Well-tuned Parameters Research Articles

Effects of PSO Algorithm Parameters on the MPPT System Under Partial Shading Condition

Maximum Power Point Tracking (MPPT) systems enable photovoltaic (PV) panels to work at their Maximum PowerPoint (MPP). To do this, several algorithms have been developed, including conventional, intelligent, and meta-heuristic. Once a partial shading condition (PSC) occurs, more than one peak emerges in the power-voltage curve of photovoltaic arrays. Under PSCs, conventional algorithms get stuck at the local maximum point and fail to reach the global maximum point. Being an alternative method, Particle Swarm Optimization (PSO) algorithm has been frequently employed for MPPT systems under PSCs. This algorithm has some parameters that affect its performance to reach the global MPP of the PV panel. Therefore, with well-tuned parameters, the effectiveness of the PSO will increase for the different PSCs. In this study, the effects of the cognitive learning and social learning parameters of the PSO algorithm are investigated under different PSCs. To achieve this, an MPPT system, including a boost-type DC-DC converter, is created in MATLAB®/Simulink®. Simulation studies show that the PSO algorithm fails to track global MPP with constant cognitive and social learning parameters under changing partial shading conditions. Furthermore, the results show that these two parameters affect the time to reach the MPP of the PSO algorithm.

Cumulative displacement-based detection of damper malfunction in bridges using data-driven isolation forest algorithm

Long-span cable-supported bridges are flexible and prone to significant deformations under temperature, wind, and vehicle loads. Such large deformations induce frequent longitudinal movement at the end of the bridge girder, which can cause malfunction or failure in the components such as expansion joints, bearings, or longitudinal viscous dampers. This paper develops a condition assessment approach using the girder end displacement measurements, which primarily includes four steps, namely, data preprocessing, feature characterization, datasets preparation and splitting, and anomaly detection. The data preprocessing step considers the physics-based correlation between the temperature and displacement, which is used to pre-check noisy data issues in the displacement monitoring data. Subsequently, hourly cumulative displacement is calculated as the feature to feed into the machine learning model. The third step prepares the datasets and splits them into training and testing datasets. Then the isolation forest algorithm is implemented for anomaly detection with well-tuned parameters of the number of decision trees and the number of samples in each decision tree. Structural health monitoring data of a suspension bridge are collected to verify the proposed approach, and the detection targets are specified as viscous damper malfunction and a specific holiday period. The proposed approach demonstrates its ability to detect both events successfully, which can shed light on the predictive and preventive maintenance of the bridge.

Adaptive-passive tuned mass damper for structural aseismic protection including soil–structure interaction

As one of the most traditional vibration mitigation devices, tuned mass dampers (TMDs) are applied widely in aseismic protection of building structures. The control effect of a passive TMD is depended on its parameters, especially the frequency ratio. Nevertheless, passive TMDs are high sensitive to the frequency detuning issue, and a mistuned TMD will lose its aseismic protection. Soil-structure interaction (SSI) will deviate the structural frequency and lead to a mistuned TMD. Besides, different soil conditions will lead to different frequency deviations of the primary structure. However, it may be difficult to obtain parameters of the soil exactly. To solve this problem, a recently developed adaptive-passive eddy current pendulum TMD (APEC-PTMD) is applied to a benchmark 40-story tall building including SSI in this study, and four different soil conditions are considered. The APEC-PTMD can identify the optimal TMD frequency in the building with different soil types, and then retune itself through adjusting the pendulum length, and also the air gap between the conduct plate and permanent magnets to reset its damping ratio. Therefore, no prior knowledge of the soil condition is needed for the APEC-PTMD. To verify the aseismic protection effect, 44 far-field earthquake excitations are chosen, and two passive TMDs are used for comparison, while one is optimized for the base-fixed structure and the other is optimized based on the dense-soil model. The passive TMD will become mistuned when SSI is considered or soil parameters are changed, while APEC-PTMD can adapt to the structural dominant frequency with different SSI. Therefore, it always works as a passive TMD with well-tuned parameters. Results show that the APEC-PTMD has a better aseismic protection than the passive TMD, especially in the soft-soil model, and it has an excellent control effect compared to the without TMD case at the same time.

Multiphysics Modelling and Simulation of Thrombolysis via Activated Platelet-Targeted Nanomedicine

PurposeThis study establishes a multiphysics simulation platform for both conventional and targeted thrombolysis using tissue plasminogen activator (tPA). Based on our computational results, the effects of therapeutic parameters on the dynamics of thrombolysis and the risk of side effects are investigated.MethodsThe model extends our previously developed one-dimensional(1D) mathematical models for fibrinolysis by incorporating targeted thrombolysis. It consists of two parts: (i) a coupled mathematical model of systemic pharmacokinetics (PK) and pharmacodynamics (PD) and local PD in a 1D occluded artery, and (ii) a mechanistic model for a targeted thrombolytic system via activated platelet-targeted tPA-loaded nanovesicles (tPA-NV), with model parameters derived from our in vitro experiments. A total of 16 therapeutic scenarios are simulated by varying the clot location and composition as well as the dosing regimen with free tPA or tPA-NV.ResultsOur simulation results indicate that tPA-NV offers several advantages over free tPA for thrombolysis. It reduces systemic exposure of tPA, thereby minimising the risk of bleeding complications. Simulations with different tPA-NV doses reveal that tPA-NV at 10% of the recommended dose can be as effective as the standard regimen with the full recommended dose of free tPA, demonstrating the potential of our tPA-NV as a new thrombolytic strategy with a reduced tPA dose. Moreover, faster recanalisation can be achieved with tPA-NV, especially for platelet-rich(or fibrin-poor) clots.ConclusionsOur simulation platform for thrombolysis with well-tuned model parameters can be used to evaluate and optimise treatment regimens of existing and new thrombolytic therapies via benefit/risk assessment under various therapeutic scenarios.

Machine Learning Gyro Calibration Method Based on Attitude Estimation

This paper proposes a novel machine learning (ML) gyro calibration method that can achieve higher-accuracy gyro calibration and attitude estimation than the classical extended Kalman filter (EKF) approach when high-accuracy measurements are unavailable. The gyro calibration process is modeled as a time-series problem based on the standard EKF output. Then, a designed ML model is trained by the collected time-series data so that it can conduct gyro calibration by generating an ML correction to the gyro measurement. The proposed ML calibration method does not make assumptions in the form of gyro measurement errors, but directly learns it from data when high-accuracy information is available. Therefore, it is possible to outperform the EKF bias calibration when there is only low-accuracy information. To validate the method, a torque-free CubeSat is simulated using sun sensors and magnetometers to generate higher- and lower-accuracy attitude measurements, respectively. The simulation results show that the ML gyro calibration achieves smaller residual errors compared with the standard EKF. Meanwhile, the EKF attitude estimation accuracy is also improved, as the attitude integration is more accurate using the ML-calibrated gyro measurement. Four ML models based on different principles are examined, including multilayer perceptron, convolutional neural network, recurrent neural network, and Gaussian processes. It is found that, usually, a more sophisticated ML model can capture more gyro error information, but all four models can achieve similar performance with well-tuned parameters.

Enhanced Skin Condition Prediction Through Machine Learning Using Dynamic Training and Testing Augmentation

In recent years, deep learning has taken the spotlight in automated medical bioimaging. However, the performance of current state-of-the-art score stems primarily from well-tuned parameters and architecture. There is still only limited research focused on dynamic data augmentation, even in the fields of machine learning and computer vision. In this study, we propose a dynamic training and testing augmentation capable of increasing performance significantly. The searching augmentation framework used in this study requires fewer GPU hours than a conventional search algorithm, which needs to train a new model every time augmentation is proposed. Speeding up of the search algorithm is achieved by using Bayesian optimization on a trained model, so we do not have to train a new model every time a new augmentation policy is proposed. The performance of our method is compared with that of a single model and the ensemble model that happens to be the winner of the ISIC 2019 challenge. Furthermore, we use the latest compact yet significantly accurate network architecture EfficientNet as the backbone system. Our method delivers a superior result, and this study also shares the searched augmentation policy utilized, which requires extraordinary resources. Thus, other researchers can use the searched augmentation policies for dermoscopic images to improve performance.

Analysis of Synchronverter Self-Synchronization Dynamics to Facilitate Parameter Tuning

This paper proposes a self-synchronizing synchronverter controller design that leverages the addition of a virtual resistance (along with a suitable coordinate transformation) to compute feedback signals during self synchronization prior to grid connection. With respect to analysis, we exploit separation-of-time-scales arguments and develop appropriate reduced-order models, which are well-suited for studying phase-angle and voltage-magnitude self-synchronization dynamics independently. Our work provides analytical justification for the effects of pertinent controller parameters and system initial conditions on self-synchronization dynamics observed empirically in time-domain simulations. As such, it offers practical guidance on favourable parameter-value settings to achieve fast self synchronization, and it yields accurate estimates for self-synchronization times with well-tuned parameters. Through numerical simulations and experiments, we illustrate the efficacy of the proposed controller design and verify the validity of subsequent analyses.

PID Controller Tuning Based on Phase Margin (PM) for Wireless Temperature Control

Performance of proportional-integral-derivative (PID) controller which tuning based on phase margin was investigated. A process simulator modified for wireless process control purpose and antennas with modules connected for wireless communication applications. The impact of robust error control code on wireless temperature control experiments is investigated by means of three tuning parameters of the PID controller which are initially determined by using non-parametric methods. The well-tuned control parameters were assessed by means of a MATLAB/SISO tool (Control System Designer tool) for tuning singe input–single output controllers based on root locus diagram. To determine the bias values of the system, an initial steady state was obtained and the system output was monitored at the constant heater capacity (10%) for 300 s. At the end of 5 min, the control key in MATLAB/Simulink block diagram was changed and the PID control algorithm was activated for different set point changes were given to the system at the same time and the effect of the parameters was observed. It was seen that the wireless temperature control successfully carried out desired set values which selected different set points by using the robust error control with well-tuned parameters. Integral of Squared Error (ISE) and Integral of Absolute Error (IAE) values of different set points were calculated for this PID controller.

Automatic feature learning using multichannel ROI based on deep structured algorithms for computerized lung cancer diagnosis

This study aimed to analyze the ability of extracting automatically generated features using deep structured algorithms in lung nodule CT image diagnosis, and compare its performance with traditional computer aided diagnosis (CADx) systems using hand-crafted features. All of the 1018 cases were acquired from Lung Image Database Consortium (LIDC) public lung cancer database. The nodules were segmented according to four radiologists’ markings, and 13,668 samples were generated by rotating every slice of nodule images. Three multichannel ROI based deep structured algorithms were designed and implemented in this study: convolutional neural network (CNN), deep belief network (DBN), and stacked denoising autoencoder (SDAE). For the comparison purpose, we also implemented a CADx system using hand-crafted features including density features, texture features and morphological features. The performance of every scheme was evaluated by using a 10-fold cross-validation method and an assessment index of the area under the receiver operating characteristic curve (AUC). The observed highest area under the curve (AUC) was 0.899±0.018 achieved by CNN, which was significantly higher than traditional CADx with the AUC=0.848±0.026. The results from DBN was also slightly higher than CADx, while SDAE was slightly lower. By visualizing the automatic generated features, we found some meaningful detectors like curvy stroke detectors from deep structured schemes. The study results showed the deep structured algorithms with automatically generated features can achieve desirable performance in lung nodule diagnosis. With well-tuned parameters and large enough dataset, the deep learning algorithms can have better performance than current popular CADx. We believe the deep learning algorithms with similar data preprocessing procedure can be used in other medical image analysis areas as well.

Composite nonlinear feedback control for output regulation problem of linear discrete-time systems with input saturation

Transient performance for output regulation problems of linear discrete-time systems with input saturation is addressed by using the composite nonlinear feedback(CNF) control technique. The regulator is designed to be an additive combination of a linear regulator part and a nonlinear feedback part. The linear regulator part solves the regulation problem independently which produces a quick output response but large oscillations. The nonlinear feedback part with well-tuned parameters is introduced to improve the transient performance by smoothing the oscillatory convergence. It is shown that the introduction of the nonlinear feedback part does not change the solvability conditions of the linear discrete-time output regulation problem. The effectiveness of transient improvement is illustrated by a numeric example.

An experimental comparison of proportional-integral, sliding mode, and robust adaptive control for piezo-actuated nanopositioning stages

This paper presents a comparative study of the proportional-integral (PI) control, sliding mode control (SMC), and robust adaptive control (RAC) for applications to piezo-actuated nanopositioning stages without the inverse hysteresis construction. For a fair comparison, the control parameters of the SMC and RAC are selected on the basis of the well-tuned parameters of the PI controller under same desired trajectories and sampling frequencies. The comparative results show that the RAC improves the tracking performance by 17 and 37 times than the PI controller in terms of the maximum tracking error e(m) and the root mean tracking error e(rms), respectively, while the RAC improves the tracking performance by 7 and 9 times than the SMC in terms of e(m) and e(rms), respectively.

Driving Performance Advancement for SRM Using a Novel Adaptive Learning Control on Dominant Parameters

In this paper, an adaptive learning control (ALC) based on the repetitive tuning of the dominant energized parameters is proposed to minimize the torque ripple and energy conversion loss of the switched reluctance motors (SRMs). A simple and effective modulating strategy on excitation commutation angles and duty cycle is repeatedly applied to the SRM until desired performance tracking is achieved. According to the load variation, an initial suboptimum operating condition, derived from the expert’s knowledge and experience as well as accurate magnetization characteristics of the SRM, is offered to promise a well starting dynamics, learning ratio and convergence speed. Driving performance enhancement can be obtained on account of the best programming of phase current profiles via the well-tuned parameters to make minimum energy conversion loss and torque ripple reachable. Evolution and simulation based on the DSP-based processor has been developed to realize the proposed control approach and setup the SRM drive system. Experimental results on a 4-phase 8/6 pole SRM are given to show the effectiveness and performance enhancements of the presented control mechanism.

Modelling and simulation of steam turbine processes: individual models for individual tasks

Within power plants, several physical, chemical and mechanical processes are conducted to transfer the energy, stored in fossil fuel, into electrical energy. This energy conversion is divided into several stages. Hitherto, the largest conventional power plants employ steam turbines as prime movers to drive a generator. Hence, a steam turbine is one module to convert heat energy into mechanical energy. And thus it is one link in the chain of energy conversions with the aim of generating electrical energy. Today, steam turbine industry faces numerous challenges concerning efficiency, commissioning time, start-up times, operation, availability, safety, cost-effectiveness, etc. Many of these tasks can be supported by simulating the transient operational behaviour of the turbine in advance. For example, the commissioning time can be shortened if the turbine controllers are initialized with well-tuned pre-set parameters; cost-effectiveness can be increased by setting aside unnecessary devices and exactly determining material specifications; safety may be increased by predicting the impacts of failures and thus taking the necessary precautions. Different tasks require different details regarding the employed turbine simulation model. Thus, the turbine controller may be well tuned with less complex simulation models of turbine, generator and electrical grid, whereas detailed studies of failures, mainly the transient behaviour which may lead to serious damages, may require detailed modelling of the turbine-internal thermodynamic processes. Here, a brief overview of models which simulate the transient thermodynamic behaviour of a steam turbine is presented. Three different approaches will be introduced and compared with respect to different operating situations. Also, special attention is directed towards the time dependence of critical states, mainly turbine speed and pressure development in certain areas. The first model is based on a simple, linear approach and is suitable of giving a quick overview. The second one incorporates more details and is useful if the operating point is close to the design point. Finally, the last model incorporates mass and energy balances as well as the major non-linearities. Hence it depicts the turbine behaviour over a large range of operating points.

Parameter tuning for induction-algorithm-oriented feature elimination

Feature selection has long been an active research topic in machine learning. Beginning with an empty set of features, it selects features most necessary for learning a target concept. Feature elimination, a newer technique, starts out with a full set of features and eliminates those most unnecessary for learning the target concept. Feature elimination tends to be more effective, can capture interacting features more easily, and suffers less from feature interaction than feature selection. Because the most unnecessary features are eliminated from the beginning, they will not mislead the induction process in terms of efficiency or accuracy. Induction-algorithm-oriented feature elimination, with particular parameter configurations, can achieve higher predictive accuracy than existing popular feature selection approaches. We propose two sets of well-tuned parameters based on empirical analysis. To understand how to achieve the best performance possible from IAOFE, we conducted a comprehensive analysis of IAOFE parameter tuning.