Optimal Model Parameters Research Articles

A Meta-Heuristic Optimization Procedure for the Identification of the Nonlinear Model Parameters of Hydraulic Dampers Based on Experimental Dataset of Real Working Conditions

Abstract Hydralic dampers are widely implemented in railway vehicle suspension stages, especially in high-speed passenger trains. They are designed to be mounted in different positions to improve comfort, stability, and safety performances. Numerical simulations are often used to assist the design and optimization of these components. Unfortunately, hydraulic dampers are highly nonlinear due to the complex fluid dynamic phenomena taking place inside the chambers and through the by-pass orifices. This requires accurate damper models to be developed to estimate the influence of the nonlinearities of such components during the dynamic performances of the whole vehicle. This work aims at presenting a new parametric damper model based on a nonlinear lumped element approach. Moreover, a new model tuning procedure will be introduced. Differently from the typical sinusoidal characterization cycles, this routine is based on experimental tests of real working conditions. The set of optimal model parameters will be found through a metaheuristic iterative approach able to minimize the differences between numerical and experimental damper forces. The performances of the optimal model will be compared with the ones of the most common Maxwell model generally implemented in railway multibody software programs.

Improving snow depth simulations on Arctic Sea ice by assimilating a passive microwave-derived record

The snow depth can modulate sea ice changes and is a necessary input parameter to obtain altimeter-derived sea ice thickness values. First, we determine the optimal model parameters for simulating snow depth based on a two-layer snow depth model. Then, we use the particle filter (PF) approach to assimilate satellite-derived snow depth observations to improve the snow depth simulations. Next, we generate daily snow depth estimates from 2012 to 2020 at a 50-km resolution. With the use of three airborne Operation IceBridge (OIB) datasets (i.e., the NSIDC OIB quick look product, NSIDC OIB L4 product and NOAA OIB product), this work reveals that compared to the original simulated snow depths, the snow depth estimates are improved, with root mean square error (RMSE) decreases of 0.8 cm (12%), 1.3 cm (22%) and 1.1 cm (15%), respectively, corresponding to the three OIB products. With the use of Multidisciplinary Drifting Observatory for the Study of Arctic Climate (MOSAiC) snow buoy and Snow and Ice Mass Balance Array (SIMBA) buoy data, it can be verified that the application of PF improves snow depth simulations in the Central Arctic. The variations in the monthly and seasonal snow depth estimates retrieved from the proposed method agree well with those in the estimates retrieved from two other existing algorithms. Based on the presented snow depth estimates, we can perform long-term snow depth and sea ice analysis. Snow depth estimates improve the understanding of Arctic environmental change and promote the future development of sea ice models.

Constructing GRACE-Based 1 km Resolution Groundwater Storage Anomalies in Arid Regions Using an Improved Machine Learning Downscaling Method: A Case Study in Alxa League, China

Using the Gravity Recovery and Climate Experiment (GRACE) satellite to monitor groundwater storage (GWS) anomalies (GWSAs) at the local scale is difficult due to the low spatial resolution of GRACE. Many attempts have been made to downscale GRACE-based GWSAs to a finer resolution using statistical downscaling approaches. However, the time-lag effect of GWSAs relative to environmental variables and optimal model parameters is always ignored, making it challenging to achieve good spatial downscaling, especially for arid regions with longer groundwater infiltration paths. In this paper, we present a novel spatial downscaling method for constructing GRACE-based 1 km-resolution GWSAs by using the back propagation neural network (BPNN) and considering the time-lag effect and the number of hidden neurons in the model. The method was validated in Alxa League, China. The results show that a good simulation performance was achieved by adopting varying lag times (from 0 to 4 months) for the environmental variables and 14 hidden neurons for all the networks, with a mean correlation coefficient (CC) of 0.81 and a mean root-mean-square error (RMSE) of 0.70 cm for each month from April 2002 to December 2020. The downscaled GWSAs were highly consistent with the original data in terms of long-term temporal variations (the decline rate of the GWSAs was about −0.40 ± 0.01 cm/year) and spatial distribution. This study provides a feasible approach for downscaling GRACE data to 1 km resolution in arid regions, thereby assisting with the sustainable management and conservation of groundwater resources at different scales.

Multi-step wind speed prediction based on LSSVM combined with ESMD and fractional-order beetle swarm optimization

Accurate and stable wind speed prediction can alleviate the uncertain impacts of wind power generation caused by nonlinear characteristics of wind speed, and then improve the reliability of wind power. In this paper, a hybrid model for wind speed prediction based on mode decomposition, parameter optimization and basic prediction model is proposed. First, the extreme-point symmetric mode decomposition (ESMD) is employed to adaptively decompose the denoised wind speed time series into sub-sequences with different frequencies. Second, a fractional-order beetle swarm optimization (FO-BSO) for parameter optimization of the Least squares support vector machine (LSSVM) is proposed. Through benchmark functions and non-parametric statistical test, the advantages of the FO-BSO in accuracy, stability and convergence speed are verified. Subsequently, the ESMD-FO-BSO-LSSVM prediction model is established, and three groups of wind speed datasets with different sampling locations and sampling frequencies are selected for simulation experiments. The results show that the coefficient of determination of 1-step prediction of the proposed model in three datasets are 0.9856, 0.9713, 0.9940, which has 2.43%, 3.38%, 3.08% average promotion than that of 7 comparative models. And the accuracy and stability of ESMD-FO-BSO-LSSVM model in multi-step wind speed prediction have also achieved better performance than 7 competitors.

Bayesian Collaborative Learning for Whole-Slide Image Classification.

Whole-slide image (WSI) classification is fundamental to computational pathology, which is challenging in extra-high resolution, expensive manual annotation, data heterogeneity, etc. Multiple instance learning (MIL) provides a promising way towards WSI classification, which nevertheless suffers from the memory bottleneck issue inherently, due to the gigapixel high resolution. To avoid this issue, the overwhelming majority of existing approaches have to decouple the feature encoder and the MIL aggregator in MIL networks, which may largely degrade the performance. Towards this end, this paper presents a Bayesian Collaborative Learning (BCL) framework to address the memory bottleneck issue with WSI classification. Our basic idea is to introduce an auxiliary patch classifier to interact with the target MIL classifier to be learned, so that the feature encoder and the MIL aggregator in the MIL classifier can be learned collaboratively while preventing the memory bottleneck issue. Such a collaborative learning procedure is formulated under a unified Bayesian probabilistic framework and a principled Expectation-Maximization algorithm is developed to infer the optimal model parameters iteratively. As an implementation of the E-step, an effective quality-aware pseudo labeling strategy is also suggested. The proposed BCL is extensively evaluated on three publicly available WSI datasets, i.e., CAMELYON16, TCGA-NSCLC and TCGA-RCC, achieving an AUC of 95.6%, 96.0% and 97.5% respectively, which consistently outperforms all the methods compared. Comprehensive analysis and discussion will also be presented for in-depth understanding of the method. To promote future work, our source code is released at: https://github.com/Zero-We/BCL.

Critical Evaluation and Thermodynamic Re-Optimization of the Si-P and Si-Fe-P Systems.

Thermodynamic modeling of the Si-P and Si-Fe-P systems was performed using the CALculation of PHAse Diagram (CALPHAD) method based on critical evaluation of available experimental data in the literature. The liquid and solid solutions were described using the Modified Quasichemical Model accounting for the short-range ordering and Compound Energy Formalism considering the crystallographic structure, respectively. In the present study, the phase boundaries for the liquidus and solid Si phases of the Si-P system were reoptimized. Furthermore, the Gibbs energies of the liquid solution, (Fe)3(P,Si)1, (Fe)2(P,Si)1, and (Fe)1(P,Si)1 solid solutions and FeSi4P4 compound were carefully determined to resolve the discrepancies in previously assessed vertical sections, isothermal sections of phase diagrams, and liquid surface projection of the Si-Fe-P system. These thermodynamic data are of great necessity for a sound description of the entire Si-Fe-P system. The optimized model parameters from the present study can be used to predict any unexplored phase diagrams and thermodynamic properties within the Si-Fe-P alloys.

뉴로모픽 환경에서 QoS를 고려한 최적의 SNN 모델 파라미터 생성 기법

IoT edge services utilizing neuromorphic hardware architectures are suitable for autonomous IoT applications as they perform intelligent processing on the device itself. However, spiking neural networks applied to neuromorphic hardware are difficult for IoT developers to comprehend due to their complex structures and various hyper-parameters. In this paper, we propose a method for generating spiking neural network (SNN) models that satisfy user performance requirements while considering the constraints of neuromorphic hardware. Our proposed method utilizes previously trained models from pre-processed data to find optimal SNN model parameters from profiling data. Comparing our method to a naive search method, both methods satisfy user requirements, but our proposed method shows better performance in terms of runtime. Additionally, even if the constraints of new hardware are not clearly known, the proposed method can provide high scalability by utilizing the profiled data of the hardware.

Encoding nonlinear and unsteady aerodynamics of limit cycle oscillations using nonlinear sparse Bayesian learning

This article investigates the applicability of a recently proposed, nonlinear sparse Bayesian learning (NSBL) algorithm to identify and estimate the complex aerodynamics of limit cycle oscillations. NSBL provides a semi-analytical framework for determining the data-optimal sparse model nested within a (potentially) over-parameterized model. This is particularly relevant to nonlinear dynamical systems where modelling approaches involve the use of physics-based and data-driven components. In such cases, the data-driven components, where analytical descriptions of the physical processes are not readily available, are often prone to overfitting, meaning that the empirical aspects of these models will often involve the calibration of an unnecessarily large number of parameters. While an overparameterized model may fit the observed data well, such models may be inadequate for making predictions in regimes that are different from those wherein the data were recorded. In view of this, it is desirable to not only calibrate the model parameters, but also identify the optimal compromise between data fit and model complexity. In this article, we exhibit the optimal model discovery for an aeroelastic system wherein the structural dynamics are well-known and described by a differential equation model, coupled with a semi-empirical aerodynamic model for laminar separation flutter, resulting in low-amplitude limit cycle oscillations (LCO). To illustrate the performance of the algorithm, in this article, we use synthetic data and demonstrate the ability of the algorithm to correctly rediscover the optimal model and model parameters, given a known data-generating model. The synthetic data are generated from a forward simulation of a known differential equation model with parameters selected so as to mimic the dynamics observed in wind-tunnel experiments. Subsequently, we demonstrate the performance of the algorithm for model selection using noisy LCO data from wind tunnel experiments. As there is no ground truth available for the experimental data case, we provide a comparison between NSBL and Bayesian model selection to validate the results, and demonstrate the use of NSBL as an efficient alternative to traditional methods.

Language independent optimization of text readability formulas with deep reinforcement learning

Abstract Readability formulas are used to assess the level of difficulty of a text. These language dependent formulas are introduced with pre-defined parameters. Deep reinforcement learning models can be used for parameter optimization. In this article we argue that an Actor-Critic based model can be used to optimize the parameters in the readability formulas. Furthermore, a selection model is proposed for selecting the most suitable formula to assess the readability of the input text. English and Persian data sets are used for both training and testing. The experimental results of the parameter optimization model show that, on average, the F-score of the model for English increases from 24.7% in the baseline to 38.8%, and for Persian from 23.5% to 47.7%. The proposed algorithm selection model further improves the parameter optimization model to 65.5% based on F-score for both English and Persian.

Functional data clustering via information maximization

ABSTRACT A novel method for clustering functional data is introduced that utilizes information maximization. The method employs unsupervised learning to develop a probabilistic classifier that maximizes the mutual information or squared loss mutual information between data points and their corresponding cluster assignments. A significant advantage of this method is that it involves only continuous optimization of model parameters, which is simpler than discrete optimization of cluster assignments and avoids the drawbacks of generative models. Unlike existing methods, this method does not necessitate the estimation of probability densities of Karhunen-Loeve expansion scores under different clusters and does not require the common eigenfunction assumption. The efficacy of the proposed method is demonstrated through simulation studies and real data analysis. Additionally, the technique permits out-of-sample clustering, and its performance is comparable to that of supervised classifiers.

Capacity Consistency Prediction and Process Parameter Optimization of Lithium‐Ion Battery based on Neural Network and Particle Swarm Optimization Algorithm

AbstractThe grading capacity of lithium‐ion battery is an important basis for evaluating battery quality. Aiming at the difficulty of determining the critical control factors and the threshold of the control parameters in the lithium‐ion battery manufacturing process, a capacity prediction and process parameter optimization model of lithium‐ion battery is proposed by combining the back propagation (BP) and particle swarm optimization (PSO) algorithms. First, the BP method is applied to establish the nonlinear mapping relationship between process data and grading capacity, which is regarded as the capacity consistency prediction model. Second, using the prediction model as fitness function and combining with PSO algorithm, the optimization model of process parameters is established. Finally, under the given initial process parameters from the lithium‐ion battery pilot line, it is carried out to obtain the best process parameter formula. The results show that the BP method has an accurate capacity consistency prediction effect. Combined with PSO algorithm, the optimized process parameters are obtained, which significantly improves the capacity consistency of lithium‐ion batteries. The results serve as an engineering application method to guide the selection and confirmation of process parameters at the battery design stage.

Modeling and process parameter optimization of laser cutting based on artificial neural network and intelligent optimization algorithm

Modeling and process parameter optimization of laser cutting based on artificial neural network and intelligent optimization algorithm

Temperature- and pH-Dependent Kinetics of the Aqueous Phase Hydrogen Sulfide Scavenging Reactions with MEA-Triazine

A novel kinetic model for the aqueous phase hydrogen sulfide scavenging reactions using MEA-triazine (HET) is proposed. The assumptions of the model are based on experimental observations obtained by NMR spectroscopy, supporting the existence of 3,5-bis(2-hydroxyethyl)hexahydro-1,3,5-thiadiazine (TDZ) as a quantitative reaction intermediate and showing the protonation behavior of HET and the lack of protonation of 5-(2-hydroxyethyl)hexahydro-1,3,5-dithiazine (DTZ). Experimental kinetic data were obtained with a new in situ Raman spectroscopy setup, which enabled monitoring the time-variation of bisulfide concentrations in a batch stirred reacting system at temperatures of up to 75 °C for HET/HS– initial concentration ratios from 0.5 to 5. The optimal model parameters were regressed from the experimental data using a brute force optimization method. The rate constants of the first and second scavenging reactions were estimated to be 0.435 and 0.004 L mol–1 s–1 at 25 °C, and the activation energies were 68 and 57 kJ mol–1, respectively.

Efficient Calibration of Groundwater Contaminant Transport Models Using Bayesian Optimization.

Numerical modeling is a significant tool to understand the dynamic characteristics of contaminants transport in groundwater. The automatic calibration of highly parametrized and computationally intensive numerical models for the simulation of contaminant transport in the groundwater flow system is a challenging task. While existing methods use general optimization techniques to achieve automatic calibration, the large numbers of numerical model evaluations required in the calibration process lead to high computing overhead and limit the efficiency of model calibration. This paper presents a Bayesian optimization (BO) method for efficient calibration of numerical models of groundwater contaminant transport. A Bayes model is built to fully represent calibration criteria and derive the objective function for model calibration. The efficiency of model calibration is made possible by the probabilistic surrogate model and the expected improvement acquisition function in BO. The probabilistic surrogate model approximates the computationally expensive objective function with a closed-form expression that can be computed efficiently, while the expected improvement acquisition function proposes the most promising model parameters to improve the fitness to the calibration criteria and reduce the uncertainty of the surrogate model. These schemes allow us to find the optimized model parameters effectively by using a small number of numerical model evaluations. Two case studies for the calibration of the Cr(VI) transport model demonstrate that the BO method is effective and efficient in the inversion of hypothetical model parameters, the minimization of the objective function, and the adaptation of different model calibration criteria. Specifically, this promising performance is achieved within 200 numerical model evaluations, which substantially reduces the computing budget for model calibration.

A multi-behavior recommendation method exploring the preference differences among various behaviors

User behavior data has been widely used in recent research of recommendation systems. Existing work usually utilize only single behavior instead of multi-behavior. However, there are various typed of user behaviors in practical scenarios. For example, there are view, purchase, add-to-cart and add-to-favorite behaviors in a real-world e-commerce platform. For recommendation systems, using more user information is better for exploiting intrinsic user characteristics, as the single behavior records are usually too sparse to conduct preference mining. Therefore, multi-behavior based recommendation methods are increasingly emphasized by researchers. In this paper, we propose a novel framework (FPD) for utilizing multi-behavior data: generating embedding and building training losses. We compute an additional supplementary score by capturing user's preference difference among behaviors, instead of merely using initial embedding to obtain a ordinary prediction score. For better optimization model parameters, we use various behaviors to build multiple training losses. We optimize the loss function of the non-sampling strategy and set personal positive weights for each user. Experimental results on two real-world datasets demonstrates that our model outperformed the state-of-the-art methods in terms of various evaluation measurements. Furthermore, extensive ablation experiments and visualization analysis are conducted to verify the effectiveness of the proposed idea and to further explain the principle of the proposed method proposed.

Modeling Instantaneous Firing Rate of Deep Brain Stimulation Target Neuronal Ensembles in the Basal Ganglia and Thalamus

ObjectiveDeep brain stimulation (DBS) is an effective treatment for movement disorders, including Parkinson disease and essential tremor. However, the underlying mechanisms of DBS remain elusive. Despite the capability of existing models in interpreting experimental data qualitatively, there are very few unified computational models that quantitatively capture the dynamics of the neuronal activity of varying stimulated nuclei—including subthalamic nucleus (STN), substantia nigra pars reticulata (SNr), and ventral intermediate nucleus (Vim)—across different DBS frequencies. Materials and MethodsBoth synthetic and experimental data were used in the model fitting; the synthetic data were generated by an established spiking neuron model that was reported in our previous work, and the experimental data were provided using single-unit microelectrode recordings (MERs) during DBS (microelectrode stimulation). Based on these data, we developed a novel mathematical model to represent the firing rate of neurons receiving DBS, including neurons in STN, SNr, and Vim—across different DBS frequencies. In our model, the DBS pulses were filtered through a synapse model and a nonlinear transfer function to formulate the firing rate variability. For each DBS-targeted nucleus, we fitted a single set of optimal model parameters consistent across varying DBS frequencies. ResultsOur model accurately reproduced the firing rates observed and calculated from both synthetic and experimental data. The optimal model parameters were consistent across different DBS frequencies. ConclusionsThe result of our model fitting was in agreement with experimental single-unit MER data during DBS. Reproducing neuronal firing rates of different nuclei of the basal ganglia and thalamus during DBS can be helpful to further understand the mechanisms of DBS and to potentially optimize stimulation parameters based on their actual effects on neuronal activity.

Simulation Failure-Robust Bayesian Optimization for Data-Driven Parameter Estimation

Advances in modeling and computation have resulted in high-fidelity digital twins capable of simulating the dynamics of a wide range of industrial systems. These simulation models often require calibration, or the estimation of an optimal set of parameters in some goodness-of-fit sense, to reflect a system’s observed behavior. While searching over the parameter space is an inevitable part of the calibration process, simulation models are rarely designed to be valid for arbitrarily large parameter spaces. The application of existing calibration methods, therefore, often results in repeated model evaluations using parameters that can cause the simulations to be impractically slow or even result in catastrophic failure. In general, the shape of subregions in the parameter space that could result in simulation failure is unknown. In this article, we propose a novel failure-robust Bayesian optimization (FR-BO) algorithm that learns these failure regions (FRs) from online simulations and informs a Bayesian optimization algorithm to avoid FRs while optimizing model parameters. This results in acceleration of the optimizer’s convergence and prevents wastage of time trying to simulate parameters with high failure probabilities. The effectiveness of the proposed FR-BO algorithm is demonstrated via a well-known benchmark example where we compare against state-of-the-art gradient matching techniques, and a practical example related to parameter estimation for digital twins of buildings.

High-Performance Adaptive Weak Fault Diagnosis Based on the Global Parameter Optimization Model of a Cascaded Stochastic Resonance System.

Stochastic resonance (SR), as a type of noise-assisted signal processing method, has been widely applied in weak signal detection and mechanical weak fault diagnosis. In order to further improve the weak signal detection performance of SR-based approaches and realize high-performance weak fault diagnosis, a global parameter optimization (GPO) model of a cascaded SR system is proposed in this work. The cascaded SR systems, which involve multiple multi-parameter-adjusting SR systems with both bistable and tri-stable potential functions, are first introduced. The fixed-parameter optimization (FPO) model and the GPO models of the cascaded systems to achieve optimal SR outputs are proposed based on the particle swarm optimization (PSO) algorithm. Simulated results show that the GPO model is capable of achieving a better SR output compared to the FPO model with rather good robustness and stability in detecting low signal-to-noise ratio (SNR) weak signals, and the tri-stable cascaded SR system has a better weak signal detection performance compared to the bistable cascaded SR system. Furthermore, the weak fault diagnosis approach based on the GPO model of the tri-stable cascaded system is proposed, and two rolling bearing weak fault diagnosis experiments are performed, thus verifying the effectiveness of the proposed approach in high-performance adaptive weak fault diagnosis.

Development of a Machine Learning Forecast Model for Global Horizontal Irradiation Adapted to Tibet Based on Visible All-Sky Imaging

The Qinghai-Tibet Plateau is rich in renewable solar energy resources. Under the background of China’s “dual-carbon” strategy, it is of great significance to develop a global horizontal irradiation (GHI) prediction model suitable for Tibet. In the radiation balance budget process of the Earth-atmosphere system, clouds, aerosols, air molecules, water vapor, ozone, CO2 and other components have a direct influence on the solar radiation flux received at the surface. For the descending solar shortwave radiation flux in Tibet, the attenuation effect of clouds is the key variable of the first order. Previous studies have shown that using Artificial intelligence (AI) models to build GHI prediction models is an advanced and effective research method. However, regional localization optimization of model parameters is required according to radiation characteristics in different regions. This study established a set of AI prediction models suitable for Tibet based on ground-based solar shortwave radiation flux observation and cloud cover observation data of whole sky imaging in the Yangbajing area, with the key parameters sensitively tested and optimized. The results show that using the cloud cover as a model input variable can significantly improve the prediction accuracy, and the RMSE of the prediction accuracy is reduced by more than 20% when the forecast horizon is 1 h compared with a model without the cloud cover input. This conclusion is applicable to a scenario with a forecast horizon of less than 4 h. In addition, when the forecast horizon is 1 h, the RMSE of the random forest and long short-term memory models with a 10-min step decreases by 46.1% and 55.8%, respectively, compared with a 1-h step. These conclusions provide a reference for studying GHI prediction models based on ground-based cloud images and machine learning.

Continuous adaptation of a digital twin model for a pilot flotation plant

Model-based methods have a key role in achieving the technical, economical, and environmental performance improvements of the mineral processing systems. However, unmodeled process phenomena and disturbances leading to unreliable modeling results, may prevent the efficient online utilization of these methods at the operational level after the deployment of the model. This study demonstrates the feasibility of online adaptation of a dynamic, mechanistic process models in mineral beneficiation application at a pilot environment. At first, a digital twin of the grinding and flotation stages of a pilot-scale plant was developed. In the experimental campaign, a change from copper-zinc-pyrite ore to a mixture of pyrite-rich and non-sulfide gangue-rich material was carried out. Thus, during experiments, a notable change in the flotation performance was observed, which could not be replicated by a constant-parameter digital twin model. The proposed parameter adaptation framework, encompassing stochastic optimization in moving time windows, was found to be suitable for finding new optimal model parameters during the changing experimental conditions by using the elemental grades in different flotation stages. In addition, simulation studies are presented to highlight the challenges of digital twin parameter adaptation in mineral processing applications, where often only a sparse and disturbance influenced data from the key process variables are available. The adaptive digital twin allows applying the dynamic, mechanistic process models efficiently for predictive simulations in operational decisions leading to more sustainable and resource-efficient minerals processing.