Ashraf Saad

Evolving an artificial neural network classifier for condition monitoring of rotating mechanical systems

We present the results of our investigation into the use of genetic algorithms (GAs) for identifying near optimal design parameters of diagnostic systems that are based on artificial neural networks (ANNs) for condition monitoring of mechanical systems. ANNs have been widely used for health diagnosis of mechanical bearing using features extracted from vibration and acoustic emission signals. However, different sensors and the corresponding features exhibit varied response to different faults. Moreover, a number of different features can be used as inputs to a classifier ANN. Identification of the most useful features is important for an efficient classification as opposed to using all features from all channels, leading to very high computational cost and is, consequently, not desirable. Furthermore, determining the ANN structure is a fundamental design issue and can be critical for the classification performance. We show that a GA can be used to select a smaller subset of features that together form a genetically fit family for successful fault identification and classification tasks. At the same time, an appropriate structure of the ANN, in terms of the number of nodes in the hidden layer, can be determined, resulting in improved performance.

Hybrid intelligent systems for network security

Society has grown to rely on Internet services, and the number of Internet users increases every day. As more and more users become connected to the network, the window of opportunity for malicious users to do their damage becomes very great and lucrative. The computer industry is combating the rising threat of malicious activity with new hardware and software products such as Intrusion Detection Systems, Intrusion Prevention Systems, and Firewalls. However, malicious users are constantly looking for ways to by-pass the security features of these products, and many times they will succeed. This paper describes a novel concept implemented for the purpose of computer and network security with hopes of using it to combat malicious user activity. A hybrid-intelligent system based on Bayesian Learning Networks and Self-Organizing Maps was created and used for classifying network and host based data collected within a Local Area Network. The KDD-CUP-99 data set was used to test this classification system, and the experimental results show that there is an advantage to using a hybrid system such as this because there was a significant improvement in classification accuracy compared to a non-hybrid Bayesian Learning approach when network-only data is used for classification.

An IEEE 1394 - FireWire - based embedded video system for surveillance applications

In order to address the need for portable inexpensive systems for video surveillance, we built a computer vision system that provides digital video data transfer from a CCD camera using embedded software/hardware via the IEEE 1394 protocol, also known as FireWire or i.Link, and Ethernet TCP/IP interfaces. Controlled by an extended version of the IEEE 1394-based digital camera specification (DCAM), the system is made up of entirely open interfaces that provide many image resolutions and frame rates, and digital signal processing and compression capabilities. We present the design of the system and its performance on running a digital signal processing application. We also give details about the software and hardware architecture, as well as our directions for future research and development.

A knowledge visualization tool for teaching and learning computer engineering knowledge, concepts, and skills

Current educational environments rely primarily on a text-based syllabus approach to describing course content as well as whole curricula. This traditional approach is too limiting. It fails to delineate the relationship of concepts and skills covered in a given course to those covered by other courses. It also fails to show the knowledge base that a learner will have acquired at the end of his/her course of study, as such base relates to lifelong learning. New demands placed on leaching and learning environments call for a more engaging experience for learners. The approach we have adopted is based on using a visualization tool that can engage both learners and instructors in an active learning approach to construct a map for the knowledge, concepts, and skills that a learner possesses and those that he/she acquires during each course of the computer engineering curriculum. We present our design for such a tool that is currently under development.

Mobile robotics as the platform for undergraduate capstone electrical and computer engineering design projects

Research grade mobile robots are no longer restricted to robotics laboratories. Now that they are available and affordable through several vendors, they can be used for undergraduate capstone electrical and computer engineering design projects. Several important educational considerations have supported this decision, including: providing students with a platform for their projects that enables integrating both hardware and software development as well as dealing with real-time programming issues. We present our experience in using the AmigoBot from ActiveMedia for capstone senior design projects. Our work has focused on integrating a variety of sensors to the robot control architecture, most notably the integration of an IEEE 1394 Firewire - digital CCD camera. That in turn has enabled students to develop and test computer vision algorithms. Additional system capabilities included wireless communication of video streams captured by the digital camera. Integrating the ability to remotely control the AmigoBot can also enable using it for indoor surveillance applications.

Genetic Algorithms for Artificial Neural Net-based Condition Monitoring System Design for Rotating Mechanical Systems

Stable single film bipolar membranes of prolonged life and improved performance particularly for use in electrodialysis water splitting process, are prepared by introducing a more stable interface in the membrane structure. After the cationic exchange groups are preformed on an insoluble cross-linked aromatic polymeric matrix, the dissociable anionic exchange groups may be introduced more intimately chemically bonded in position by using multi-functional compounds containing mixed tertiary, secondary and primary amine groups, so that the resulting interface is comparatively more stable, and is less likely to be neutralized, therefore, attaining longer life-time and a higher level of performance.

Genetic Algorithm-Evolved Bayesian Network Classifier for Medical Applications

This paper presents the development of a Bayesian Network (BN) classifier for a medical application. Patient age classification is based on statistical features extracted from electrocardiogram (ECG) signals. The computed ECG features are converted to a discrete form to lower the dimensionality of the signal and to allow for conditional probabilities to be calculated for the BN. Two methods of network discovery from data were developed and compared: a greedy hill-climb search and a search method based on evolutionary computing. The performance comparison of these two methods for network structure discovery shows a large increase in classification accuracy with the GA-evolved BN as measured by the area under the curve of the Receiver Operating Characteristic curve.

National science foundation advanced technological education projects and centers of excellence for information technology education: an overview

This paper gives an overview of initiatives for information technology (IT) education undertaken by projects and centers of excellence for Advanced Technological Education (ATE) with support from the National Science Foundation (NSF). While these centers have a common goal to advance IT education regionally and nationally, they differ in the means of doing so. Several aspects of these projects and centers are highlighted, including challenges and opportunities that are beneficial to other ATE projects and centers. Current issues pertinent to IT curriculum development, including curriculum guidelines and accreditation of IT programs, are presented and discussed. The paper concludes with recommendations and future steps that are of import to IT educators at large. This paper will be presented at SIGITE 2005 in conjunction with a panel discussion involving principal investigators of current NSF ATE projects and centers for IT education.

Generic architecture for power aware routing in wireless sensor networks

We describe the design and implementation of a generic architecture to provide a collective solution for power-aware routing to a wide range of problems in wireless sensor network environments. Power aware-routing is integral to the solutions proposed for a variety of wireless sensor network problems. This can lead to challenging technical problems and deployment barriers when attempting to integrate the proposed solutions. We have identified coupling between sources and sinks as the main factor for different design choices for a range of problems. We developed a core-based hierarchical routing framework and protocol for efficient power-aware routing that is used to decouple the sources from sinks. We show the efficiency of our approach via simulation.

An evolutional network architecture for developmental knowledge bases

We present a network architecture that supports developmental knowledge bases. Knowledge representation in our architecture is based on knowledge quanta. The network architecture described herein is evolutional in nature and utilizes spreading activation algorithms for network formation. In doing so, our research is seeking answers to fundamental questions pertinent to the way people learn as well.