Sagar Kamarthi

Feature extraction from wavelet coefficients for pattern recognition tasks

An efficient feature extraction method based on the fast wavelet transform is presented. The paper especially deals with the assessment of process parameters or states in a given application using the features extracted from the wavelet coefficients of measured process signals. Since the parameter assessment using all wavelet coefficients will often turn out to be tedious or leads to inaccurate results, a preprocessing routine that computes robust features correlated to the process parameters of interest is highly desirable. The method presented divides the matrix of computed wavelet coefficients into clusters equal to row vectors. The rows that represent important frequency ranges (for signal interpretation) have a larger number of clusters than the rows that represent less important frequency ranges. The features of a process signal are eventually calculated by the Euclidean norms of the clusters. The effectiveness of this new method has been verified on a flank wear estimation problem in turning processes and on a problem of recognizing different kinds of lung sounds for diagnosis of pulmonary diseases.

Accelerating neural network training using weight extrapolations

The backpropagation (BP) algorithm for training feedforward neural networks has proven robust even for difficult problems. However, its high performance results are attained at the expense of a long training time to adjust the network parameters, which can be discouraging in many real-world applications. Even on relatively simple problems, standard BP often requires a lengthy training process in which the complete set of training examples is processed hundreds or thousands of times. In this paper, a universal acceleration technique for the BP algorithm based on extrapolation of each individual interconnection weight is presented. This extrapolation procedure is easy to implement and is activated only a few times in between iterations of the conventional BP algorithm. This procedure, unlike earlier acceleration procedures, minimally alters the computational structure of the BP algorithm. The viability of this new approach is demonstrated on three examples. The results suggest that it leads to significant savings in computation time of the standard BP algorithm. Moreover, the solution computed by the proposed approach is always located in close proximity to the one obtained by the conventional BP procedure. Hence, the proposed method provides a real acceleration of the BP algorithm without degrading the usefulness of its solutions. The performance of the new method is also compared with that of the conjugate gradient algorithm, which is an improved and faster version of the BP algorithm.

Optimal pricing of reusable and recyclable components under alternative product acquisition mechanisms

Product recovery facilities (PRFs), which process discarded product returns as well as sell the recovered components, play a vital role in the promotion of product reuse and recycle. The financial woes of many PRFs can be attributed to the product recovery costs and the inventory control of recovered components. Fluctuations in the demand for recovered components and unpredictability of the pattern and timing of discarded product returns make inventory management difficult. Pricing of recovered components is an effective strategy to control inventory and boost the revenues. This work determines the optimal prices of reusable and recyclable components when a PRF has to adhere to a legislation which limits the disposal quantity. In the first and second scenarios considered, the PRF passively accepts product returns, whereas in the third and fourth, it proactively acquires them. Single type discarded products are processed in the first and third scenarios and multi-type products in the second and fourth. An analysis is carried out to study the effect of return quantities, component yield, product yield, disposal regulation, and the product recovery, holding, and disposal costs on the prices of reusable and recyclable components and the following performance metrics: inventory levels, disposal quantities, and overall profit.

Solving large scale disassembly line balancing problem with uncertainty using reinforcement learning

Due to increasing environmental concerns, manufacturers are forced to take back their products at the end of products’ useful functional life. Manufacturers explore various options including disassembly operations to recover components and subassemblies for reuse, remanufacture, and recycle to extend the life of materials in use and cut down the disposal volume. However, disassembly operations are problematic due to high degree of uncertainty associated with the quality and configuration of product returns. In this research we address the disassembly line balancing problem (DLBP) using a Monte-Carlo based reinforcement learning technique. This reinforcement learning approach is tailored fit to the underlying dynamics of a DLBP. The research results indicate that the reinforcement learning based method is able to perform effectively, even on a complex large scale problem, within a reasonable amount of computational time. The proposed method performed on par or better than the benchmark methods for solving DLBP reported in the literature. Unlike other methods which are usually limited deterministic environments, the reinforcement learning based method is able to operate in deterministic as well as stochastic environments.

Wavelet networks for sensor signal classification in flank wear assessment

It is known that the force and vibration sensor signals in a turning process are sensitive to the gradually increasing flank wear. Based on this fact, this paper investigates a flank wear assessment technique in turning through force and vibration signals. Mainly to reduce the computational burden associated with the existing sensor-based methods for flank wear assessment, a so-called wavelet network is investigated. The basic idea in this new method is to optimize simultaneously the wavelet parameters (that represent signal features) and the signal-interpretation parameters (that are equivalent to neural network weights) to eliminate the feature extraction phase without increasing the computational complexity of the neural network. A neural network architecture similar to a standard one-hidden-layer feedforward neural network is used to relate sensor signal measurements to flank wear classes. A novel training algorithm for such a network is developed. The performance of this n ew method is compared with a previously developed flank wear assessment method which uses a separate feature extraction step. The proposed wavelet network can also be useful for developing signal interpretation schemes for manufacturing process monitoring, critical component monitoring, and product quality monitoring.

Performance evaluation of solar photovoltaic arrays including shadow effects using neural network

This paper proposes a neural network based approach to estimating the maximum possible output power of a solar photovoltaic array under the non-uniform shadow conditions at a given geographic location. Taking the solar irradiation levels, the ambient temperature, and the Suns position angles as inputs, a multilayer feed-forward neural network estimates the output power of the solar photovoltaic array. Training data for the neural network is generated by conducting a series of experiments on a shaded solar panel at different hours of a day for several days. After training the neural network, its accuracy and generalization properties are verified on test data. It is found that the neural network, which is an approximation of the actual shading function, is able to estimate the maximum possible output power of the solar PV arrays accurately. Further, the network is able to estimate the maximum output power for field data and gives rise to the possibility that the proposed approach can be used for making decision regarding the installation of solar PV arrays in the field.

The structure and analysis of nanotechnology co-author and citation networks

Research activities and collaborations in nanoscale science and engineering have major implications for advancing technological frontiers in many fields including medicine, electronics, energy, and communication. The National Nanotechnology Initiative (NNI) promotes efforts to cultivate effective research and collaborations among nano scientists and engineers to accelerate the advancement of nanotechnology and its commercialization. As of August 2008, there have been over 800 products considered to benefit from nanotechnology directly or indirectly. However, today’s accomplishments in nanotechnology cannot be transformed into commercial products without productive collaborations among experts from disparate research areas such as chemistry, physics, math, biology, engineering, manufacturing, environmental sciences, and social sciences. To study the patterns of collaboration, we build and analyze the collaboration network of scientists and engineers who conduct research in nanotechnology. We study the structure of information flow through citation network of papers authored by nano area scientists. We believe that the study of nano area co-author and paper citation networks improve our understanding of patterns and trends of the current research efforts in this field. We construct these networks based on the publication data collected for years ranging 1993 through 2008 from the scientific literature database “Web of Science”. We explore those networks to find out whether they follow power-law degree distributions and/or if they have a signature of hierarchy. We investigate the small-world characteristics and the existence of possible community structures in those networks. We estimate the statistical properties of the networks and interpret their significance with respect to the nano field.

Solar Photovoltaic Array's Shadow Evaluation Using Neural Network with On-Site Measurement

This paper proposes a method to accurately predict the maximum output power of the solar photovoltaic arrays under the shadow conditions by using neural network, a combined method using the multilayer perceptrons feed forward network and the backpropagation algorithm. Using the solar irradiation levels, the ambient temperature and the suns position angles as the input signals, and the maximum output power of the solar photovoltaic array as an output signal, the training data for the neural network is received by measurement on a particular time, when solar panel is shaded. After training, the neural network models accuracy and generalization are verified by the test data. This model, which is called the shading function, is able to predict the shadow effects on the solar PV arrays for long term with low computational efforts.

Evaluation of Production Facilities in a Closed-Loop Supply Chain: A Fuzzy TOPSIS Approach

It has become common for manufacturing facilities involved in production of new products to also carry out collection and re-processing of used products. While environmental consciousness has become an obligation to the facilities in the production of new products due to governmental regulations and public perspective on environmental issues, potentiality of the facilities to re-process used products directly affects the profitability of the facilities. Although many papers in the literature deal with performance evaluation of facilities, none of them address these two factors. To this end, a TOPSIS (Technique for Order Preference by Similarity to an Ideal Solution) approach, which evaluates production facilities in terms of both environmental-consciousness and potentiality, is proposed. Furthermore, since most of the criteria that fall under these two factors are intangible, triangular fuzzy numbers (TFNs) are employed to rate them in the evaluation process. A numerical example demonstrates the feasibility of the proposed method.

A cluster-based wavelet feature extraction method and its application

In this paper, a new cluster-based approach is proposed for extracting features from the coefficients of a two-dimensional discrete wavelet transform. The wavelet coefficients from the matrix of each frequency channel are segregated into non-overlapping clusters in an unsupervised mode using a set of application-specific representative images. In practical situations, this set of representative images can be the same as the ones kept aside for training a classifier. The proposed method divides the matrices of computed wavelet coefficients into disjoint clusters that are centered around the position of dominant coefficients. The features that can distinguish images of one class from those of other classes are obtained by computing energies of the clusters. The feature vectors so obtained are then presented as input patterns to an image classifier, such as a neural network. Experimental results based on the applications for texture classification and wood surface defect detection have shown that the proposed cluster-based wavelet feature extraction method is able to effectively extract important intrinsic information content from the test images, and increase the overall classification accuracy as compared with conventional feature extraction methods.