Tuning Effort Research Articles

Isolated microgrid frequency stabilization through nonlinear model predictive control of a gas engine generator set

Gas engine generator sets are applied in microgrids due to their capability to provide reliable, responsive and efficient distributed power generation, enhancing grid resilience and enabling the integration of renewable energy sources. This article proposes a methodology for stabilizing the frequency of an isolated microgrid subjected to a measurable disturbance by controlling the power of a premixed turbocharged natural gas engine. Control difficulties of this task include strongly nonlinear dynamics, time delay in the air-fuel path and constrained operating conditions. Conventional control methods like proportional-integral control have difficulties solving these problems, and require extensive tuning efforts and gain scheduling to deliver acceptable performance, prompting adoption of advanced control strategies. The presented approach utilizes a cascaded control architecture with a nonlinear model predictive controller (NMPC) for high level control and engine specific low-level controllers for controlling the intake manifold pressure and air-fuel ratio. The NMPC captures the inherent nonlinear dynamics, accounts for an input delay and handles state-dependent constraints. The cascaded control architecture enables the usage of a comparably simple model for the NMPC design, reducing the computational cost and the parametrization effort. This additionally enables application of the high level controller for different gas engine types and allows design of the controller in a simple simulation environment prior to the experiments on the real engine. Experimental results highlight the NMPC’s ability to control the engine at its physical limits over the entire load range without inducing frequency oscillations using just one parameter set. This emphasizes the robustness of the presented approach, making it a promising solution for real-world applications.

MTDiff: Visual anomaly detection with multi-scale diffusion models

Advancements in computer vision have fueled rapid developments in unsupervised anomaly detection, but current methods often encounter limitations when addressing anomalies with varying scales, and the intricate pipelines that require significant tuning efforts further hinder the usability. In this work, we propose MTDiff, a novel anomaly detection method comprising diffusion models built on different scales. In essence, the constructed scale-specific branches and their incorporation can enhance the pattern coverage, thus improving performance. MTDiff involves two parts: reconstruction that repairs the anomalous region to pseudo-normal, and detection that carefully compares and localizes the anomalies. Instead of the typical forward process of diffusion, we construct a partial Markov chain to improve the reconstruction quality. During the discrimination, we construct a simple but effective detector that operates on feature-level to better utilize the rich contextual information. MTDiff comes with a concise training pipeline, with optimized diffusion iterations ensuring the efficiency. Sufficient experiments reveal that it outperforms the state-of-the-art approaches, showing superior stability and robustness in both image- and pixel-level anomaly detection. The related code is available at https://github.com/vergilben/MTDiff.

Spatio-temporal deep recurrent convolutional neural network for infrared focal plane arrays non-uniformity correction

The inherent non-uniformity generated in infrared focal plane arrays (IRFPA) due to inconsistent responses of detectors, and it significantly degrades the performance of infrared imaging systems. Traditional Non-Uniformity Correction (NUC) methods mainly leverage the temporal low-frequency property of non-uniformity to remove it. In recent years, Convolutional Neural Network (CNN) has gained attention as a powerful tool for conducting NUC using solely spatial information from a single frame. In this study, we propose a deep neural network with a recurrent spatio-temporal feature fusion module to enhance CNN performance using inter-frame information. Initially, the proposed network uses CNN to coarsely extract FPN features solely from the spatial domain. Subsequently, the network merges it with the historical estimation under the guidance of two gating signals, which calculated from spatio-temporal information with gated convolutional layers. Both stages are jointly trained to ensure that the entire network operates end-to-end and adapts to different scenarios, eliminating complex parameter tuning efforts. Experiments demonstrated that the proposed network shows both the advantages of traditional and CNN-based NUC methods. It has a higher convergence starting point, faster convergence speed, and a 1 ∼ 2 dB improvement in PSNR compared to both types of methods.

Revisiting CEC 2022 ranking: A new ranking method and influence of parameter tuning

Optimization competitions give an impulse to develop optimization algorithms. However, there is no common agreement on how to rank the contestants. This paper proposes a method of assessing the performance of the algorithms. The proposed mark is easily interpretable by humans and can be compared with previously published marks. The proposed method was used to create a ranking for contestants of the CEC 2022 competition on single objective bound constrained numerical optimization. The resulting ranking is different from the official one. The ranks of some algorithms differ by up to five. As the proposed ranking is more focused on the results at the end of the budget, winning algorithms are better suited for most real-world applications. Since the tuning effort influences the algorithm’s results, the paper also examines the influence of parameter tuning on the place achieved in both rankings by the top four algorithms. The parameters to tune were extracted from the papers that introduced the algorithms and from the source codes. The results showed that all considered algorithms were not carefully tuned for CEC. Generally, tuning using the target performance metric helps, but tuning using the other metric is harmful. The profit of tuning is up to a 33% increase in the number of trails that found the global optimum. The ablation analyses of the algorithms’ parameters showed that only a few parameters strongly influence the results. Frequently, parameters not listed in the papers are among the most important.

Enhancing Electric Vehicle Performance with a Hybrid PI-Sliding Mode Controller for Battery Supercapacitor Integration

Nowadays, most of the works are based on electric vehicle usage for sustainable transportation using traditional energy storage device, such as battery. Usage of batteries in electric vehicles is having several disadvantages, for example, life span, temperature, and charge estimation. In this paper, a novel control scheme for battery and supercapacitor- (SC-) based hybrid energy storage system (HESS) using hybrid proportional and integral- (PI-) sliding mode control (SMC) for electric vehicle (EV) applications is introduced and implemented. This HESS with hybrid controller proves the usage of batteries in EVs to its fullest potential. The conventional control strategy for HESS follows two-loop voltage and current PI controllers with low-pass filter (LPF) and involves tuning of multiple control parameters with variations of source and load disturbances. Performance of the system is affected by tuning PI controller constants. A slow response time with linear PI controllers is long which is not advisable for starting and sudden jerk conditions of EVs. Moreover, the PI controller performance is affected by the system parameter variations during load changes. And these parameters are dynamic in nature due to nonideal conditions. In this paper, a hybrid PI-sliding mode controller (SMC) scheme is designed to control the bidirectional DC-DC converters to overcome the drawbacks of aforementioned issues. The combined PI-SMC controller reduces the tuning effort and reduces the effect of shift in operating point in controller performance. Linear modeling is done using small signal analysis for each subsystems. Permanent magnet synchronous machine (PMSM) is used as electric vehicle. The entire system and its controllers are simulated using MATLAB-Simulink, and detailed comparison is carried between conventional PI and proposed hybrid PI-SMC scheme to regulate the DC link voltage. The results are tabulated and show that the hybrid PI-SMC scheme outperforms in transient and steady-state conditions than the traditional PI controller. A scaled hardware prototype of 48 W set-up is developed using dSPACE-1104, and the experimental results have been carried out to verify the proposed system’s feasibility.

Adaptive Anytime Multi-Agent Path Finding Using Bandit-Based Large Neighborhood Search

Anytime multi-agent path finding (MAPF) is a promising approach to scalable path optimization in large-scale multi-agent systems. State-of-the-art anytime MAPF is based on Large Neighborhood Search (LNS), where a fast initial solution is iteratively optimized by destroying and repairing a fixed number of parts, i.e., the neighborhood of the solution, using randomized destroy heuristics and prioritized planning. Despite their recent success in various MAPF instances, current LNS-based approaches lack exploration and flexibility due to greedy optimization with a fixed neighborhood size which can lead to low-quality solutions in general. So far, these limitations have been addressed with extensive prior effort in tuning or offline machine learning beyond actual planning. In this paper, we focus on online learning in LNS and propose Bandit-based Adaptive LArge Neighborhood search Combined with Exploration (BALANCE). BALANCE uses a bi-level multi-armed bandit scheme to adapt the selection of destroy heuristics and neighborhood sizes on the fly during search. We evaluate BALANCE on multiple maps from the MAPF benchmark set and empirically demonstrate performance improvements of at least 50% compared to state-of-the-art anytime MAPF in large-scale scenarios. We find that Thompson Sampling performs particularly well compared to alternative multi-armed bandit algorithms.

A Non-parametric Graph Clustering Framework for Multi-View Data

Multi-view graph clustering (MVGC) derives encouraging grouping results by seamlessly integrating abundant information inside heterogeneous data, and has captured surging focus recently. Nevertheless, the majority of current MVGC works involve at least one hyper-parameter, which not only requires additional efforts for tuning, but also leads to a complicated solving procedure, largely harming the flexibility and scalability of corresponding algorithms. To this end, in the article we are devoted to getting rid of hyper-parameters, and devise a non-parametric graph clustering (NpGC) framework to more practically partition multi-view data. To be specific, we hold that hyper-parameters play a role in balancing error item and regularization item so as to form high-quality clustering representations. Therefore, under without the assistance of hyper-parameters, how to acquire high-quality representations becomes the key. Inspired by this, we adopt two types of anchors, view-related and view-unrelated, to concurrently mine exclusive characteristics and common characteristics among views. Then, all anchors' information is gathered together via a consensus bipartite graph. By such ways, NpGC extracts both complementary and consistent multi-view features, thereby obtaining superior clustering results. Also, linear complexities enable it to handle datasets with over 120000 samples. Numerous experiments reveal NpGC's strong points compared to lots of classical approaches.

Modelling and Control of an Urban Air Mobility Vehicle Subject to Empirically-Developed Urban Airflow Disturbances

Urban air mobility is expected to play a role in improving transportation of people and goods in growing urban areas while contributing to sustainable urban growth and zero-emissions future aviation. The research presented herein computationally investigated the performance of control laws for a generic Urban Air Taxi (UAT) subjected to empirically-developed urban airflow disturbances. This involved developing a representative flight dynamics model of a UAT in steady level cruise flight with an inner-loop autopilot. Active Disturbance Rejection Control (ADRC) and Proportional-Integral-Derivative (PID) control laws were implemented to investigate the controlled and uncontrolled acceleration responses and compare them to the acceleration limits in ISO 2631. Using a linear flight dynamics model, ADRC demonstrated improved performance over PID control with equal initial tuning effort. PID was able to reduce passenger accelerations to unharmful, though still uncomfortable, levels while ADRC further reduced the lateral accelerations to comfortable levels.

Robust Space Launcher Control with Time-Varying Objectives

This paper presents a novel approach to robust linear time-varying control design for space launchers. It allows the controller to explicitly account for the launcher’s time-varying dynamics and changing control objectives along the ascent trajectory. The latter are readily incorporated via time-varying weighting functions into a mixed sensitivity design. For a traceable and transparent control design, a physically motivated weighting scheme is applied, which requires little tuning effort. The controller is obtained via a novel observer-based finite horizon linear time-varying synthesis approach. It provides a transparent structure which is easy to implement. The approach is used to design a pitch controller for a flexible expendable launch vehicle in atmospheric ascent tracking the open-loop guidance reference signal.

An audio‐based risky flight detection framework for quadrotors

AbstractDrones have increasingly collaborated with human workers in some workspaces, such as warehouses. The failure of a drone flight may bring potential risks to human beings' life safety during some aerial tasks. One of the most common flight failures is triggered by damaged propellers. To quickly detect physical damage to propellers, recognise risky flights, and provide early warnings to surrounding human workers, a new and comprehensive fault diagnosis framework is presented that uses only the audio caused by propeller rotation without accessing any flight data. The diagnosis framework includes three components: leverage convolutional neural networks, transfer learning, and Bayesian optimisation. Particularly, the audio signal from an actual flight is collected and transferred into time–frequency spectrograms. First, a convolutional neural network‐based diagnosis model that utilises these spectrograms is developed to identify whether there is any broken propeller involved in a specific drone flight. Additionally, the authors employ Monte Carlo dropout sampling to obtain the inconsistency of diagnostic results and compute the mean probability score vector's entropy (uncertainty) as another factor to diagnose the drone flight. Next, to reduce data dependence on different drone types, the convolutional neural network‐based diagnosis model is further augmented by transfer learning. That is, the knowledge of a well‐trained diagnosis model is refined by using a small set of data from a different drone. The modified diagnosis model has the ability to detect the broken propeller of the second drone. Thirdly, to reduce the hyperparameters' tuning efforts and reinforce the robustness of the network, Bayesian optimisation takes advantage of the observed diagnosis model performances to construct a Gaussian process model that allows the acquisition function to choose the optimal network hyperparameters. The proposed diagnosis framework is validated via real experimental flight tests and has a reasonably high diagnosis accuracy.

Real-life data-driven model predictive control for building energy systems comparing different machine learning models

By considering forecasts and exploiting storage effects, model predictive control can achieve significant energy and cost savings in the building sector. However, due to the high individual modeling effort, model predictive control lacks practical applicability. For that reason, data-driven process models, approximating the system behavior based on measurements, have become increasingly popular in recent years. Still, scientific literature lacks consent about the most promising model types and efficient workflows to integrate different machine learning models into a model predictive controller. With this work, we present a workflow to provide efficient model predictive controllers based on measurement data automatically. The main idea is to translate different machine learning models into optimization syntax to enable efficient optimization with full access to gradients. We currently consider artificial neural networks, gaussian process regression, and simple linear regression process models. We use a generic model ontology to automatize the controller generation further and test the methodology on two real-life use cases. The first use case is the application of five office rooms with smart thermostat valves. The second use case is a test hall with an air handling unit and a concrete core activation. Using only two days of initial training data, we deploy controllers based on the different model types for six weeks in the offices and apply online learning to improve the models continuously. We observe only minor differences in controller performance despite the artificial neural networks showing the highest prediction accuracy. The second use case shows that the simple linear models require less controller tuning effort. Thus, for practical applications, we recommend linear regression models.

Simulation of Hadronic Interactions with Deep Generative Models

Accurate simulation of detector responses to hadrons is paramount for all physics programs at the Large Hadron Collider (LHC). Central to this simulation is the modeling of hadronic interactions. Unfortunately, the absence of first-principle theoretical guidance has made this a formidable challenge. The state-of-the-art simulation tool, Geant4, currently relies on phenomenology-inspired parametric models. Each model is designed to simulate hadronic interactions within specific energy ranges and for particular types of hadrons. Despite dedicated tuning efforts, these models sometimes fail to describe the data in certain physics processes accurately. Furthermore, finetuning these models with new measurements is laborious. Our research endeavors to leverage generative models to simulate hadronic interactions. While our ultimate goal is to train a generative model using experimental data, we have taken a crucial step by training conditional normalizing flow models with Geant4 simulation data. Our work marks a significant stride toward developing a fully differentiable and data-driven model for hadronic interactions in High Energy and Nuclear Physics.

Cotton Gin Stand Machine-Vision Inspection and Removal System for Plastic Contamination: Hand Intrusion Sensor Design

Plastic contamination in cotton lint poses significant challenges to the U.S. cotton industry, with plastic wrap from John Deere round module harvesters being a primary contaminant. Despite efforts to manually remove this plastic during module unwrapping, some inevitably enters the cotton gin’s processing system. To address this, a machine-vision detection and removal system has been developed. This system uses inexpensive color cameras to identify plastic on the gin stand feeder apron, triggering a mechanism that expels the plastic from the cotton stream. However, the system, composed of 30–50 Linux-based ARM computers, requires substantial effort for calibration and tuning and presents a technological barrier for typical cotton gin workers. This research aims to transition the system to a more user-friendly, plug-and-play model by implementing an auto-calibration function. The proposed function dynamically tracks cotton colors while excluding plastic images that could hinder performance. A critical component of this auto-calibration algorithm is the hand intrusion detector, or “HID”, which is discussed in this paper. In the normal operation of a cotton gin, the gin personnel periodically have to clear the machine, which entails running a stick or their arm/hand under the detection cameras. This results in the system capturing a false positive, which interferes with the ability of auto-calibration algorithms to function correctly. Hence, there is a critical need for an HID to remove these false positives from the record. The anticipated benefits of the auto-calibration function include reduced setup and maintenance overhead, less reliance on skilled personnel, and enhanced adoption of the plastic removal system within the cotton ginning industry.

A model reference adaptive control framework for floating offshore wind turbines with collective and individual blade pitch strategy

The utilization of floating offshore wind turbines is regarded as a promising solution for offshore renewables, wherein blade pitch control plays a crucial role in both operation and maintenance. Among the various blade pitch control methods, strategies that involve more states pertaining to the control objectives demonstrate superior tracking performance. However, these approaches necessitate additional linearization and parameter tuning efforts, particularly when addressing varying operating conditions in real-world unsteady environments. In this study, a decoupled model reference adaptive control framework with inner control laws for ideal responses for both collective and individual blade pitch is employed, in the context of floating offshore wind turbine control problems during operation beyond rated conditions. According to the simulations conducted on a benchmark wind turbine simulator under a series of turbulent wind speeds, the proposed integrated controller achieved reductions in power fluctuation and damage equivalent load in blades compared to corresponding baseline controllers, at the expense of acceptable deterioration in tower base loads and blade actuator activities. The proposed scheme demonstrates its adaptability in serving the entire operating range, mitigating the need for parameter tuning due to variations in wind speed, and exhibits potential to involve more control purposes.

Segmentation-based closed-loop layer height control for enhancing stability and dimensional accuracy in wire-based laser metal deposition

Laser metal deposition (LMD) with wire is a versatile additive manufacturing process used for the production of near-net-shape metal components as well as for modification and repair applications. Major advantages of using wire as the feedstock material as opposed to powder include the elimination of hazardous metal dust in the process environment and the lower costs. However, the process is highly sensitive to disturbances and requires a significant effort for parameter tuning. Thus, in order to achieve a stable process as well as defined geometric properties over many layers, dedicated approaches for monitoring and control are essential. In particular, maintaining a constant distance between the workpiece surface and the deposition head is an important prerequisite for process stability. Therefore, in this work, a layer height control system for wire-based LMD was implemented. The objective was to ensure a constant layer height corresponding to the specified height increment even in case of disturbances. Using a laser line scanner, the height profile of the part was obtained after each deposited layer. The weld beads were then divided into small segments to obtain a fully discrete height profile, and the wire feed rate for the next layer was set by an individual controller within each segment. The implemented control system was tested for its effectiveness under different disturbances, whereby significant height differences could be fully compensated within few layers. The developed segmentation approach was found to be an effective method to ensure the dimensional accuracy of LMD components. This work thus constitutes a major contribution to the advancement of fully automated additive manufacturing of metallic components.

Direct Predictive Voltage Control for Grid-Connected Permanent Magnet Synchronous Generator System

In this paper, a direct predictive voltage control (DPVC) is proposed for the grid-connected permanent magnet synchronous generator (PMSG) system. In the proposed strategy, the dc-link voltage regulation and power regulation terms are merged into one cost function, thereby eliminating the conventional cascade control structure. The strategy selects the voltage vector that not only will generate a dc-link voltage closer to the setpoint at instant <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$k+2$</tex-math></inline-formula> , but will generate an active power that can further reduce that voltage error for the future instant. To cope with the mismatch between the actual and nominal parameters, Kalman filter has been added to compensate for the steady-state error. Besides, to enhance the dynamic performance and to save the effort of parameter tuning, the weighting factor is further eliminated from the cost function by sorting the terms into two groups and independently evaluating each group in turn. To fully validate the effectiveness of the proposed strategy, the experimental test waveforms of the DPVC strategy with and without weighting factor have been presented and compared with that of the conventional PI-MPC strategy in various testing conditions.

Enhanced Position Estimation Based on Frequency Adaptive Generalized Comb Filter for Interior Permanent Magnet Synchronous Motor Drives

Accurate rotor position estimation is critical to ensure the high-performance operation of position sensorless interior permanent magnet synchronous motor (IPMSM) drives. However, the back electromotive force (BEMF) estimated by the model-based observers is usually nonideal, which may further degrade the position estimation accuracy. In this paper, a frequency adaptive generalized comb filter (FAGCF) with a quadrature phase-locked loop (PLL) is adopted to suppress the rotor position estimation errors by filtering out the dominant distortions in the BEMF, including the (6k±1)th harmonics caused by inverter nonlinearity and flux spatial harmonics, as well as the DC offset from the inaccurate current measurements. This proposed filter is constructed by multiple digital delay blocks and therefore has the advantage of easy implementation with less online parameters tuning effort. Meanwhile, a linearized model is derived for PLL parameter design, with which the stability of the FAGCF-PLL position estimation system can also be analyzed theoretically. The effectiveness of the proposed sensorless control strategy has been validated with the experimental results on a 1.5-kW IPMSM drive.

Enhanced two consecutive samples based de-modulation technique for atomic force microscopy application

This article investigates robust amplitude detectors suitable for atomic force microscopy (AFM) while discussing better alternatives. An AFM instrument’s measurement unit is responsible for providing the amplitude information obtained from the tip of a cantilever beam to identify the surface smoothness of a test material. Therefore, two efficient approaches are suggested to leverage Lyapunov’s theory while adhering to better noise suppression and DC-offset rejection capabilities. Nevertheless, an enhanced two samples-based Lyapunov’s demodulation approach is proposed to detect the amplitude information rapidly. Consequently, the modifications applied to the conventional method help reduce the tuning efforts and structural complications. The proposed solution remains structurally simpler and useful for high- and low-frequency probes. Furthermore, the extensive design guidelines for all techniques and the simulation results are presented. Different amplitude signals are synthetically generated from several rough pseudo-test surfaces for early verification and sent to a real-time digital controller to judge the proposal’s efficacy.

Finite Set Sensorless Control With Minimum a Priori Knowledge and Tuning Effort for Interior Permanent-Magnet Synchronous Motors

Finite Set Sensorless Control With Minimum a Priori Knowledge and Tuning Effort for Interior Permanent-Magnet Synchronous Motors

ALADA: A lite automatic data augmentation framework for industrial defect detection

Industrial defect detection is a critical and challenging task in the quality control of manufacturing production. Competent in feature extraction and pattern recognition, deep learning shows great power for classifying and locating defects in industrial products. However, insufficient data records and diverse categories of defects restrict the detection accuracy of data-driven neural networks. One direction to solve such a problem is data augmentation, which aims at generating synthetic copies from existing data and improving the generalizability of detection models. However, confirming a suitable augmentation policy involves either human experience or substantial experiments for augmentation parameters. To address these challenges, in this work, a lite automatic data augmentation (ALADA) framework is proposed to jointly optimize the data augmentation policies and the neural network for industrial defect detection. First, a lite search space is formulated to efficiently sample augmentation policies and generate augmented images for joint optimization. To reduce the hyperparameter tuning efforts for retraining with searched policies, a three-step bi-level optimization scheme is proposed to replace the retraining strategy and update the model and augmentation parameter alternately. To solve the non-differentiable problem in the joint optimization scheme, policy gradient sampling is implemented to estimate the gradient flow efficiently. Experimental results on three industrial defect detection datasets, namely, Tianchi-TILE, GC10-DET, and NEU-DET, reveal that our proposed automatic augmentation framework outperforms the state-of-the-art augmentation methods and effectively improves the accuracy of the baseline defect detection model. The proposed ALADA scheme also alleviates the missed detection of defects in four practical industrial circumstances: textured background, uneven brightness, low contrast, and intraclass difference.