A Siamese hybrid neural network framework for few-shot fault diagnosis of fixed-wing unmanned aerial vehicles

Li Shaobo,Li Chuanjiang,Zhang Ansi,Yang Lei,Zio Enrico,Pecht Michael,Gryllias Konstantinos 한국CDE학회 2022 Journal of computational design and engineering Vol.9 No.4

As fixed-wing unmanned aerial vehicles (FW-UAVs) are used for diverse civil and scientific missions, failure incidents are on the rise. Recent rapid developments in deep learning (DL) techniques offer advanced solutions for fault diagnosis of unmanned aerial vehicles. However, most existing DL-based diagnostic models only perform well when trained on massive amounts of labeled data, which are challenging to collect due to the complexity of the FW-UAVs systems and service environments. To address these issues, this paper presents a novel framework, Siamese hybrid neural network (SHNN), to achieve few-shot fault diagnosis of FW-UAVs in an intelligent manner. “State map” strategy is firstly proposed to transform raw flight data into similar and dissimilar sample pairs as input. The proposed SHNN framework consists of two identical networks that share weights with each other, and each subnetwork is designed with a hybrid one-dimensional conventional neural network and long short-term memory model as feature encoder, whose generated feature embedding is used to measure the similarity of input pairs via a distance function in the metric space. In comprehensive experiments on a real flight dataset of an FW-UAV, the SHNN framework achieves competitive results compared to other models, demonstrating its effectiveness in both binary and multi-class few-shot fault diagnosis.

An Extended Generative Feature Learning Algorithm for Image Recognition

( Bin Wang ),( Chuanjiang Li ),( Qian Zhang ),( Jifeng Huang ) 한국인터넷정보학회 2017 KSII Transactions on Internet and Information Syst Vol.11 No.8

Image recognition has become an increasingly important topic for its wide application. It is highly challenging when facing to large-scale database with large variance. The recognition systems rely on a key component, i.e. the low-level feature or the learned mid-level feature. The recognition performance can be potentially improved if the data distribution information is exploited using a more sophisticated way, which usually a function over hidden variable, model parameter and observed data. These methods are called generative score space. In this paper, we propose a discriminative extension for the existing generative score space methods, which exploits class label when deriving score functions for image recognition task. Specifically, we first extend the regular generative models to class conditional models over both observed variable and class label. Then, we derive the mid-level feature mapping from the extended models. At last, the derived feature mapping is embedded into a discriminative classifier for image recognition. The advantages of our proposed approach are two folds. First, the resulted methods take simple and intuitive forms which are weighted versions of existing methods, benefitting from the Bayesian inference of class label. Second, the probabilistic generative modeling allows us to exploit hidden information and is well adapt to data distribution. To validate the effectiveness of the proposed method, we cooperate our discriminative extension with three generative models for image recognition task. The experimental results validate the effectiveness of our proposed approach.

Efficient walking gait with different speed and step length: Gait strategies discovered by dynamic optimization of a biped model

Kang An,Chuanjiang Li,Zuhua Fang,Chengju Liu 대한기계학회 2017 JOURNAL OF MECHANICAL SCIENCE AND TECHNOLOGY Vol.31 No.4

The selection of walking gait for biped robots depends on the requirement of walking environment. Walking with different situations of walking speeds and step lengths, the gait strategies are different. In this paper, we study the energetically optimal walking gait strategies under the different walking situations using a simple biped walking model with dynamic optimization method. The walking model with mass legs and three actuations, which is designed upon Srinivasan’s model, is built for the purpose of the paper. Dynamic optimization is used for a free search with minimal constraints. The analysis of the COT of the optimal gaits and its two components COT swing and COT push-off show that the COT is increasing with the increase of the walking speed. For a certain walking speed, the minimal value of COT can be found with a corresponding step length. According to the joint torques output strategies, we discover four gait patterns including two typical walking gaits patterns that the hip torque impulse is only at the beginning or at the end of the swing phase, respectively, and two other new transitional gait patterns.

Consensus and Security Control of Multi-agent Systems Based on Set-membership Estimation with Time-varying Topology under Deception Attacks

Yanfei Zhu,Hang Liu,Chuanjiang Li,Jiahao Yu 제어·로봇·시스템학회 2022 International Journal of Control, Automation, and Vol.20 No.11

This paper is concerned with the consensus and security of leader-follower multi-agent systems with time-varying topology, where each follower’s sensor-to-controller channel and controller-to-actuator channel are subject to unknown but bounded (UBB) random deception attacks. A novel set-membership estimation and control framework is presented to realize system consensus, input-to-state stability and security performance, in which each controller receives arbitrary estimates from itself and neighbor set-membership estimators. Based on the Lyapunov functional and stochastic control method, a sufficient condition on the existence of estimators and controllers is derived. Then, a convex optimization algorithm is established to deal with the proposed framework. Finally, the effectiveness of the method is verified by numerical simulation.

Simultaneous Stability of Large-scale Systems via Distributed Control Network with Partial Information Exchange

Yanfei Zhu,Fuwen Yang,Chuanjiang Li,Yilian Zhang 제어·로봇·시스템학회 2018 International Journal of Control, Automation, and Vol.16 No.4

This paper is concerned with the simultaneous stability of the multi-mode large-scale systems composed of the interaction subsystems. A novel distributed control network consisting of multiple network-based controllers with the partial information exchange is adopted to simultaneously stabilize the large-scale systems in multiple operation modes. In the distributed control network (DCN), a partial state information exchange approach is developed to save the real-time communication and computation resources. To compensate for the effects of dynamic couplings between interaction subsystems, the designed controllers use both the local states and the neighbors’ partial information with packet dropouts for local feedback design. Then, a series of Lyapunov functions are constructed to derive a matrix-inequality-based sufficient condition for the existence of the desired controllers. Based on an orthogonal complement technique, the gains of the controllers in DCN are parameterized. The iterative algorithm for the solution of simultaneous stabilization problem is also developed. Finally, a numerical example is performed to show the relevant feature of the proposed method.

Controlling Dynamic Formations of Mobile Agents Governed by Euler-Lagrange Dynamics

Liangming Chen,Qingkai Yang,Chuanjiang Li,Guangfu Ma 제어·로봇·시스템학회 2021 International Journal of Control, Automation, and Vol.19 No.5

This paper studies the problem of controlling dynamic formations of mobile agents governed by EulerLagrange dynamics. Here a formation is said to be dynamic if as time evolves, the desired formation undergoes translation, scaling and rotation. First, a constant-gain formation control algorithm is designed such that all agents can converge to the desired dynamic formation, in which the graphic information is needed for the selection of constant gains. Then, another fully distributed formation control algorithm is further proposed by employing variablegain control techniques, which enables each agent to be independent of the knowledge of the overall interaction graph needed otherwise in the control gain. Instead of moving with a desired translational velocity, a centroidtracking formation control algorithm is also proposed such that the centroid of the formation tracks a desired trajectory. The parametric uncertainties are taken into consideration in the proposed formation control algorithms. Finally, simulation examples are provided to validate the effectiveness of the proposed control algorithms.

Construction of fat1 Gene Expression Vector and Its Catalysis Efficiency in Bovine Fetal Fibroblast Cells

Liu, Boyang,Yang, Runjun,Li, Junya,Zhang, Lupei,Liu, Jing,Lu, Chunyan,Lian, Chuanjiang,Li, Zezhong,Zhang, Yong-Hong,Zhang, Liying,Zhao, Zhihui Asian Australasian Association of Animal Productio 2012 Animal Bioscience Vol.25 No.5

The FAT-1 protein is an n-3 fatty acid desaturase, which can recognize a range of 18- and 20-carbon n-6 substrates and transform n-6 polyunsaturated fatty acids (PUFAs) into n-3 PUFAs while n-3 PUFAs have beneficial effect on human health. Fat1 gene is the coding sequence from Caenorhabditis elegans which might play an important role on lipometabolism. To reveal the function of fat1 gene in bovine fetal fibroblast cells and gain the best cell nuclear donor for transgenic bovines, the codon of fat1 sequence was optimized based on the codon usage frequency preference of bovine muscle protein, and directionally cloned into the eukaryotic expression vector pEF-GFP. After identifying by restrictive enzyme digests with AatII/XbaI and sequencing, the fusion plasmid pEF-GFP-fat1 was identified successfully. The pEF-GFP-fat1 vector was transfected into bovine fetal fibroblast cells mediated by Lipofectamine2000$^{TM}$. The positive bovine fetal fibroblast cells were selected by G418 and detected by RT-PCR. The results showed that a 1,234 bp transcription was amplified by reverse transcription PCR and the positive transgenic fat1 cell line was successfully established. Then the expression level of fat1 gene in positive cells was detected using quantitative PCR, and the catalysis efficiency was detected by gas chromatography. The results demonstrated that the catalysis efficiency of fat1 was significantly high, which can improve the total PUFAs rich in EPA, DHA and DPA. Construction and expression of pEF-GFP-fat1 vector should be helpful for further understanding the mechanism of regulation of fat1 in vitro. It could also be the first step in the production of fat1 transgenic cattle.

The Physiological Occlusion of the Central Canal May Be a Prerequisite for Syringomyelia Formation

Chuan Jiang,Xinyu Wang,Chunli Lu,Qian Li,Longbing Ma,Wei Li,Shengyu Cui,Kang Li,Xiang Wang,Yuxin Feng,Fengzeng Jian 대한척추신경외과학회 2023 Neurospine Vol.20 No.4

Objective: Syringomyelia is a common central nervous system disease characterized by the dilation of the central canal (CC). Regarding the pathogenesis of syringomyelia, cerebrospinal fluid (CSF) circulation obstruction in the subarachnoid space (SAS) of the spinal cord has been widely accepted. However, clinical and animal studies on obstructing the CSF in SAS failed to form syringomyelia, challenging the theory of SAS obstruction. The precise pathogenesis remains unknown. Methods: We utilized an extradural compression rat model to investigate the pathogenesis underlying syringomyelia. Magnetic resonance imaging enabled detection of syringomyelia formation. To assess CSF flow within the SAS, Evans blue was infused into the cisterna magna. Histological analysis allowed morphological examination of the CC. Furthermore, CSF flow through the CC was traced using Ovalbumin Alexa-Flour 647 conjugate (OAF-647). Scanning electron microscopy (SEM) enabled visualization of ependymal cilia. Results: The findings showed that the dura mater below the compression segment exhibited lighter coloration relative to the region above the compression, indicative of partial obstruction within the SAS. However, the degree of SAS occlusion did not significantly differ between syringomyelia (SM-Y group) and those without (SM-N group). Intriguingly, hematoxylin and eosin staining and CSF tracing revealed occlusion of the CC accompanied by reduced CSF flow in the SM-Y group compared to SM-N and control groups. SEM images uncovered impairment of ependymal cilia inside the syringomyelia. Conclusion: CC occlusion may represent a physiological prerequisite for syringomyelia formation, while SAS obstruction serves to initiate disease onset. The impairment of ependymal cilia appears to facilitate progression of syringomyelia.