Papers by Keyword: Linear Regression

Abstract: This research integrates experimental testing and Machine Learning (ML) techniques to predict the weld quality of Tungsten Inert Gas (TIG) and Shielded Metal Arc Welding (SMAW). A balanced dataset comprising weld parameters and mechanical test results including tensile strength, impact energy, and bend test outcomes was compiled for mild steel and stainless steel specimens with thickness ranging from 6mm to 10mm. Experimental results revealed that TIG welding produced superior tensile strength (up to 572 MPa) and impact energy (up to 58J) compared to SMAW. A Random Forest classifier achieved 100% accuracy in classifying weld quality as Good or Defect, while linear Regression produced tensile strength with an R2 of 0.68, Mean Absolute Error (MAE) of 17.5 MPa, and Root Mean Squared Error (RMSE) of 20.27 MPa. These results confirm the viability of ML techniques as non-destructive tools for weld quality prediction and mechanical property estimation. The framework developed in this research contributes to intelligent welding process control and supports the transition toward efficient, data driven manufacturing.

Abstract: In this study, we used a linear regression machine learning model to predict the stress-strain curve of AZ91/graphene composites. The proposed model successfully made predictions with an accuracy of approximately 0.99 (99%) and a small error. The mechanical properties obtained from the curves, such as the yield and ultimate tensile strength, were in excellent agreement with the actual and predicted values. This linear regression model is also well-suited for predicting the stress-strain curve of composites.

Abstract: Dijkstra algorithms are typically used to find the shortest path from a source node to a destination node. It is widely used in various applications due to its reliability and less complexity. This paper presents the extended Dijkstra Algorithm with lower latency and consumes less computing memory intended for implementation in many AGVs networks for effective decentralized task distribution path planning. This paper proposed linear regression normalization across the node network in Dijkstra architecture to reduce computing time and memory consumption. The issue addressed through this optimization focused on reducing the possibilities of collision between AGVs and deadlock. The extended Dijkstra algorithm significantly reduces computing time compared to the traditional Dijkstra algorithm. In addition, the proposed solutions suggest better AGV routing for collision avoidance and deadlock prevention possibilities.

Abstract: Stock analysis and forecasting is a very challenging study due to the unpredictable and volatile database environment. However, their patterns are often unique as they are influenced by many uncertainties, such as financial results of companies (Earnings per share), risk transactions, market sentiment, government policies, and conditions such as epidemics. Even though they are challenging our goal is to predict the accurate values within a shorter span of a dataset. In this paper we have compared and analyzed the best ML model that predicts the exact closing amount of the next few days, using three to four months of nifty50 Indian stock from Yahoo Finance. Five regression models are involved in this analysis, Linear Regression (LR), Decision Tree (DT), Support Vector Regression (SVR), SARIMAX (Integrated Seasonal Integrated Season with EXogenous features), Gated Recurrent Unit (GRU – deep learning). The performance metrics like RMSE (Root Mean Squared Error), MAE (Mean Absolute Error) and MAPE (Mean Absolute Percentage Error) are used. On the basis of our comparison, we would like to conclude that GRU provides a low error value in all three performance metrics and also gives accurate predictions compared to the other five regression models used.

Abstract: Coronavirus becomes cerebral pain every day throughout the world. Many cases of coronavirus continue to grow, directly irritating human daily exercises and devastating the economy of nations. The Indian Government announced a one day Janta curfew on March 22, 2020. After three days on March 25 2020, 19 days of lockdown were declared in the country for mitigation of the COVID-19 pandemic. Four lockdowns and six unlock periods were implemented to control the pandemic, but lockdown is the major obstacle to the economy. In unlocking period government open the economic activity stepwise to boost the economy. Coronavirus infection is under control during a lockdown time, but the infection becomes pandemic unlock 1.0, 2.0 and 3.0 period. In Unlock 4.0 and unlock 5.0 coronavirus cases growth goes down but in unlock period 6.0, a sudden spike in confirmed cases. It is due to the festival session and relaxation provided by the Government in the unlock 6.0. The research aimed to forecast the trend towards the COVID-19 pandemic in India with data from June 01, 2020, by applying the ARIMA and Prophet model. Based on several presumptions, the findings of the analysis have shown that, after the unlock-up period is completed, it has been predicted that India's pandemic is expected to decrease by approximately about December 2020 and that it will crest around within the initial weeks of March 2021.

Abstract: One of the owner’s common problem before executing construction projects is the complexity in estimating the project cost in an early stage. Inaccurate cost estimation will force the owner to make further arrangement to the project budget. This study aims to develop an initial cost estimation model for superstructures of Precast I-girder Bridge. Cost estimation model was developed based on thirteen data of detail engineering design of I-girder bridge in Daerah Istimewa Yogyakarta (DIY). Factors influencing the cost of the superstructures of the I-girder bridge were identified. Bridge span and width, the size of the sidewalk, and railing’s type are considered as variables affecting the cost of superstructures. These variables are then arranged into two different analysis Multiple Linear Regression (MLR) analysis and Artificial Neural Network (ANN), in order to obtain the best estimation model. The results of the analysis showed that bridge span and width were the significant factors influencing cost. The correlation value of bridge span is 89.0%, bridge width is 74.2%, the size of the sidewalk is 66.1%, and railing’s type is 46.1% as identified factors that affect the cost of the superstructure. A comparative model of two approaches shows that the ANN has better accuracy than that of MLR, although the difference was not significant.

Abstract: Travel time value of public transport passengers is one of important variables in decision making about transport policy. Giving subsidy for public transport and allocate it to the right passenger will result in more benefits for the passenger of public transport. And as an effect it will increase the number of passengers then increase the public transport usage. As a result, it will reduce the number of private car usage which will reduce the air pollution and oil consumption and finally support sustainability transport. In this study, Travel time value based on questionnaire data, which is designed with stated preference with route choice approach. The multiple linear regression model is used to analyzed factors that influence public transport values of travel time, and traveler's income and trip purpose are introduced as categories. The study results indicate, in general, travel time value of regional trips is higher than urban travel. In Urban trip, travel time values for school is higher than values for work and other travels, while, for intercity travel, time value for work are higher than school and other travels. It can be comprehended since in this study the passenger of the urban trip who the travel purpose is work has low income. In common, when the personal income increase the travel time value also increase. But within this study, the public transport passengers who have no income consist of students, and their school regulations give hard punishment when their students come late. Based on that, the passenger with no income their travel time value is higher than the passengers have the lowest income.

Abstract: In order to acquire high machining quality and minimum machining time, cutting forces are usually modeled to understand the milling process, simulate or predict cutting forces, and optimize the machining parameters. In this paper, side milling tests were conducted on superalloy Inconel 718 with a solid carbide end mill, and the cutting forces vs. cutting time were measured. The average cutting forces were extracted from the measured instantaneous cutting forces under different feed rates of experiments, and the components of the shear forces and edge forces were determined by using the linear regression of the experimental data. The cutting force coefficients, including shear force coefficients and edge force coefficients, were identified. In addition, the algorithms of the mathematical model were implemented in Matlab. The predicted cutting forces were in good agreement with the experimentally measured forces, and the validation of the cutting force model was demonstrated.

Abstract: The intermittent drawbacks due to fluctuating temperature factor inside photovoltaic (PV) cells have clearly affected the overall energy performance especially in stochastic weather conditions. Temperature element in the tropical regions is a crucial factor to be determined based on Standard Testing Condition (STC) and Nominal Operating Cell Temperature (NOCT) correlations. Based on the crucial implication of heat dissipation, this study shares some insights of five level heat contour covering the surrounding temperature, PV surface temperature, PV bottom temperature and 2-level of two feet height located under PV array. The field data in real-time approach has been brought up in line to support the energy balance modelling for PV applications with localized heat contour analysis using statistical evaluations. The regression analysis of the 3471 data sampling for the period of 5 days (7AM till 7 PM) produces very good results with correlation coefficient, R² = 0.97.

Abstract: Described the problem of recovering from discrete data in general smooth function. The algorithms of data smoothing, providing a smooth pair of generalized, built according to the linearly independent functions, polynomials defined on the mating intervals analysis. The possibility of using for this purpose the methods of constrained optimization. As the objective function is used the quadratic loss function. The smoothness of the conjugation of polynomials is provided with equality constraints. It was established that in the absence of constraints of inequality type solution of the recovery can be obtained in explicit form. The validity of the developed algorithms on examples of the treatment of special test signals. The results indicate the possibility of their use for the analysis multiextremal dependencies, which include the intensity of the acoustic signals functioning technical systems.