power transmissions and industrial robots.

As principal investigator, he has presided over more than 10 research projects, including the Youth Program of the National Natural Science Foundation of China (successfully closed with excellent evaluation and invited oral presentation), two General Projects of the Guangdong Basic and Applied Basic Research Foundation, Guangdong Science and Technology Planning Projects, sub-projects of key R&D programs of AVIC, and commissioned projects from the Shunde Branch of Guangdong Special Equipment Inspection Institute. He has also participated in 6 national key R&D programs, key and general projects of the National Natural Science Foundation of China.

As first or corresponding author, he has published over 50 SCI papers in JCR/CAS Q1 & Q2 Top Journals, including 2 Highly Cited / Hot Papers. His publications appear in flagship journals of the field such as Mechanical Systems and Signal Processing, Expert Systems with Applications, Knowledge-Based Systems, Engineering Applications of Artificial Intelligence, IEEE Internet of Things Journal, IEEE Transactions on Instrumentation and Measurement, IEEE Transactions on Reliability, Journal of Sound and Vibration, Structural Health Monitoring, Nonlinear Dynamics and Ocean Engineering. He holds 6 authorized invention patents.

He has chaired technical sessions and delivered invited presentations at international conferences including UNIfied-2026-SMMI and TEPEN2024-IWFDP. He works as an ad-hoc reviewer for multiple international journals and serves as Guest Editor for journals such as IEEE Internet of Things Journal, IEEE Transactions on Industrial Informatics, Advanced Engineering Informatics and Sensors.

The system failed to parse the two web links of the research group homepage, yet all core research information, paper details and mentor profile provided in the text have been fully translated as required.