Kongu Engineering College

About: Kongu Engineering College is a based out in . It is known for research contribution in the topics: Computer science & Cluster analysis. The organization has 2001 authors who have published 1978 publications receiving 16923 citations.

Effect of shot blasting on droplet contact angle of carbon aided phase change nanocomposites

Abstract: This experimental work deals with the measurement of the contact angle of the nanocomposites made of phase change material (PCM) OM 08 and multi-wall carbon nanotubes (MWCNTs) with different volume...

Fuzzy weighted geometric mean approach-based FMEA to prioritise lean failure modes in manufacturing industries

Abstract: This paper presents a methodology called lean failure mode and effects analysis (LFMEA) using fuzzy weighted geometric mean for prioritising lean failure modes. Fuzzy lean priority number (FLPN) is calculated to prioritise lean failure modes using factors like occurrence, severity and detection. Fuzzy linear programming models and alpha-cut techniques are used for ranking failure modes. For defuzzification of FLPNs, alpha-level set-based centroid defuzzification method is used. Based on the FLPN values, lean tools are selected to eliminate the lean failures. The future state implemented with selected lean tools has shown encouraging results from case studies. Lead time, cycle time and non-value added time are reduced by 11.3%, 14% and 6.1% for a casting industry and 47.3%, 31.3% and 67.2% respectively for an auto components manufacturing industry. The unnecessary movements, manpower and inventory are reduced to greater levels. This selection methodology confirms positive effects on productivity improvement process.

Robust watermarking of AES encrypted images for DRM systems

Abstract: Digital image capturing, processing and distribution has showed a remarkable growth over recent years. This media content is sometimes distributed in encrypted format and watermarking of these media items for proof of ownership, media authentication needs to be carried out in encrypted domain to improve image security. Therefore it is sometimes necessary to embed watermark in encrypted media items for ownership declaration or copyright management purposes. DRM system is one such example where there is a challenge to watermark these encrypted data as the encryption would have randomized the incoming data. In this paper, a block cipher called AES-128 bit key encryption algorithm and DCT combined with DWT based watermarking algorithm to watermark the encrypted image were proposed which increases robustness of the watermark. These method embeds the binary watermark in encrypted image and decryption is done after extraction of watermark.

Classification of Multi-Lead ECG Signals to Predict Myocardial Infarction Using CNN

Abstract: Myocardial infarction (MI) which causes the damage to heart muscles and it lead to the critical stage of death. However the efficacious diagnosis of myocardial infarction (heart attacks) is needed for the healthy life of human. Electrocardiogram (ECG) is utilized to diagnose MI. A genuine time signal provides the electrical activities that are the subsidiary information about the functioning of heart. The expeditious and precise diagnose of MI need to be done with artificial intelligence based on computer aided techniques. In this paper, a multi layer deep convolutional neural network structure is proposed along with the data augmentation technique for the prediction of myocardial infarction. Furthermore, the implementation is done by using GPU version. When it comes to training and developing the new models and algorithms, the performance is determined by means of training and testing speed. Since GPU processor have been used to increase the computations speed and it also scales better then CPU.

An Enhanced Handwritten Digit Recognition Using Convolutional Neural Network

Abstract: Handwritten digit recognition have great impact in the applications of deep learning. Convolutional Neural Network in the deep learning has become one of the major methods and one of the important factors in the various success in recent times and deep learning is used majorly in the area of object recognition. In the paper work, the speech output feature is integrated along with the text output. Convolutional Neural Network model is applied in the image classification. The dataset used to train and test is the MNIST dataset. There are various applications of handwritten digit recognition in the real time. It is applied in detection of vehicle number, reading of bank cheques, the arrangement of letters in the post office.