As a widely adopted control strategy in industry, PID control is widely recognized for its simplicity, efficiency and reliability. However, in the face of complex and variable nonlinear industrial controlsystems, tra...
详细信息
The use of the deep Convolutional Neural network (CNN) in breast cancer classification of mammogram images has been widely investigated to aid radiologists in better clinical diagnoses. Multiple levels of convolution ...
详细信息
ISBN:
(纸本)9798350372113;9798350372106
The use of the deep Convolutional Neural network (CNN) in breast cancer classification of mammogram images has been widely investigated to aid radiologists in better clinical diagnoses. Multiple levels of convolution and non-linearity repetitions in CNN's architecture are required to extract significant data to be represented. However, the vanishing gradient effect occurs when deeper network training as a product of the partial derivative of loss function on each weightage update can cause no meaningful network learning, even with additional epochs. Overcoming this using the activation function of rectified linear unit (ReLU) by allowing neurons to be activated to allow non-linearity when the output is more than zero could lessen the problem. However, restrictive allowance of non-linearity for <0 for final feature extraction when producing output probability on highly complex data such as mammogram images leads to dropped networkperformance. To overcome this, this study proposed an adaptive ReLU based on genetic algorithm (GA) profiling to determine the best threshold value for allowing neuron activation based on mutation and adaptation to improve the restrictive capability of the original ReLU. We modified the adaptive ReLU on the final learning layer of two CNN architectures and observed the performance on a public mammogram dataset of INbreast. Our experiments show improved accuracy from 95.0% to 98.5% and improved classification performance compared to other well-known activation functions. Applying evolutionary-based GAs to activation functions can represent an exciting frontier in meta-learning for neural networks.
This article investigates the problem of predefined time prescribed performance tracking control for robotic manipulators, considering bounded external disturbances and unknown input saturation. At first, a Gaussian e...
详细信息
This article investigates the problem of predefined time prescribed performance tracking control for robotic manipulators, considering bounded external disturbances and unknown input saturation. At first, a Gaussian error function is utilized as a replacement for non-smooth unknown input saturation. Then, a predefined-time performance function is designed to achieve the predefined convergence. Based on this foundation, a controller is developed utilizing a predefined-time terminal sliding surface. The position tracking error exhibits globally predefined time convergence and prescribed transient steady state performance. Stability analysis of the designed controller is conducted utilizing the Lyapunov method. Finally, numerical simulations are employed to verify the effectiveness and superiority of the proposed scheme.
In the realm of smart manufacturing, Industrial controlsystems (ICSs) play a critical role in supervising essential operations across various manufacturing facilities. As ICSs become increasingly integral to critical...
详细信息
ISBN:
(纸本)9798350363029;9798350363012
In the realm of smart manufacturing, Industrial controlsystems (ICSs) play a critical role in supervising essential operations across various manufacturing facilities. As ICSs become increasingly integral to critical infrastructure, their susceptibility to cyber threats escalates. Despite the positive impact of abundant digital data communication on ICS performance, rising concerns about data security underscore the pivotal role of intrusion detection systems (IDS) in averting the consequences of network security attacks. However, challenges persist, particularly in the detection of zero-day attacks. Furthermore, while many IDSs exhibit high performance for known attack patterns, they often function as black boxes, presenting interpretability challenges for network operators to take appropriate actions. This paper introduces a novel zero-day intrusion detection system, thoughtfully crafted to address the interdependence within ICS network traffic. Our approach designs a multi-head-attention mechanism, not only for precise classification of network traffic but also to mitigate zero-day attacks. Subsequently, we incorporate a novel autoencoder architecture, purposefully designed to capture distinctive patterns present in both known attacks and normal traffic, leveraging the output from the attention layer. Furthermore, our approach integrates Explainable Artificial Intelligence mechanisms, employing the multi-head attentions model to offer a more detailed description of the underlying process involved in traffic categorization. The model's efficacy is demonstrated through experiments conducted using water system ICS testbeds, underscoring its performance and reliability in detecting diverse cyber-attacks and inferring zero-day attacks with an accuracy surpassing 89%, while also providing interpretable results.
A DC motor is a common actuator in process controlsystems that converts electrical energy into mechanical energy. This paper examined the performance of a deep learning-based neural network predictive controller (NNP...
详细信息
The comprehensive performance and user experience of large-scale distributed video conferencesystems are constrained by network traffic, and conventional network services are difficult to ensure the stability and rel...
详细信息
This paper is concerned with the adaptive resilient event-triggered control problem for networked switched systems with nonperiodic denial of service (DoS) attacks. In order to save limited network bandwidth and reduc...
详细信息
ISBN:
(纸本)9798350387780;9798350387797
This paper is concerned with the adaptive resilient event-triggered control problem for networked switched systems with nonperiodic denial of service (DoS) attacks. In order to save limited network bandwidth and reduce the negative impact on system performance caused by DoS attacks, a novel adaptive resilient event-triggered scheme (ARETS) is developed, which can save the communication cost and improve the flexibility of the system. With considering the ARETS, dwell time switching and transmission delays, the criterion of asymptotic stability and L2-gain performance is derived such that the system under DoS attacks owns security performance. On this basis, the co-design conditions of the ARETS and the controller are proposed. Finally, a simulation example is given to verify the effectiveness and merits of the proposed method.
Advent of new communication technologies made it possible to replace point-to-point control structures with networked systems as a breakthrough toward Industry 4.0. However, overcoming communication imperfections is s...
详细信息
ISBN:
(纸本)9798350376357;9798350376340
Advent of new communication technologies made it possible to replace point-to-point control structures with networked systems as a breakthrough toward Industry 4.0. However, overcoming communication imperfections is still a major issue. In this paper, a group of network-based cascade controlsystems is investigated in which, network communication is considered within the controller-actuator channel. In order to reduce conservatism and increase robustness of the control system, time-varying delay and packet dropouts are considered as separate terms using an augmented continuous-time state-space model. Based on Lyapunov theorem, the H-infinity stability conditions are attained dealing with external disturbances. An identified linear model for an industrial gas turbine unit is employed for performance assessment of the designed controllers. Obtained results confirm capability of the proposed method to achieve larger gain margins and more robustness against uncertain delays and data loss.
The growth of network infrastructure over the years, including the increasing number of devices and users connecting to networks, the increasing complexity of network environments, and the need for more flexible and e...
详细信息
ISBN:
(纸本)9781665493932
The growth of network infrastructure over the years, including the increasing number of devices and users connecting to networks, the increasing complexity of network environments, and the need for more flexible and efficient network management, has led to the emergence of Software-Defined networking (SDN) as a feasible means of support. SDN is a networking approach that uses software to abstract, manage, and control the network infrastructure. In SDN, the control plane, which determines how data is routed through the network, is separated from the data plane, which forwards the traffic. This separation allows the control plane to be more flexible and programmable, enabling administrators to easily configure and manage the network using the software. There are several routing algorithms and techniques that can be used to improve networkperformance in Software-Defined networking. Shortest Path Routing is a common routing algorithm along with Segment Routing and may be capable of speeding up the network's overall performance by decreasing the time it takes for traffic to reach its destination. Load Balancing is a technique that involves distributing traffic across multiple resources, such as servers or network devices, to optimize the use of resources and improve the performance of the network. For this, there exist several techniques such as Round Robin and Ant Colony Optimization. Another such technique used to optimize the use of resources and improve the overall performance of the network is Traffic Engineering-where software is used to route traffic through the network based on various criteria, such as the type of traffic, the source, and destination of the traffic, or the current load on the network. There have been numerous studies in these fields investigating the feasibility of novel systems and solutions that have been proposed for improving networkperformance. This paper goes into the methodologies followed for these systems and the impacts they have had
Many medical scenarios require the accurate and rapid infusion of medical liquids, such as blood, saline, and medications. However, the centralized platforms used in existing infusion systems struggle to balance perfo...
详细信息
ISBN:
(纸本)9798400709746
Many medical scenarios require the accurate and rapid infusion of medical liquids, such as blood, saline, and medications. However, the centralized platforms used in existing infusion systems struggle to balance performance and cost, making it difficult to achieve accurate control of liquid flow in these situations. To address this challenge, this paper proposes an accurate-rapid infusion system based on a distributed network. Taking performance, safety, and cost into account, the system utilizes a pressure bag as the main actuator, and a corresponding mathematical model is developed. A distributed measurement & controlnetwork is then constructed to execute complex control algorithms. Unlike centralized platforms, the distributed network can be constructed quickly and flexibly, enabling computility sharing for edge computing. Experimental results demonstrate that the system controls liquid flow accurately and stably in the steady state, with the error mean of all algorithms being less than 9.7 mL/min and the standard deviation (STD) less than 7.9 mL/min. This paper introduces a novel method for future research on rapid infusion systems.
暂无评论