Showing papers by "College of Engineering, Pune published in 2017"

Machine learning for predictive maintenance of industrial machines using IoT sensor data

Abstract: The industrial Internet of Things (IIoT) is the use of Internet of Things (IoT) technologies in manufacturing which harnesses the machine data generated by various sensors and applies various analytics on it to gain useful information. The data captured by the machines is usually accompanied by a date time component which proves vital for predictive modelling. This paper explores the use of AutoRegressive Integrated Moving Average (ARIMA) forecasting on the time series data collected from various sensors from a Slitting Machine, to predict the possible failures and quality defects, thus improving the overall manufacturing process. The use of Machine Learning thus proves a vital component in IIoT having use cases in quality management and quality control, lowering the cost of maintenance and improving the overall manufacturing process.

Short term stock price prediction using deep learning

Abstract: Short — term price movements, contribute a considerable measure to the unpredictability of the securities exchanges. Accurately predicting the price fluctuations in stock market is a huge economical advantage. The aforementioned task is generally achieved by analyzing the company, this is called as fundamental analysis. Another method, which is undergoing a lot of research work recently, is to create a predictive algorithmic model using machine learning. To train machines to take trading decisions in such short — period of time, the latter method needs to be adopted. Deep Neural Networks, being the most exceptional innovation in Machine Learning, have been utilized to develop a short-term prediction model. This paper plans to forecast these short — term prices of stocks. 10 unique stocks recorded on New York Stock Exchange are considered for this review. The review essentially focuses on the prediction of these short — term prices leveraging the power of technical analysis. Technical Analysis guides the framework to understand the patterns from the historical prices fed into it, and attempts to probabilistically forecast the fleeting future prices of the stock under review. The paper discusses about two distinct sorts of Artificial Neural Networks, Feed Forward Neural Networks and Recurrent Neural Networks. The review uncovers that Feed Forwards Multilayer Perceptron perform superior to Long Short-Term Memory, at predicting the short — term prices of a stock.

On Hopf bifurcation in fractional dynamical systems

Abstract: Fractional order dynamical systems admit chaotic solutions and the chaos disappears when the fractional order is reduced below a threshold value [1]. Thus the order of the dynamical system acts as a chaos controlling parameter. Hence it is important to study the fractional order dynamical systems and chaos. Study of fractional order dynamical systems is still in its infancy and many aspects are yet to be explored. In pursuance to this in the present paper we prove the existence of fractional Hopf bifurcation in case of fractional version of a chaotic system introduced by Bhalekar and Daftardar-Gejji [2]. We numerically explore the (A, B, α) parameter space and identify the regions in which the system is chaotic. Further we find (global) threshold value of fractional order α below which the chaos in the system disappears regardless of values of system parameters A and B.

IoT based solid waste management system for smart city

Abstract: Today, waste management from its inception to its disposal is one of the important challenges for the municipal corporations in all over the world. Dust bins placed across cities set at open places are flooding because of increment in the waste each day and making unhygienic condition for the citizens, to maintain a strategic distance from such a circumstance we have proposed wireless solid waste management system for smart cities which allows municipal corporations to monitor status of dustbins remotely over web server and keep cities clean very efficiently by optimizing cost and time required for it. As soon as dustbin has reached its maximum level, waste management department gets alert via SMS via gsm module placed at dustbin so department can send waste collector vehicle to respective location to collect garbage. The objective of the project is to enhance practicality of IoT based solid waste collection and management system for smart city.

Conducting polyaniline nanocomposite-based paints for corrosion protection of steel

Abstract: The incorporation of inorganic fillers of different nature and size into conducting polyaniline (PANI)-based paint formulation extends the possibility of developing protective coatings with a self-healing capability and improved corrosion protection performance. The resulting PANI-based coatings are characterized as nanocomposite systems if the filler has nano-size dimensions. Nanofillers such as metal and metal oxide nanoparticles, clay, carbon nanotubes, graphene, and other inorganic pigments combined with PANI give rise to a variety of PANI nanocomposites with interesting properties and potential applications. The present review article concerns applications of PANI nanocomposites in steel anticorrosion technology. The advantages of PANI nanocomposite coatings over the parent polyaniline coating are highlighted. The synergistic effect of PANI and nanofiller leads to enhancement of the mechanical, physical, and chemical properties of coatings allowing the self-healing property of PANI to appear through either the anodic protection mechanism resulting in the oxide repairing at pinholes or the controlled inhibitor release mechanism by which the PANI-based nanocomposite coating liberates corrosion inhibitors (dopant ions) on demand upon the generation of a defect on the coating leading to the oxidation of the metal and hence to the reduction of PANI.

Smart flood disaster prediction system using IoT & neural networks

Abstract: Floods are the natural disasters that cause catastrophic destruction and devastation of natural life, agriculture, property and infrastructure every year. Flooding is influenced by various hydrological & meteorological factors. A number of researches have been done in flood disaster management and food prediction systems. However, it has now become significant to shift from individual monitoring and prediction frameworks to smart flood prediction systems which include stakeholders and the flood affecting people equally with help of recent technological advancements. Internet of Things (IoT) is a technology that is a combination of embedded system hardware and wireless communication network which further transfers sensed data to computing device for analysis in real-time. Researches in direction of flood prediction have shifted from mathematical models or hydrological models to algorithmic based approaches. Flood data is dynamic data and non-linear in nature. To predict floods, techniques such as artificial neural networks are used to devise prediction algorithms. Here an IoT based flood monitoring and artificial neural network (ANN) based flood prediction is designed with the aim of enhancing the scalability and reliability of flood management system. The main aim of this system is to monitor humidity, temperature, pressure, rainfall, river water level and to find their temporal correlative information for flood prediction analysis. The IoT approach is deployed for data collection from the sensors and communication over Wi-Fi and an ANN approach is used for analysis of data in flood prediction.

Nonlinear Control for Dual Objective Active Suspension Systems

Abstract: In this paper, an active suspension system employing a new nonlinear control law is proposed to address the problem of achieving the dual objective of providing ride comfort and trying to keep the suspension deflection within the constraint of rattle space. The control is a nonlinear function of the magnitude of the suspension deflection and an estimate of the effect of the road disturbance. The control scheme is analyzed and assessed for the large classes of road profiles through simulation and by experimentation on a laboratory setup. The performance of the proposed scheme is compared with a passive suspension system.

Diagnosis of liver diseases using machine learning

Abstract: Liver Diseases account for over 2.4% of Indian deaths per annum. [14] Liver disease is also difficult to diagnose in the early stages owing to subtle symptoms. Often the symptoms become apparent when it is too late. [1] This paper aims to improve diagnosis of liver diseases by exploring 2 methods of identification-patient parameters and genome expression. The paper also discusses the computational algorithms that can be used in the aforementioned methodology and lists demerits. It proposes methods to improve the efficiency of these algorithms.

An overview of DC-DC converter topologies and controls in DC microgrid

Abstract: DC Microgrid has a promising future due to its better compatibility with distributed renewable energy resources, higher efficiency and higher system reliability. This paper presents a comprehensive literature review of DC-DC Converters topologies used in DC Microgrids. The advantages and limitations of classical and recent converter topologies are discussed. The Hierarchical Control Strategies-Primary and Secondary controls have been reviewed. Primary control relies only on local measurements used for proper load sharing among converters. Secondary control is a coordinated control with some form of communication for additional functionalities. A brief note on protection and the key challenges faced in DC Microgrid operation have also been discussed. This paper gives a brief idea about the recent developments and overall operation of DC Microgrid.

A comparative study of cluster based outlier detection, distance based outlier detection and density based outlier detection techniques

Abstract: As there is an increasing demand of data, outlier detection is coming across as a popular field of research. Outlier is stated as an observation which is dissimilar from the other observations present in the data set. It is advantageous in the fields like medical industry, crime detection, fraudulent transaction, public safety etc. Outlier can be learnt in different fields like big data, time series data, high dimension data, biological data, uncertain data and many more. Most of the time 10% of the whole sample data set is incorrect, not accessible or missing sometimes. This paper studies and compares the popular outlier detection algorithms namely, Cluster based outlier detection, Distance based outlier detection and Density based outlier detection. Comparative study of these outlier detection techniques is performed to find out most efficient outlier detection method for calculation of the outlier.

Poisson noise reduction from X-ray images by region classification and response median filtering

Abstract: Medical imaging is perturbed with inherent noise such as speckle noise in ultrasound, Poisson noise in X-ray and Rician noise in MRI imaging. This paper focuses on X-ray image denoising problem. X-ray image quality could be improved by increasing dose value; however, this may result in cell death or similar kinds of issues. Therefore, image processing techniques are developed to minimise noise instead of increasing dose value for patient safety. In this paper, usage of modified Harris corner point detector to predict noisy pixels and responsive median filtering in spatial domain is proposed. Experimentation proved that the proposed work performs better than simple median filter and moving average (MA) filter. The results are very close to non-local means Poisson noise filter which is one of the current state-of-the-art methods. Benefits of the proposed work are simple noise prediction mechanism, good visual quality and less execution time.

A two-way FSI analysis of multiphase flow in hydrodynamic journal bearing with cavitation

Abstract: This work deals with a study of a three-dimensional CFD analysis and multi-phase flow phenomena for hydrodynamic journal bearing with integrated cavitation. The simulations are carried out considering the realistic bearing deformations by two-way fluid–structure interactions (FSI) along with cavitation using ANSYS®Workbench software. The design optimization module is used to generate the optimized solution of the attitude angle and eccentricity for the combination of operating speed and load. Bearings with and without cavitation are investigated. A drop in maximum pressure value is observed when cavitation is considered in the bearing. The rise in oil vapor distribution is noted with an increase in shaft speed which lowers the magnitude of the pressure build up in the bearing. The bearing deformations are analyzed numerically and found increasing with an increase in shaft speed. The experimental data obtained for pressure distribution showed good agreement with numerical data along with a considerable reduction in computation time.

Smart Poultry Farm Automation and Monitoring System

Abstract: A combination of wireless sensors and GPRS network can be used for controlling and monitoring environmental parameters in a poultry farm. Various environmental parameters like temperature, humidity, ammonia gas have a big role in operations of Poultry. Operator can get updates regarding the internal environmental situation of poultry farm by accessing the data using a web page. A combination of hardware and software is used which will initiate the action automatically to control the environmental parameters according to preset standards, if there are any changes in parameters which exceed the presets system will act automatically and help to take actions to control the environmental parameters. Sensors are used to control temperature, water level, smoke, gas and food dispensing. All these sensors are connected with the raspberry pi which can control and monitor all data. The data is transmitted using GPRS, and detailed record of poultry farm with status of environmental conditions is maintained at a webpage. System thus focuses on maintaining best possible environmental conditions with minimal human effort.

Data mining and visualization for prediction of multiple diseases in healthcare

Abstract: A vast amount of data is generated in the fields of healthcare and diagnostics, doctors have to make a direct contact with patients to determine the wounds, injuries and diseases by which the patient is affected. This paper highlights the application of classifying and predicting a specific disease by implementing the operations on medical data generated in the field of medical and healthcare. In this project an efficient multiclass Naive Bayes algorithm is used for prediction of a particular disease by training it on a set of data before implementation. Wrong clinical decisions taken by medical practitioners can cause any harm or result in serious loss of life of a patient which is hard to afford by any hospital. To acquire a precise and cost effective treatment, technology based Data Mining Systems can be constructed to make worthy decisions. The main aim of this project is to build a basic decision support system which can determine and extract previously unseen patterns, relations and concepts related with multiple disease from a historical database records of specified multiple diseases. The proposed system can solve difficult queries for detecting a particular disease and also can assist medical practitioners to make smart clinical decisions which traditional decision support systems were not able to. The decisions taken by medical practitioners with the help of technology can result in effective and low cost treatments. There is an insufficiency of technology and analysis system and methods to discover connections, concepts and patterns in the medical data. Data mining is an engineering study of extracting previously undiscovered patterns from a selected set of data. In this paper, data mining methods namely, Naive Bayes and J48 algorithms are compared for testing their accuracy and performance on the training medical datasets. The medical datasets will be visualized by different visualization techniques like 2D/3D graphs, pie charts and other methods. The algorithms mentioned above are compared and evaluated on basis of their accuracy and time consumption factors. The algorithm which gives out high accuracy in the comparative study is selected for implementation for developing the system.

Arduino UNO and GSM based wireless health monitoring system for patients

Abstract: In India, near about 20% of the total population loses their lives due to interrupted health monitoring system i.e. in most of the hospitals, doctor visits patients either in morning shift or in evening shift or in both shifts. What happens if patient's health becomes critical in between that interval or when a doctor is not available with a patient. The answer is; a patient may lose her\his life. So to avoid this critical situation; we are proposing a smart embedded system device which monitors patients health continuously. This system monitors patients heart rate, body temperature and saline liquid level (if any). If any of the above parameters goes beyond the threshold value, this smart device informs doctors or care taker and ask for corrective actions to save patients life.

Graphene oxide-modified polyaniline pigment for epoxy based anti-corrosion coatings

Abstract: In the present work, a set of polyaniline–graphene oxide (PANI–GO) nanocomposites which exhibit superior properties in terms of shelf life, processability and conductivity due to the synergistic effect of GO and PANI, have been synthesized by varying the concentration of highly non-conducting GO with respect to aniline. The obtained materials were characterized by UV–Vis, FTIR, XRD, Raman, TGA as well as FESEM, TEM analysis. The results reveal that nanocomposites show better dispersibility, crystallinity, thermal stability, and conductivity. Further, the synthesized composites have been tested for their anti-corrosion properties. The potentiodynamic results reveal that PANI nanocomposites with 1% GO exhibited long-term anti-corrosion behavior with a corrosion rate of 6.5 × 10−5 mm year−1, which is much lower than its individual components and commercial-grade red oxide. Also, it possesses highest impedance modulus ~33 kΩ cm2 and real impedance ~32 kΩ cm2, maximum coating resistance ~14.81 × 103 Ω cm2 and minimum coating capacitance after 96 h of immersion in 3.5% mass NaCl than those exhibited by all other coated samples. Higher concentration of GO could not retard the corrosion rate confirming that hydrophilicity of GO play an important role in the redox mechanism of PANI.

Detection of fire using image processing techniques with LUV color space

Abstract: Vision based fire detection system have recently gained popularity as compared to traditional fire detection system based on sensors. The popularity and need of video surveillance at residential, Industrial, public and business locations have supported the widespread use of vision based fire detection system. The colour of fire is the basic technique for identification of fire in an image. However, the colour of fire varies from red, orange, yellow to white. Also, there are non-fire objects with fire-like colour. In order to improve the accuracy of fire detection system, colour detection is combined with various other techniques. Edge detection, motion detection, area covered by flames, existence of smoke, growth of fire and background segmentation are some techniques which are combined by various researchers and used to correctly classify the fire images and fire-like non fire images in a video. There are also various thresholds that are used to differentiate fire in any frame. These thresholds need to be adjusted based on the type of area and its brightness level. Also, the difference in the subsequent frames and area covered by the flames supports the existence of fire if it is greater than the threshold. This paper presents the comparative analysis of five recent vision based fire detection system. These fire detection systems are based on flame colour detection combined with other features such as motion and area of frame. The fire detection system based on LUV colour space and hybrid transforms is proposed.

Optimizing the size of a CFRP patch to repair a crack in a thin sheet

Abstract: One of the several techniques to repair cracks in structural sheets consist in bonding polymer composite patches. The effectiveness of the repair for restoring the quasistatic strength of the structure depends largely on the adhesively bonded interface. The interface fails due to interfacial separation caused by the high peeling and shearing stresses. The geometrical dimensions, that is, patch length and width, have significant effect on the interface separation and they need to be optimized. The failure strength of the patch was determined by a numerical analysis using the cohesive zone model. Twenty-five numerical analyses were carried out as per the L-25 Taguchi orthogonal array followed by ANOVA which indicated the greater contribution of the patch width toward the failure of the patch. The failure stresses thus obtained were used to generate a response surface in ANSYS Design Explorer Module. A design criterion in terms of the percentage increase of the failure stress over the yield stress of...

Platforms for big data analytics: Trend towards hybrid era

Abstract: The primary objective of this paper is to present detailed analysis of various platforms suitable for Big Data processing. In this paper, various software frameworks available for Big Data analytics are surveyed and in-detail assessment of their strengths and weaknesses is discussed. In addition to this, widely used data mining algorithm are discussed for their adaptation for Big Data analysis w.r.t their suitability for handling real-world application problems. Future trends of Big Data processing and analytics can be predicted with effective implementation of these well established and widely used data mining algorithms by considering the strengths of software frameworks and platforms available. Hybrid approaches (integration of two or more platforms) may be more appropriate for a specific data mining algorithm and can be highly adaptable as well as perform real-time processing.

Sound based human emotion recognition using MFCC & multiple SVM

Abstract: Emotion recognition using human speech is one of the latest challenges in speech processing and Human Machine Interaction (HMI) for the purpose of addressing varied operational needs for the real world applications Besides human facial expressions, speech has been proven to be one of the most valuable modalities for automatic recognition of human emotions Speech is a spontaneous medium of perceiving emotions which provides in-depth Here in this paper, we have used MFCC for extraction of features and Multiple Support Vector Machine (SVM) as a classifier We have performed extensive experiment on happy, anger, sad, disgust, surprise and neutral emotion sound database Performance analysis of multiple SVM revealed that non-linear kernel SVM achieved greater accuracy than linear SVM