1,256 publications from this institution
One crucial aspect that yet remains fairly unknown while can inform us about the behavior of deep neural networks is their decision boundaries. Trust can be improved once we understand how and why deep models carve out a particular form of decision boundary and thus make particular decisions. Robustness against adversarial examples is directly related to the decision boundary as adversarial examples are basically 'missed out' by the decision boundary between two classes. Investigating the decision boundary of deep neural networks, nevertheless, faces tremendous challenges. First, how we can generate instances near the decision boundary that are similar to real samples? Second, how we can leverage near decision boundary instances to characterize the behaviour of deep neural networks? Motivated to solve these challenges, we focus on investigating the decision boundary of deep neural network classifiers. In particular, we propose a novel approach to generate instances near decision boundary of pre-trained DNNs and then leverage these instances to characterize the behaviour of deep models.
This paper is concerned with the problem of finite-time <i xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">H</i> <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">∞</sub> synchronization for a class of discrete-time switched neural networks with quantized feedback controllers. First, we take advantage of mode-dependent average dwell time approach and semi-time-dependent (STD) Lyapunov function method to establish new criteria for the finite-time synchronization. Based on the newly developed criteria, a finite-time bounded real lemma is presented and further, the quantized state feedback controller is designed to ensure the finite-time <i xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">H</i> <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">∞</sub> synchronization of the underlying neural networks. The conservatism is further reduced by the proposed mode-dependent and STD controller. Finally, we use two illustrative examples to show the effectiveness and superiority of the results obtained in this paper.
Summary This paper is focused on the problem of adaptive sliding mode control design for uncertain neutral‐type stochastic systems under a prescribed H ∞ performance. A simplified state observer is put forward to estimate the unknown state variables, which could be properly incorporated for establishing a new linear‐type switching surface and the associated adaptive variable structure controller. By virtue of the adaptive control design, unknown matched perturbation and potential uncertainties can be counteracted, and the system trajectories are guaranteed to reach the predefined switching surface within finite moment in almost surely sense, and performance analysis of the closed‐loop dynamics during the sliding surface is carried out with a specified H ∞ performance. At last, two illustrative examples through computer simulations are provided to verify the effectiveness and applicability of the proposed scheme.
In this paper, the problem of model predictive control for drum water level of boiler systems is investigated. The parameter uncertainties are time-varying norm-bounded and the non-linearity is assumed to satisfy the boundedness condition. The aim is to design a state-feedback controller which minimizes an upper bound on a quadratic objective function at each sampling instant. The hard constraint on the variance of the input is also considered. By linear matrix inequality (LMI) approach, sufficient conditions are obtained, which guarantee the robustly asymptotic stability of the closed-loop feedback system. An example of a boiler drum system is included to demonstrate the effectiveness of the proposed techniques.
With the rapid development of automation technologies, aerospace engineering systems, including aircraft, satellite, and spacecraft, have become increasingly susceptible to system/component malfunctions. Failure to take appropriate responses to even relatively minor defects can result in highly destructive events. A conventional feedback control design may result in an unsatisfactory performance or even instability in the event of malfunctions. Because of this, fault diagnosis (FD) and fault-tolerant control (FTC) technologies that can ensure the safety of handicapped systems have attracted significant interest. FTC design and relevant techniques have provided a flexible framework for dealing with these challenges since the 1970s. There has been significant progress since the 1970s by the active research community, through symposiums and seminars, as well as the vast number of publications on the subject. The research illustrates that FD and FTC are effective and applicable in many engineering plants, especially for aerospace engineering systems. However, it still remains a challenging research area in applications relating to aircraft, spacecraft, and satellites.
This paper addresses the problem of delay-dependent robust and reliable, H <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">∞</sub> static output feedback (SOF) control for uncertain discrete-time piecewise-affine (PWA) systems with time-delay and actuator failure in a singular system setup. The Markov chain is applied to describe the actuator faults behaviors. In particular, by utilizing a system augmentation approach, the conventional closed-loop system is converted into a singular PWA system. By constructing a mode-dependent piecewise Lyapunov-Krasovskii functional, a new H <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">∞</sub> performance analysis criterion is then presented, where a novel summation inequality and S-procedure are succeedingly employed. Subsequently, thanks to the special structure of the singular system formulation, the PWA SOF controller design is proposed via a convex program. Illustrative examples are finally given to show the efficacy and less conservatism of the presented approach.
Logistics network represents a complex system where different elements that are logistic locations interact with each other. This interaction contains delays caused by time needed for delivery of the material. In this paper, we study local input-to-state stability of such logistics networks. Their behaviour is described by a functional differential equation with a constant time-delay. An appropriate Lyapunov–Razumikhin function and the small gain condition are utilized to establish some conditions for stability analysis of the network under consideration. Our stability conditions for the logistics network are based on the information about the interconnection properties between logistic locations and their production rates. Finally, numerical results are provided to demonstrate an application of the proposed approach.
Large wind farms are gaining prominence due to increasing dependence on renewable energy. In order to operate these wind farms reliably and efficiently, advanced maintenance strategies such as condition based maintenance are necessary. However, wind turbines pose unique challenges in terms of irregular load patterns, intermittent operation and harsh weather conditions, which have deterring effects on life of rotating machinery. This paper reviews the state-of-the-art in the area of diagnostics and prognostics pertaining to two critical failure prone components of wind turbines, namely, low-speed bearings and planetary gearboxes. The survey evaluates those methods that are applicable to wind turbine farm-level health management and compares these methods on criteria such as reliability, accuracy and implementation aspects. It concludes with a brief discussion of the challenges and future trends in health assessment for wind farms.
No abstract is provided for this article.
In this paper, we investigate the design of distributed damping systems (DDSs) for the overall seismic protection of multiple adjacent buildings. The considered DDSs contain interstory dampers implemented inside the buildings and also interbuilding damping links. The design objectives include mitigating the buildings seismic response by reducing the interstory-drift and story-acceleration peak-values and producing small interbuilding approachings to decrease the risk of interbuilding collisions. Designing high-performance DDS configurations requires determining convenient damper positions and computing proper values for the damper parameters. That allocation-tuning optimization problem can pose serious computational difficulties for large-scale multibuilding systems. The design methodology proposed in this work—(i) is based on an effective matrix formulation of the damped multibuilding system; (ii) follows an H ∞ approach to define an objective function with fast-evaluation characteristics; (iii) exploits the computational advantages of the current state-of-the-art genetic algorithm solvers, including the usage of hybrid discrete-continuous optimization and parallel computing; and (iv) allows setting actuation schemes of particular interest such as full-linked configurations or nonactuated buildings. To illustrate the main features of the presented methodology, we consider a system of five adjacent multistory buildings and design three full-linked DDS configurations with a different number of actuated buildings. The obtained results confirm the flexibility and effectiveness of the proposed design approach and demonstrate the high-performance characteristics of the devised DDS configurations.
This paper presents a novel method to address a Proportional Integral observer design for the actuator and sensor faults estimation based on Takagi–Sugeno fuzzy model with unmeasurable premise variables. The faults are assumed as time-varying signals whose kth time derivatives are bounded. Using Lyapunov stability theory and L 2 performance analysis, sufficient design conditions are developed for simultaneous estimation of states and time-varying actuator and sensor faults. The Proportional Integral observer gains are computed by solving the proposed conditions under Linear Matrix Inequalities constraints. A simulation example is provided to illustrate the effectiveness of the proposed approach.
This chapter is dedicated to present some methods for vibration protection of statues and cultural heritage objects against earthquakes and ambient vibrations. The chapter opens with short explanation about the dynamic performance test that has been done on a full-scale copy of the famous statue of Michelangelo Buonarroti Pieta Rondanini and its isolation system. Based on the results of the tests, a multidegree-of-freedom model is developed. This updated model is used to verify the proposed vibration control strategies. To reduce the level of vibration transmitted to the statue, combination of inerter with TMD (TMDI) is proposed. Optimal parameters of the passive device are found by using numerical methods. Furthermore, a scaled isolation system is designed and manufactured according to the test results of the full-scale structure. To improve the performance of the isolator, a designed and manufactured ball-screw type inerter is introduced to the scaled structure. The effectiveness of the manufactured inerter on the dynamic behavior of the isolation system is demonstrated by using dynamic performance tests on a shaking table. The results provide some numerical and experimental results considering inerter-based methods in order to vibration protection of statues and cultural heritage.
This paper focuses on the fuzzy variable structure control for uncertain systems with disturbance. Specifically, the fuzzy control is introduced to estimate the control disturbance, the switching control is included to compensate for the approximation error, and they possess the characteristic of simpleness in design and effectiveness in attenuating the control chattering. Some typical numerical examples are presented to demonstrate the effectiveness and advantage of the fuzzy variable structure controller proposed.
