Jian Kang’s research while affiliated with Northwestern Polytechnical University and other places

The deflector jet pressure servo valve (DJPSV), a critical component of the aircraft brake servo system, requires a precise foundational model for performance analysis, optimization, and enhancement. However, the complexity of the jet process within the V-groove of the deflector plate presents challenges for accurate mathematical modeling. To address this issue, the paper takes the DJPSV as the research object, carries out detailed mathematical modeling of its components, analyzes the influencing factors of the performance of the key component—the front stage—and optimizes the design of the key factors. First, integrating FLUENT velocity field analysis, this study proposes a novel perspective to rationally simplify and parametrically model the injection process in 3D space. Subsequently, a systematic deduction of the mathematical model for DJPSV is undertaken. Employing the AMESim platform and the secondary development module AMESet, a comprehensive simulation model is constructed, facilitating the study of static-dynamic valve characteristics. Additionally, utilizing the Morris theory and an intelligent algorithm, sensitivity analysis, and structural optimization on the critical component, the pre-stage. The results reveal that the width of the receiving diverter wedge (M), the width of the V-groove outlet (b1), and the distance from the V-groove outlet to the receiving diverter wedge (h) exert the most significant influence on the differential pressure of the pre-stage, which are the key parameters affecting the output differential pressure of the pre-stage. The experiment verifies the accuracy of the simulation model, offering a vital theoretical foundation for valve development and related areas.

Achieving asymptotic and concurrent fault diagnosis in hydraulic system remains a challenging endeavor due to the inherent attributes of the hidden occurrence, simultaneous manifestation, coupling, and limited sample size. To address the above issues, this paper proposes a hierarchical multi-output fault detection and diagnosis framework, namely, HMDF, based on a hierarchical learning strategy to leverage an improved convolutional neural network (CNN) and support vector machine (SVM). Both a multi-channel CNN and a multi-branch CNN are employed to extract and downscale features collected by the sensors at diverse sampling frequencies first, and then, such features are subsequently subjected to classification using SVM. The hierarchical learning strategy enables the identification of different fault states, both at the component and the intra-component level. Additionally, a modified whale optimization algorithm is also utilized to optimize the classification process of SVM. Extensive experiments are conducted to test the proposed HMDF with the hydraulic system datasets. Results show that HMDF achieves a diagnostic accuracy of up to 98.9% for the dataset, surpassing traditional methods reliant on manual extraction of time–frequency features, and it also exhibits superior classification performances with a small sample size. The HMDF is expected to offer a generalized framework for the multi-output fault detection and diagnosis in hydraulic systems and other complex components.

Hydraulic systems play a central role in the transmission and control of various industrial equipment, and the consequences of failures can be severe. The high pressure and closed nature of hydraulic systems, coupled with poor measurability of parameters, limit the applicability of many fault diagnosis methods. In order to achieve automatic fusion of multi-sensor data and accurate fault diagnosis, this paper proposes a novel intelligent fault diagnosis model (IFDM) that combines Variational Mode Decomposition (VMD) with residual networks incorporating attention mechanisms. Utilizing adaptive VMD along with multiscale dispersion entropy allows for the automatic extraction of features from multi-sensor data without the need for specialized knowledge, making it more suitable for industrial applications. The residual networks with attention mechanisms allocate weights to each channel, enabling the model to better learn the relationships between channels and enhance its focus on different fault features. Experimental results demonstrate that the proposed method can accurately diagnose various levels of faults in multiple components of hydraulic systems, achieving a classification accuracy exceeding 99% with superior generalization capabilities.

In this study, we comprehensively investigate the structure and operational principles of the Rotary Direct Drive Electro-Hydraulic Pressure Servo Valve (RDDPV). Our objective is to establish the dynamics equations governing the motor, slide valve, and bias mechanism of the valve. Additionally, we construct a mathematical model for the servo valve controller, while ensuring the linearization of the controller model. Furthermore, we conduct an in-depth analysis of the static characteristics of the valve, including linearity, dead zone, hysteresis loop, and zero drift. Regarding the dynamic characteristics, we establish a dynamic mathematical model for the RDDPV valve. Subsequently, we subject the servo valve to analysis with a focus on frequency response and dynamic response, using the control current as the input and the pressure as the output. To perform these analyses, we employ the software package SIMULINK of MATLAB, facilitating dynamic simulations. Remarkably, the simulation results exhibit the valve's conformity to design requirements, underscoring its suitability for subsequent research and development endeavors. Through our rigorous investigation, we offer essential technical support for the forthcoming stages of the valve's research and development, thereby laying a robust foundation for its further advancement.

According to the characteristics of rotary direct drive electro-hydraulic pressure servo valve, the lumped parameter modeling and three-dimensional (3D) field analysis methods are integrated. The lumped parameter method is used for mathematical modeling for system parts with clear physical concept, and for complicate structures difficult to simply and derive, the 3D mathematical model is employed to analyze the magnetic field, flow field and structural field. Moreover, numerical fitting is carried out to the field analysis results, and the model recognition methods, such as neural network, are used for modeling. Specific mathematical algorithms are used to represent the mathematical relationship between the input parameters (such as structural parameters and environmental parameters) and the feature output parameters. Finally, the algorithm which can be used to evaluate the relationship between the input parameters (such as structural parameters and environmental parameters) of the analyzed object and its static and dynamic performances is obtained. For similar research objects, the user only needs to change the values of input parameters to complete the analysis and evaluation of the static and dynamic performances of system.

This paper proposes a learning-based model predictive control scheme. This scheme divides the predictive model into a known nominal model and an unknown model residual. Model residual is learned using Gaussian process regression. The learned stochastic model is solved quickly using differential dynamic programming, taking into account control input constraints. The simulation results show that compared with state of art optimal control methods, this scheme has good robustness to model residual, accelerates the solution of high-dimensional problems, and can strictly constrain the control inputs according to the actual situation. Based on this learning-based model predictive control scheme, this paper also proposes an online learning gait generator for the uncertainty problem in the locomotion control of biped robots. The zero moment point is strictly constrained during training to ensure safety. The simulation results show that the gait generator is robust to unknown load and unknown external force.

This work proposes a parameters auto-tuning strategy for biped locomotion in whole-body stabilization control (inverse kinematics based and inverse dynamics based) and active impedance control based on Bayesian optimization(BO). Using the domain knowledge, the parameter space is divided into three sub-spaces and optimized by decoupling BO and alternating BO algorithms. The effectiveness of the proposed method is demonstrated in simulation using a torque-controlled biped robot that we developed. The 32 control parameters are tuned in less than 400 evaluations. In addition, the auto-tuned parameters are robust to different top-level velocity inputs and show compliant behavior with balance in push recovery scenarios. To the best of our knowledge, this is the first work to automatically tune the parameters of the three controllers (inverse kinematics, inverse dynamics and active impedance control) jointly.