Search

The paper investigated the efficacy of entropy-based ANN-PSO model combining Artificial Neural Networks (ANN) and Particle Swarm Optimization (PSO) for estimation and optimization of quality characteristics associated with pulsed Nd:YAG laser cutting of aluminium alloy. In the ANN-PSO model, ANN trained using backpropagation with the Bayesian regularization algorithm is employed for estimation and computation of objective function value during optimization with PSO. The entropy method is used to compute the real weight of different output quality characteristics during formulation of the combined objective function. An experiment has been conducted based on full factorial experimental design, where cutting speed, pulse energy, and pulse width are considered as controllable input parameters while kerf width, kerf deviation, surface roughness, and material removal rate are measured as output parameters. Further, the experimental dataset has been used in the ANN-PSO model for prediction and optimization. The prediction accuracy of the ANN module is indicated by a low mean absolute error of 1.74%. Experimental validation of optimized output also results in less than 2% error only. ANOVA study suggests cutting speed as the most influencing factor.

Some people are too ignorant of their surroundings. Most of them, usually, throw trash illegally in the streets or abandoned areas. In this paper, to get rid of such a problem, we have proposed a real-time monitoring mechanism to find the illegal trash throwing person for an intelligent surveillance system. The proposed mechanism of finding people and trash from the video frame can be done using subsequently applying the following several steps. At first, we have applied the shadow removal technique and then applied the background subtraction procedure. To reduce the noise levels, we have implemented labeling & filtering operations. For extracting feature vectors, a histogram of oriented gradients (HOG) is used. Finally, by applying a support vector machine classifier (SVM), we have identified both humans and trash from the video frames. The key contribution of this proposed framework is to gain better performance by applying the shadow removal technique before the background subtraction procedure. Besides, the temporary bounding box concept has been taken into consideration during the trash detection phase.

In the present work, artificial neural networks (ANN) have been used to model the complex relationship between input-output parameters of metal inert gas (MIG) welding processes. Four ANN training algorithms such as back propagation neural network (BPNN) with gradient descent momentum (GDM), BPNN with Levenberg Marquardt (LM) algorithm, BPNN with Bayesian regularization (BR), and radial basis function networks (RBFN) method have been used for prediction modelling. An experimentation based on full factorial experimental design has been conducted on MIG welding of austenitic stainless steel of grade-304 where welding current, welding speed, and voltage have been considered as input parameters, and tensile strength has been considered as measurable output parameter. The dataset so constituted is used for ANN modelling. Altogether, 40 different ANN architectures have been trained and tested using the above-mentioned algorithms, and 3-11-1 ANN architecture trained using BPNN with BR has been considered to show best prediction capability with mean % absolute error of 0.354%.

Present paper attempted to model complex relationship between CO2 laser–MIG hybrid welding parameters and it has been completed using different algorithms of artificial neural networks (ANN). Input parameters for the study include laser power, welding speeds and wires feed rate and tensile strength of the joint is considered as output. A full factorial experimental dataset is used for the purpose. Variants of back propagation neural networks (BPNN) and Radial Basis Function Networks have been used as training algorithm. Altogether 65 different ANN architecture have been trained and tested using 6 different training algorithms to find out ANN with best prediction capability. 3-11-1 ANN architecture trained using BPNN with Bayesian regularization shows best prediction capability (mean square error 3.24E − 04) and considered as Best ANN. That ANN will be useful for determining required value of welding process parameters to yield a specific welding strength and suitable for online process monitoring and control. Finally, a sensitivity analysis has been conducted and it is found that, maximum welding strength can be obtained with low wire feed rate (4 m/min), low welding speed (2 m/min) and high laser power (3 kW).

Integrated ANN-GA and ANN-SA methodology are two integrated soft computing based models that can predict and optimise input-output parameters of any manufacturing process with required optimisation accuracy eliminating the need of any closed form objective functions. In the present chapter, a case study on modelling and optimisation of CO2 LASOX cutting of mild steel plates have been carried out using the integrated ANN-GA and ANN-SA methodology to investigate efficacy of those integrated methods. In the case study cutting speed, gas pressure, laser power and standoff distance have been considered as process variables for modelling and optimisation of HAZ width, kerf width and surface roughness. 36 different ANNs have been trained and tested for ANN modelling. Finally, 4-8-3 network during training and testing through BPNN with BR results best prediction performance with MSE of 8.63E−04. Prediction capability of the best ANN (4-8-3) is found superior compared to the second order regression models developed for the purpose and is used in integrated models for optimisations. During optimisation, ANN-SA is found to show best optimisation performance with maximum absolute % error of 5.18% during experimental validation. Optimum cut quality is produced by low gas pressure and high cutting speed, laser power and stand off distance.

Nowadays, gas-assisted laser cutting of mild and carbon steels has been widely used in manufacturing industries for its accuracy and efficiency. But laser power requirement increases with thickness of the plate and increases process cost substantially. Laser assisted oxygen cutting (LASOX) is an effective method for cutting thick section cutting of mild steel with low power laser. The present chapter has explained the development of the LASOX process with its basic working principle. Researches carried out so far for modelling and optimisations of the LASOX process parameters using conventional statistical methods also have been discussed. But it has been observed that, limitations of statistical methods for modelling of complex non linear relationship between variables of laser material processing can be overcome by incorporating soft computing techniques. Those techniques are widely used in different laser material Processing. Some significant recent works have been discussed in the present chapter. But it has been yet to be incorporated in LASOX processes. Chapter ends with a notion to develop constructive integrated soft computing models for modelling and optimisation of LASOX processes.

The chapter has explained working of two integrated soft computing based models comprising of Artificial Neural Networks (ANN), Genetic Algorithms (GA) and Simulated Annealing (SA) namely, ANN-GA and ANN-SA for modelling and optimisation of Laser Assisted Oxygen Cutting (LASOX) process. Three different ANN training algorithms such as back propagation neural networks (BPNN) with Levenberg Marquardt (LM) algorithm, BPNN with Bayesian Regularisation (BR) and Radial Basis Function Networks (RBFN) have been employed for training and subsequent prediction. The ANN producing best prediction performance is used for prediction of objective function value. That process is thereby eliminating the need for closed form objective functions to produce required optimisation accuracy. The Chapter detailed the working of ANN, GA, SA and integrated models in step by step for easy understanding of the readers. The process has been explained in a generalised manner so that it can be applied for modelling and optimisation of any manufacturing process.

The paper presents an integrated model of artificial neural networks (ANNs) and non-dominated sorting genetic algorithm (NSGAII) for prediction and optimization of quality characteristics during pulsed Nd:YAG laser cutting of aluminium alloy. A full factorial experiment has been conducted where cutting speed, pulse energy and pulse width are considered as controllable input parameters with surface roughness and material removal rate as output to generate the dataset for the model. In ANN–NSGAII model, back propagation ANN trained with Bayesian regularization algorithm is used for prediction and computation of fitness value during NSGAII optimization. NSGAII generates complete set of optimal solution with pareto-optimal front for outputs. Prediction accuracy of ANN module is indicated by around 1.5 % low mean absolute % error. Experimental validation of optimized output results less than 1 % error only. Characterization of the process parameters in pareto-optimal region has been explained in detail. Significance of controllable parameters of laser on outputs is also discussed.

Three integrated ANN-GA, ANN-SA and ANN-Quasi Newton methodology has been developed and implemented according to the following way to determine optimised input parameter setting for maximum welding strength during laser-MIG hybrid welding of aluminium alloy plates. Finally, significance of optimised parameters has been determined by ANOVA. Variation of welding strength with individual process parameters have been tested through main effect plots and interaction plots. Three soft computing based integrated models such as, ANN-GA, ANN-SA and ANN-Quasi Newton have been developed.Those models predicted and optimised welding strength during hybrid CO2 laser-MIG welding process.Best ANN architecture (3-11-1 network) predicts welding strength with mean absolute percentage errors less than 2%.ANN-GA shows best optimisation performance with only 0.09% experimental validation error.Welding speed shows maximum influence on welding strength and an increase in welding speed decreases welding strength. In this paper, artificial neural networks (ANNs), genetic algorithm (GA), simulated annealing (SA) and Quasi Newton line search techniques have been combined to develop three integrated soft computing based models such as ANN-GA, ANN-SA and ANN-Quasi Newton for prediction modelling and optimisation of welding strength for hybrid CO2 laser-MIG welded joints of aluminium alloy. Experimental dataset employed for the purpose has been generated through full factorial experimental design. Laser power, welding speeds and wires feed rate are considered as controllable input parameters. These soft computing models employ a trained ANN for calculation of objective function value and thereby eliminate the need of closed form objective function. Among 11 tested networks, the ANN with best prediction performance produces maximum percentage error of only 3.21%. During optimisation ANN-GA is found to show best performance with absolute percentage error of only 0.09% during experimental validation. Low value of percentage error indicates efficacy of models. Welding speed has been found as most influencing factor for welding strength.

Laser assisted oxygen cutting (LASOX) process is an efficient method for cutting thick mild steel plates compared to conventional laser cutting process. However, scanty information is available as to modeling of the process. The paper presents an optimized SA-ANN model of artificial neural network (ANN) and simulated annealing (SA) to predict and optimize cutting quality of LASOX cutting process of mild steel plates. Optimization of SA-ANN parameters is carried out first where the ANN architecture and initial temperature for SA are optimized. The optimized ANN architecture is further trained using single hidden layer back propagation neural network (BPNN) with Bayesian regularization (BR). The trained ANN is then used to evaluate the objective function during optimization with SA. Experimental dataset employed for the purpose consists of input cutting parameters comprising laser power, cutting speed, gas pressure and stand-off distance while the resulting cutting quality is represented by heat affected zone (HAZ) width, kerf width and surface roughness. Results indicate that the SA-ANN model can predict the optimized output with reasonably good accuracy (around 3%). The proposed approach can be extended for prediction and optimization of operational parameters with reasonable accuracy for any experimental dataset.

The paper presents an artificial neural network-optimization hybrid model to predict and optimize penetration depth of CO2 LASER-MIG hybrid welding used for 5005 Al–Mg alloy. The input welding parameters are power, focal distance from the work piece surface, torch angle, and the distance between the laser and the welding torch. The model combines single hidden layer back propagation artificial neural networks (ANN) with Bayesian regularization for prediction and quasi-Newton search algorithm for optimization. In this method, training and prediction performance of different ANN architectures are initially tested, and the architecture with the best performance is further used for optimization. Finally, the best ANN architecture is found to show much better prediction capability compared to a regression model developed from the experimental data.

During precision cutting of thin sheet metals with pulsed Nd:YAG laser beam, best cut quality can obtained by uniform kerf with minimum kerf width. An integrated GA-ANN model has been developed and employed in the present work for prediction and optimisation of cut quality where trained ANN is used for computation of fitness value. A full factorial experiment on laser cutting of aluminium alloy using pulsed Nd:YAG laser has been accomplished for generation of dataset for the model. It is observed that, the ANN module can predict kerf width and kerf deviation with mean absolute % error of 0.42% and 1.05%. Optimised kerf width and kerf deviation determined by GA optimisation results in absolute % error of 1.87% and 2.00% only during comparison with results of validation experiment. By way of using this model a considerable improvement in kerf quality has been achieved with optimised parameter setting.