Optimal Tuning Parameters Research Articles

ON THE PRINCIPLES OF CONTROL OF THE PROCESS PART OF THE AIRBORNE CLASSIFIER DURING OPERATIONS OF SPECIAL-PURPOSE ABRASIVE MATERIALS

The article considers the methodical methods of practical use of a vibrodynamic classifier for sieving various abrasive materials, sets and solves the task of providing optimal tuning parameters of the electromechanical part of the vibration classifier to obtain the highest grade of abrasive grain with a high content of the main fraction and a given shape of abrasive particles. The distribution of particles of the material separated by size along the perimeter of the vibration rod was studied, depending on its parameters. It has been shown that the yield of products depends on the ratio of fractions in the starting material.

Data-adaptive M-estimators for robust regression via bi-level optimization

M-estimators are widely used in robust regression to handle heavy-tailed data corrupted by outliers. Although they have been applied to a plethora of real scenarios, it remains a challenge to practitioners how to set the tuning parameters. Often, it is set by manual tuning or according to the asymptotic efficiency rule, being sub-optimal for a real dataset with finite size. In this paper, we explore a data-driven paradigm where the optimal tuning parameters are determined by the dataset itself. Specifically, we treat the tuning parameters as hyper-parameters in robust regression, formulate the tuning problem via a novel bi-level optimization framework, and solve the regression model parameters and the tuning parameters in a joint manner. To solve this problem efficiently, especially when using neural network as the regression model, we further employ an online approximation strategy to iteratively optimize the model parameters and the tuning parameters with a proven sub-linear convergence rate. Moreover, our proposed framework is generic for any parametric regression model and M-estimator with differentiable loss function. We instantiate this framework with two popular M-estimators (Huber’s and Tukey’s) and derive the corresponding data-adaptive M-estimators. In the experiment, we present positive simulation results compared with various salient benchmarks.

Design of an Optimal Fractional Complex Order PID Controller for Buck Converter

Dynamic and robust controllers are the inherent requirement of power electronic converters, which are subjected to dynamic variations and nonlinearities. The effectiveness of fractional order controllers in non-linear system control has been well-established by studies in the past few decades. Various forms of fractional order controllers have been used in power-electronic control. Recent research indicates that complex order controllers, extensions of fractional controllers, are more robust against uncertainties and non-linearities than their integer and fractional order counterparts. Though complex order controllers have been employed in various nonlinear plants, they have not been extensively tested on power electronic applications. Also, the design and tuning of the controller is difficult. This paper investigates the effectiveness of a complex order PID controller on a typical power electronic DC-DC buck converter for the first time. Two types of complex order controllers of the form PI^{a+ib}D^c and PI^{a+ib}D^{c+id} were designed for a power electronic buck converter. The complex order controllers were implemented in Simulink and the optimal tuning of the complex order controller parameters for various performance indices was performed using different optimization algorithms. The Cohort Intelligence algorithm was found to give the most optimal results. Both the complex controllers showed more robustness towards uncertainties than the linear and fractional PID controllers. The PI^{a+ib}D^c controller gave the smoothest and fastest response under non-linearities. The dynamic performance of the complex order controller is the best and can be expected to be useful for more power electronic applications.

Transfer Learning Facilitates the Prediction of Polymer-Surface Adhesion Strength.

Machine learning (ML) accelerates the exploration of material properties and their links to the structure of the underlying molecules. In previous work [Shi et al. ACS Applied Materials & Interfaces 2022, 14, 37161-37169.], ML models were applied to predict the adhesive free energy of polymer-surface interactions with high accuracy from the knowledge of the sequence data, demonstrating successes in inverse-design of polymer sequence for known surface compositions. While the method was shown to be successful in designing polymers for a known surface, extensive data sets were needed for each specific surface in order to train the surrogate models. Ideally, one should be able to infer information about similar surfaces without having to regenerate a full complement of adhesion data for each new case. In the current work, we demonstrate a transfer learning (TL) technique using a deep neural network to improve the accuracy of ML models trained on small data sets by pretraining on a larger database from a related system and fine-tuning the weights of all layers with a small amount of additional data. The shared knowledge from the pretrained model facilitates the prediction accuracy significantly on small data sets. We also explore the limits of database size on accuracy and the optimal tuning of network architecture and parameters for our learning tasks. While applied to a relatively simple coarse-grained (CG) polymer model, the general lessons of this study apply to detailed modeling studies and the broader problems of inverse materials design.

Performance evaluation of isolated three-phase voltage source inverter with LC filter adopting different MPC methods under various types of load

As the variety of loads in microgrids continues to grow rapidly, the drawbacks of using conventional voltage and current double closed-loop PI control are becoming increasingly intolerable. These disadvantages include poor static tracking of the output and the need for complicated tuning of optimal PI control parameters. Model predictive control (MPC) is widely studied in recent years and gradually applied to the inverter as a modern control strategy. Currently, the research on MPC mainly focuses on reducing the computational burden of algorithms and improving computing skills. In this paper, the control theory of continuous-control-set model predictive control (CCS-MPC), optimal switching vector model predictive control (OSV-MPC), quasi-sliding mode model predictive control (QS-MPC) and optimal switching sequence model predictive control (OSS-MPC) are elaborated. Also, the disadvantages of adopting conventional voltage and current double closed-loop PI control on microgrids are clarified. Moreover, aiming at evaluating the influence of connecting and/or removing the common loads, such as electric vehicles to the microgrid on the output performance of microgrids in practical engineering applications, single three-phase voltage source inverter with LC filter system with two-phase pure resistance load, three-phase diode rectifier bridge nonlinear load, constant power load and constant current source load applying four different MPC methods, are simulated. The simulation results show that under different types of load, system with nonlinear load applying OSV-MPC method has the biggest value of total harmonic distortion (THD) of the output voltage and the biggest steady state root mean square (RMS) tracking error of the output voltage, which are 1.33% and 2.1 V, respectively. The performances of the output voltage of the system adopting MPC are generally improved compared to the system using conventional voltage and current double close-loop PI control. Regarding the output voltage performance of a system under load step change, a system using the OSS-MPC method with a nonlinear load exhibits the smallest overshoot of the output voltage and the smallest root mean square value of the tracking error (RMSE) between the output voltage and the reference value within 20ms after the load step change. Specifically, the overshoot and RMSE are 2.804 V and 0.810 V, respectively. The simulation results provide valuable guidance for selecting the appropriate model predictive control strategy for microgrid integrating complex loads, such as electric vehicles in practical applications.

Optimal tuning of PID controller for speed control of DC motor using equilibrium optimizer

The efficient and smooth operation of direct current (DC) motors is of high significance in various industrial applications. This study proposed a recent optimization technique, named equilibrium optimizer (EO) for optimal tuning of the parameters of proportional integral derivative (PID) controller to provide accurate and robust control in DC motors. Mathematical modelling of the motor was carried out in Simulink environment using basic electric machine theory. Conventional Ziegler-Nichols (ZN) technique is initially designed to tune the PID controller parameters, an optimization model based on the minimization of integrated absolute error and named EO-PID is formulated and subsequently implemented to optimize the tuned parameters obtained using the ZN-PID controller. Simulation results showed that there is a significant reduction in terms of rise time, settling time, overshoot and mean square error using the proposed EO-PID tuning model as compared to the ZN-PID technique. The result also showed that the proposed EO-PID model gives a better dynamic performance as it eliminates the oscillatory response common with most conventional control techniques. The results of this study have established EO as an effective optimization technique for optimal tuning of PID controller parameters, and can therefore be used by machine operators for various industrial applications.

Logistic Regression with Hyper Parameter Tuning Optimization for Heart Failure Prediction

Heart failure is a major public health concern that causes a substantial number of deaths worldwide. Risk factor analysis is required to diagnose and treat patients with heart failure. The logistic regression with hyper parameter tuning optimization is presented in this research, with ejection fraction, high blood pressure, age, and serum creatinine as relevant risk factors. This study indicates that better data preparation utilizing Deep Learning with hyper parameter adjustment be used to determine the best parameter that has a substantial influence as a risk factor for heart failure. The experiments employed data from the Faisalabad Institute of Cardiology and Allied Hospital in Faisalabad (Punjab, Pakistan), which included 299 samples. The experimental findings reveal that the proposed approach obtains a recall of 63.16% greater than related works.

High Resolution Model of the Vinton Salt-Dome Cap Rock by Joint Inversion of the Full Tensor Gravity Gradient Data with the Simulated Annealing Global Optimization Method

We present a 3D high-resolution modeling methodology based on the interpretation of gravity gradient data and its joint inversion with the simulated annealing (SA) global optimization method. The geometry of the model, used as computational domain in the solution of the forward and inverse problems, is defined with an irregular ensemble of cubic prisms that recreates the interpreted shape of the target, derived from the results of applying different interpretation methods to the gravity gradient data. In our inversion approach, the linear inverse problem resulting from the domain discretization is not solved. Instead, the cost function is explored with the SA algorithm, its low misfit region is identified, and models belonging to it are selected for obtaining the mean model, which represents the most likely model among them, as well as for estimating its uncertainty. The SA inversion algorithm we applied was numerically optimized to reduce the computational burden required by the forward problem, and it was driven by optimal tuning parameters, determined by a parametric analysis. Tests on synthetic data show the efficiency of our methodology to obtain a model that approximates the synthetic target and the usefulness of the estimated uncertainty to complement the interpretation. Finally, by applying our methodology to gravity gradient data acquired over the Vinton dome located in Louisiana, USA, we obtained results that are in agreement with geological information and previous studies.

Adaptation of the tuning parameter in general Bayesian inference with robust divergence

We introduce a novel methodology for robust Bayesian estimation with robust divergence (e.g., density power divergence or gamma -divergence), indexed by tuning parameters. It is well known that the posterior density induced by robust divergence gives highly robust estimators against outliers if the tuning parameter is appropriately and carefully chosen. In a Bayesian framework, one way to find the optimal tuning parameter would be using evidence (marginal likelihood). However, we theoretically and numerically illustrate that evidence induced by the density power divergence does not work to select the optimal tuning parameter since robust divergence is not regarded as a statistical model. To overcome the problems, we treat the exponential of robust divergence as an unnormalisable statistical model, and we estimate the tuning parameter by minimising the Hyvarinen score. We also provide adaptive computational methods based on sequential Monte Carlo samplers, enabling us to obtain the optimal tuning parameter and samples from posterior distributions simultaneously. The empirical performance of the proposed method through simulations and an application to real data are also provided.

Self-tuning vibration control using piezoelectric patches and RL shunts set to maximise electric power absorption

This paper presents experimental results on the flexural vibration response of a thin plate structure subjected to a stochastic disturbance. The plate is equipped with five piezoelectric patches connected to shunt circuitry to reduce the resonant response of a target flexural mode. The shunts are synthesized digitally to mimic a resistor-inductor connected in parallel. Firstly, the study assesses the vibration control effects produced by the shunts, contrasting the time-averaged flexural kinetic energy of the panel and the time-averaged electric power absorbed by the shunts filtered in a narrow band centred on the resonance frequency of the target mode. This analysis shows that both cost functions results on the same optimal tuning parameters, which provides a benchmark for the vibration control. Next, it investigates the online tuning of the resistive and inductive components of the shunt by implementing an extremum seeking algorithm set to maximise the time-averaged electric power absorbed locally by the shunt filtered in a narrow band centred on the resonance frequency of a target mode. The study shows that indeed the resistive and inductive components in the shunts connected to the five patches converges to the optimal values with the application of extremum seeking algorithm and this leads to a minimisation of the spectral flexural response vibration at the resonance frequency of the target mode by 18 dB.

A hybrid deep learning classifier and Optimized Key Windowing approach for drift detection and adaption

The generation of huge data with high velocity creates alteration in the distribution of the stream data, which is defined as the concept of drifts. The concept drifts negatively influence the classification accuracy and the stability of the data streams. Numerous machine learning-based models are developed to detect the concept drift in machine learning techniques. Yet, these models are inadequate for real-time applications due to time and memory constraints. Hence, this research devises dynamic streaming data analytics depending on the optimized hybrid deep learning classifier and Optimized Key Windowing (OKW) approach to effectively handle the time and memory constraints. An optimized hybrid deep learning classifier is the base classifier model developed by integrating deep Long short-term memory (LSTM) and deep Recurrent Neural Networks (RNN) to detect the concept drifts in streaming data. The model’s main advantage lies in enhanced accuracy in drift detection by a hybrid classifier due to optimal hyper parameter tuning, which is performed by the proposed intelligent preying algorithm. The other advantage of the proposed model lies in adapting the classifier with varying data patterns, which is performed using OKW. The experimentation is done using the benchmark dataset, such as Apache, Hadoop, Linux, Spark, and Cloud monitoring. The experimental analysis demonstrates the proposed model attaining higher sensitivity, accuracy, specificity, precision, and recall for Hadoop data. This shows a higher result than that of the competent techniques.

A constrained model predictive control for the building thermal management with optimal setting design

<span lang="EN-US">Today, the building sector is the most important consumer of energy. The main challenge in building management is to obtain the desired performance taking into account many aspects such as comfort requirements, variation of building physical characteristics, system constraints, and energy management. For this purpose, a predictive control approach applied to the building thermal has been designed to achieve desired performances combined with an energy optimization approach based on intrinsic system parameters. The developed approach is applied with an online identification system for effective predictive control to take into account the reel building characteristics and to choose the optimal tuning parameters. The simulation results show good performances in terms of accuracy and robustness face to internal and external disturbances with respect to system constraints.</span>

In-vivo along muscle fascicle strain heterogeneity is not affected by image registration parameters: Robustness testing of combined magnetic resonance-diffusion tensor imaging method

Coupled with diffusion tractography, non-rigid registration of high-resolution anatomical MR images allows the calculation of local strains along human skeletal muscle fascicles in-vivo. A reference study (passively imposed lengthening of gastrocnemius medial muscle) reported local shortening and lengthening occurring along the same muscle fascicles. However, the robustness of strain amplitudes and distribution patterns should be studied, as the heterogeneity of local length changes has major implications for muscle function. Using a previous image set of human medial gastrocnemius (GM) we aimed at testing: (1) the consistency of our MRI-DTI analysis workflow against changes made to the software environments, (2) the hypothesis that non-rigid demons algorithm tuning parameters (16 paired combinations were tested) are not a significant determinant of muscle fiber direction strain heterogeneity caused by passive knee extension. A profoundly altered analysis workflow did reproduce the original results well, showing a general pattern of proximally lengthened and distally shortened muscle fascicles (strain amplitude range: 21%–67%). Hierarchical shift function analyses and pairwise comparison of strain distributions between 10 equal parts of the tracked GM fascicles confirmed the hypothesis showing no significant effects of tuning parameters determining the in-vivo deformation field inhomogeneity. The findings show the robustness of the MRI-DTI method, and confirming the hypothesis, also the consistency of along muscle fascicle strain heterogeneity patterns against parameter selection. However, the strain amplitudes do vary with parameter choices. New studies are indicated to determine optimal tuning parameters to achieve accurate strain amplitudes compared to exact strains.

Tunicate Swarm Algorithm with Deep Learning Based Land Use and Cover Change Detection in Nallamalla Forest India

Every biological system on the planet is severely impacted by environmental change, and its primary driver is deforestation. Meanwhile, quantitative analysis of changes in Land Use and Land Cover (LULC) is one of the prominent ways to manage and understand land transformation; thus, it is essential to inspect the performance of various techniques for LULC mapping to recognize the better classifier to more applications of earth observation. This article develops a Tunicate Swarm Algorithm with Deep Learning Enabled Land Use and Land Cover Change Detection (TSADL-LULCCD) technique in Nallamalla Forest, India. The presented TSADL-LULCCD technique mainly focuses on the identification and classification of land use in the Nallamalla forest using LANDSAT images. To accomplish this, the presented TSADL-LULCCD technique employs a dense EfficientNet model for feature extraction. In addition, the Adam optimizer is applied for the optimal hyper parameter tuning of the dense EfficientNet approach. For land cover classification, the TSADL-LULCCD technique exploits the Deep Belief Network (DBN) approach. To tune the hyper parameters related to the DBN system, the TSA is used. The experimental validation of the TSADL-LULCCD algorithm is tested on LANDSAT-7-based Nallamalla region images. The experimental results stated that the TSADL-LULCCD technique exhibits better performance over other existing models in terms of different evaluation measures.

Intelligent Machine Learning with Metaheuristics Based Sentiment Analysis and Classification

Sentiment Analysis (SA) is one of the subfields in Natural Language Processing (NLP) which focuses on identification and extraction of opinions that exist in the text provided across reviews, social media, blogs, news, and so on. SA has the ability to handle the drastically-increasing unstructured text by transforming them into structured data with the help of NLP and open source tools. The current research work designs a novel Modified Red Deer Algorithm (MRDA) Extreme Learning Machine Sparse Autoencoder (ELMSAE) model for SA and classification. The proposed MRDA-ELMSAE technique initially performs preprocessing to transform the data into a compatible format. Moreover, TF-IDF vectorizer is employed in the extraction of features while ELMSAE model is applied in the classification of sentiments. Furthermore, optimal parameter tuning is done for ELMSAE model using MRDA technique. A wide range of simulation analyses was carried out and results from comparative analysis establish the enhanced efficiency of MRDA-ELMSAE technique against other recent techniques.

Comparison of Multi-layer Perceptron and Support Vector Machine Methods on Rainfall Data with Optimal Parameter Tuning

Comparison of Multi-layer Perceptron and Support Vector Machine Methods on Rainfall Data with Optimal Parameter Tuning

Optimized Rational Dilation Wavelet Transform for Automatic Imagined Speech Recognition

Imagined speech is a process in which a person imagines words without saying them. Electroencephalogram (EEG)-based brain-computer interfaces (BCI) systems help in automatically identifying imagined speech to facilitate persons with severe brain disorders. Extracting meaningful information from the raw EEG signal is a challenging task due to the nonstationary nature of EEG signals. Decomposing a signal into several sub-bands (SB) using rational dilation wavelet transform (RADWT) requires selecting predefined factual parameters, which is an arduous task. The main objective of this study is to propose an adaptive RADWT method capable of decomposing EEG signals by adaptively selecting the tuning parameters and classifying the EEG signals into distinct categories. The optimum tuning parameters of RADWT are obtained using particle swarm optimization and used to decompose the EEG signals into several SBs. Several statistical features are elicited from each SB and used to input six different machine learning algorithms. This work employs a 64-channel EEG dataset recorded from 15 healthy people for three categories: long words, short words, and vowels. The performance of the proposed AISR system is evaluated using seven performance evaluation metrics: accuracy, recall, precision, Cohen’s kappa, F1-score, and area-under-the-curve. The proposed system achieved the average classification accuracies of 87.26±1.12%, 89.23±0.95%, 95.5±0.68%, and 92.16±0.83% for long words, short-long words, short words, and vowels, respectively. When compared to the existing state-of-the-art, the proposed non-parametric decomposition approach and the Bagging algorithm achieved a 3%-5% improvement. The performance of the proposed method is validated using an open-access dataset.

Fuzzy TSMCSPO for Trajectory Tracking of Nuclear Reactor Dismantlement Robot Manipulator

In this study, a fuzzy logic system to tune the parameters of terminal sliding mode control with a sliding perturbation observer (TSMCSPO) in real-time is proposed according to the system’s state changes to improve the control performance of TSMCSPO for a 5-DOF robot manipulator used for nuclear reactor dismantlement. Accurate trajectory tracking control is required when using a multi-DOF robot manipulator to cut the decommissioned nuclear reactor. A TSMCSPO scheme using TSMC in SPO has been proposed in previous research, which improves the estimation and convergence performance of traditional SMCSPO. In TSMCSPO, the parameters of the controller are important to enhance the control performance. Moreover, due to the influence of the working environment and the uncertainty of the system, the characteristics of the system will be changed during the trajectory tracking control of the multi-degree-of-freedom robot manipulator. Therefore, real-time optimal parameter tuning is required. In this regard, using fuzzy logic to tune the parameters is an effective method. By designing fuzzy rules, the controller parameters can be optimized. In this study, the proposed algorithm and, the trajectory tracking control simulation of the robot manipulator in the nuclear reactor vessel internal (RVI) are implemented in MATLAB/Simulink environment. The control algorithm is verified by the experiments on a real robot manipulator platform. The outstanding control performance of the algorithm is demonstrated by comparing the trajectory tracking error between the proposed fuzzy TSMCSPO scheme and TSMCSPO with fixed parameters.

강화학습 기반 화학 공정 제어 성능 향상을 위한 보상 함수 시뮬레이션 연구

Process control is vital for operating chemical process safely and efficiently while maintaining product quality. Process control has been implemented by utilizing traditional techniques such as PID (Proportional-Integrate-Derivative) control and MPC (Model Predictive Control). However, these approaches have disadvantages when the PID tuning rules do not provide optimal tuning parameters for various operating scenarios or when the model fails to match the actual plant. RL (Reinforcement Learning) is one of the most prominent data-driven techniques for addressing such issues and has gained popularity in recent years. While RL has been extensively studied and succeeds in controlling chemical processes, the reward function has a significant impact on its performance. Therefore, it is essential to specify the reward function in order to efficiently control the process. In this study, we propose a reward function to enhance the performance of RL and compare the control performances of MPC and RLs with four distinct reward functions. The controllers track the setpoint of the product in the Van de Vusse process and are evaluated based on the deviation between the setpoint and output. RL with the proposed reward function outperformed the other controllers. Its performance was 12% greater than that of MPC and 6-30% greater than other RL controllers. In chemical processes, the control performance of RL is enhanced by incorporating the time term and positive reward in the reward function, thus outperforming conventional control approaches.

Clustering Deviation Index (CDI): a robust and accurate internal measure for evaluating scRNA-seq data clustering

Most single-cell RNA sequencing (scRNA-seq) analyses begin with cell clustering; thus, the clustering accuracy considerably impacts the validity of downstream analyses. In contrast with the abundance of clustering methods, the tools to assess the clustering accuracy are limited. We propose a new Clustering Deviation Index (CDI) that measures the deviation of any clustering label set from the observed single-cell data. We conduct in silico and experimental scRNA-seq studies to show that CDI can select the optimal clustering label set. As a result, CDI also informs the optimal tuning parameters for any given clustering method and the correct number of cluster components.