Direct Prediction Method Research Articles

Data-Driven Remaining Useful Life Estimation

This paper explores the development and application of data-driven prognostic models for estimating the Remaining Useful Life (RUL) of Nuclear Power Plant (NPP) condensers experiencing tube fouling. Due to the unavailability of run-to-failure industry sensor data, we utilized simulated data generated by the Asherah Nuclear Power Plant Simulator (ANS), initially designed by the International Atomic Energy Agency (IAEA) and programmed in Simulink for cyber security simulations. ANS's adaptability allows it to simulate Pressurized Water Reactor (PWR) behaviors given a time series of operating conditions and to introduce degradation modules to mimic fouling effects. Our study compares two primary approaches applied to data generated by ANS: inference-based and direct prediction methods. The selected inference-based approach estimates the health state of the condenser using a pipeline formed by an Auto Associative Kernel Regressor and a Hidden Markov Model (HMM), which subsequently combines the state estimates with its parameters to predict the RUL. The direct prediction method employs a Gradient Boosting Regressor Decision Tree (GBRDT) to map input variables directly to RUL. Our findings demonstrate the efficacy and limitations of each method through the case study, providing valuable insights for the adoption of data-driven RUL estimation techniques in industrial and energy applications.

Long-term temporal attention neural network with adaptive stage division for remaining useful life prediction of rolling bearings

Accurate rolling bearing remaining useful life (RUL) prediction, an effective assurance of the rotating machinery's safety and reliability, is one of the essential procedures in equipment maintenance. Current RUL prediction methods mostly adopt direct prediction methods, but it is difficult for them to guarantee prediction accuracy under the influence of longlife cycles and variable production environments. Therefore, a long-term temporal attention neural network with adaptive stage division (AD-LTAN) is proposed to predict the RUL of rolling bearings. Aiming at the large fluctuation range of the degradation starting point, an adaptive stage division model with the augmentation of early features is proposed to analyze the long-sequence signals, and then the life cycle of the bearings will be divided into different health stages. Aiming at the network memory decline under the longlife cycle of degradation, a long-term temporal attention neural network is designed to retain the long-term degradation characteristics of bearings by leveraging multilevel expansion convolution and integrating attention mechanisms to extract the fault signal features to realize the RUL prediction. The experimental results conducted on the PHM2012 and XJTU-SY datasets demonstrate that the proposed method outperforms the compared methods in terms of prediction loss (35.4 % less than their best).

Direct prediction for oceanic mesoscale eddy geospatial distribution through prior statistical deep learning

Mesoscale Eddy (ME) is a widely recognized and significant oceanic phenomenon characterized by extensive energy exchange. Accurate predictions of the future geospatial distribution of ME are crucial for maritime activities. Deep learning (DL) methods for ME prediction have consistently outperformed classical mathematical models and numerical modeling approaches. However, current DL methods based on two-dimensional gridded data suffer from error accumulation issues due to their indirect prediction approach, which involves first predicting the future environmental field and then identifying the ME patterns within it. Additionally, these models ignore the potential influence of prior knowledge related to ME on prediction outcomes. To address these issues, this paper proposes a direct prediction DL network called ES-ConvGRU, which incorporates historical prior statistics. First, we conduct an analysis of ME data from the Northwest Pacific region, extracting prior statistical knowledge such as interannual variations and seasonal characteristics. Second, mining patterns for the spatiotemporal features of both ME sequence and sea surface environmental field sequence are established. Furthermore, a nonlinear approach for mapping prior statistical knowledge and a multi-step autoregressive prediction model are designed, facilitating the integration of prior knowledge with multidimensional spatiotemporal features for ME prediction. Finally, the performance and effectiveness of ES-ConvGRU are evaluated. It achieves an average pixel accuracy of 93.21%, a mean intersection over union of 81.48%, and a frequency-weighted intersection over union of 87.55% for the 7-day distribution prediction. The results show that ES-ConvGRU exhibits notable advantages compared to the existing indirect and direct prediction methods.

A direct prediction method for 3D woven composites bending properties based on unit-cell finite element model

The deformation distribution along the thickness direction of 3D woven composites under bending loading is inhomogeneous, and hence there is difficulty in numerically predicting the bending behavior of woven composites. This paper aims to propose a direct mesoscopic method for predicting the bending behavior of woven composites using the unit-cell finite element model. Firstly, four-point bending tests were conducted in the warp and weft directions, respectively. Then, the periodic boundary conditions are established for the unit-cell model with non-uniformly distributed deformations under pure bending loading. Finally, the bending properties and damage accumulation process of the woven composites are analyzed based on the elasto-plastic damage model with different mechanical properties in tension and compression. The simulated moment-curvature curves and failure modes are in good agreement with the experimental results. It is shown that the developed unit-cell finite element model can accurately predict the bending behavior of woven composites.

Enhancing Quality Control in Battery Component Manufacturing: Deep Learning-Based Approaches for Defect Detection on Microfasteners

The management of product quality is a crucial process in factory manufacturing. However, this approach still has some limitations, e.g., depending on the expertise of the engineer in evaluating products and being time consuming. Various approaches using deep learning in automatic defect detection and classification during production have been introduced to overcome these limitations. In this paper, we study applying different deep learning approaches and computer vision methods to detect scratches on the surface of microfasteners used in rechargeable batteries. Furthermore, we introduce an architecture with statistical quality control (SQC) to continuously improve the efficiency and accuracy of the product quality. The proposed architecture takes advantage of the capability of deep learning approaches, computer vision techniques, and SQC to automate the defect detection process and quality improvement. The proposed approach was evaluated using a real dataset comprising 1150 microfastener surface images obtained from a factory in Korea. In the study, we compared the direct and indirect prediction methods for predicting the scratches on the surface of the microfasteners and achieved the best accuracy of 0.91 with the indirect prediction approach. Notably, the indirect prediction method was more efficient than the traditional one. Furthermore, using control charts in SQC to analyze predicted defects in the production process helped operators understand the efficiency of the production line and make appropriate decisions in the manufacturing process, hence improving product quality management.

Engine remaining useful life prediction based on PSO optimized multi-layer long short-term memory and multi-source information fusion

Engine as the core component of mechanical equipment, its operating state directly affects whether the equipment can operate normally. Predicting the engine remaining useful life (RUL) can monitor the health of the engine in real time and formulate a timely and reasonable maintenance plan. Aiming at the engine monitoring data with various and long time span, we propose a direct prediction method of engine RUL based on particle swarm optimization (PSO) optimized multi-layer Long Short-Term Memory (LSTM) in this paper. Firstly, the monitoring data that can well reflect the engine degradation trend is screened out, and the samples are constructed through a sliding time window. Then, a multi-layer LSTM model is constructed to mine the deep-seated features of the samples for predicting the engine RUL. Finally, the hyperparameters of the multi-layer LSTM model are optimized automatically by the PSO algorithm to optimize the performance of the model. The effectiveness of this method is verified by NASA data set. RMSE, MAE and the scoring function are used as evaluation indexes. RMSE and score of the prediction results are 12.35 and 284.1, respectively. It has higher prediction accuracy compared with traditional deep learning and machine learning methods.

Considering dynamic perception of fluctuation trend for long-foresight-term wind power prediction

With the increasing proportion of renewable energy being integrated into the power system, the stable operation of the power system is facing great challenges. Breaking through the limitation and prolonging the time period of power prediction will help to improve the consumption ability of wind power (WP), adjust the power generation plan, and ensure the safe operation of the power system. Based on the availability of long-foresight-term numerical weather prediction (NWP) and fluctuation trend mining, a wind farm cluster power forecasting method based on dynamic perception of fluctuation trend is proposed for the long-foresight-term wind power prediction (WPP). The method first matches historical power based on the fluctuation trend of NWP wind speed (WS), and then obtains statistical characteristics of similar processes in order to construct new input features. Secondly, after the prediction results are obtained, the predicted power is corrected according to the fluctuation trend to improve the prediction reliability. This method is applied to a wind farm cluster in Gansu Province, China, and the overall prediction accuracy is 2.23 % higher than that of the direct prediction method. It is verified that this method can achieve long-foresight-term prediction and increase the reliability of prediction.

Nonlinear aging knee-point prediction for lithium-ion batteries faced with different application scenarios

The capacity degradation of lithium-ion batteries (LIBs) will accelerate after long-term cycling, showing nonlinear aging features, which not only shortens the long-term life of LIBs, but also seriously endangers their safety. In this paper, by introducing the concept of nonlinear aging degree, a knee-point identification method based on the maximum distance method is established, and the nonlinear aging behavior of LIBs is identified and marked, so as to know whether the nonlinear aging phenomenon has occurred. Furthermore, two knee-point prediction methods have been proposed and compared. The direct knee-point prediction method based on stacked long short-term memory (S-LSTM) neural network and sliding window method is proposed for the scenarios of battery development, early performance evaluation and online application. For scenarios such as echelon utilization and post-safety evaluation, an indirect knee-point prediction method combining capacity prediction and knee-point identification algorithm is proposed. Through multi-dimensional comparison of the two methods, the strengths and weaknesses of their applicable scenarios are analyzed. Our work has guiding significance for finding the ideal replacement opportunity of LIBs in different scenarios, so that the user can be reminded whether to maintain or replace the battery, which greatly reduces the risk of battery safety problems.

Direct distortion prediction method for AR-HUD dynamic distortion correction.

Dynamic distortion is one of the most critical factors affecting the experience of automotive augmented reality head-up displays (AR-HUDs). A wide range of views and the extensive display area result in extraordinarily complex distortions. Existing methods based on the neural network first obtain distorted images and then get the predistorted data for training mostly. This paper proposes a distortion prediction framework based on the neural network. It directly trains the network with the distorted data, realizing dynamic adaptation for AR-HUD distortion correction and avoiding errors in coordinate interpolation. Additionally, we predict the distortion offsets instead of the distortion coordinates and present a field of view (FOV)-weighted loss function based on the spatial-variance characteristic to further improve the prediction accuracy of distortion. Experiments show that our methods improve the prediction accuracy of AR-HUD dynamic distortion without increasing the network complexity or data processing overhead.

Fuzzy logic-based adaptive tracking weight-tuned direct performance predictive control method of aero-engine

Direct performance predictive control is the most cutting-edge research in aero-engine control. The existing subspace-based improved model predictive control (SIMPC) effectively handled the problem of difficulty solving the prediction model in predictive control. However, the strong nonlinearity of aero-engine makes the controller with design parameters set constant invariant value unable to guarantee consistent control effect in all states. Even with general scheduling for design parameters, the controller's performance still behaves poorly. Therefore, this paper innovatively proposes an adaptive tuning method for the design parameters of the SIMPC controller to adapt to all engine states. Firstly, by analyzing the influence of each design parameter of the controller on the control effect, the tracking weight with strong sensitivity is selected as the dynamic tuning parameter. Then, three variables of the ratio of PLA angle, relative tracking error, and scale factor are successfully introduced to map the tracking weight value without adding additional design parameters. Thus, a fuzzy logic tuner based on these variables is then designed for adaptively tuning the tracking weight to maintain the controller's performance. Finally, results show that, compared with the existing methods, while keeping the maximum dynamic response performance, the proposed method eliminates the phenomenon of large overshoot and controls it within 2% required in the whole state range. Compared with the general scheduling method, the designed adaptive tracking weight with positive excitation caused by relative tracking error effectively shortens the response time, especially with an average performance improvement of 17% under small step changes. Whether the engine is in standard or fault conditions, the designed adaptive tracking weight is effective and feasible for undertaking the controller's consistent control effects in all states.

Evaluation of the Storms Direct Runoff Prediction Methods used for Goizha-Dabashan Watershed

The Momentum and Aron & White evaluating methods have been adopted to estimate the Nash Instantaneous Hydrograph parameters (IUH), while the two methods of excess rainfall (Ф-index and Natural Resources Conservation Service (NRCS) were applied in a model using a developed computer program in MATLAB to predict the direct runoff hydrograph for Goizha-Dabashan watershed located in the northeast of Iraq. In the verification stage, both Nash IUH optimal parameters of the storms and the average optimal values of the same parameters estimated in the calibration stage were applied and compared. The statistical tests showed a preference for the NRCS method with the momentum method in estimating direct runoff hydrograph (the average Nash-Sutcliffe Efficiency (NSE) was equal to 0.815 and 0.77 using optimal parameters verification storms and the average calibrated IUH parameters values, respectively). Also, satisfactory results (NSE was equal to 0.77 and 0.76 using storm parameters and the average calibrated IUH parameters values, respectively) were obtained by applying Aron & White with the NRCS methods, which indicated the ability of both methods for estimating direct runoff hydrograph.

A direct prediction method for wind power ramp events considering the class imbalanced problem

AbstractPredicting wind power ramp events directly based on the historical ramp event time series has drawn increasing attention recently. But the class imbalance problem of the ramp event time series significantly affects the prediction accuracy of ramp events. In the present study, a layer oversampling (LOS) method is proposed considering the relation characteristics of wind power amplitudes and the occurrence frequency of wind power ramp events. Meanwhile, a hybrid sampling method of error bootstrap‐LOS (EB‐LOS) is proposed by combining LOS with the EB oversampling method. After balancing the samples of the ramp and nonramp events by using different sampling methods, the backpropagation neural network (BPNN), and the long short‐term memory (LSTM) methods are employed to directly predict ramp events based on historical data collected from eight wind farms. Comparison results proved that the proposed EB‐LOS method achieves the best prediction performance with an average recall of 0.8196 when using the BPNN model to directly predict ramp events. The best prediction performance of the EB‐LOS method is also proved by using the LSTM model to directly predict ramp events.

Improved Model Predictive Direct Power Control for Parallel Distributed Generation in Grid-Tied Microgrids

This research proposes an improved finite control set direct power model predictive control method (FCS-DPMPC) for grid-tie distributed generation (DG). FCS-DPMPC predicts the system outcomes using the system model. During the next sampling time, a voltage vector is defined using the cost function to minimize the power ripple, consequently allowing flexibility for power regulation. Furthermore, the impact of implementing a one-step delay is studied and compensated through a model forecast pattern. In addition, a new two-step horizon technique has been developed to minimize switching frequency and computation burden. Simulation results for single DG and parallel operated DGs in a grid-tie manner confirm the effectiveness of the suggested control strategy, which signifies that this is an appropriate approach for distributed generation in microgrids.

A numerical investigation on direct and data-driven flutter prediction methods

Three alternative algorithms for dynamic aeroelastic stability analysis are investigated. They are based either on direct search of oscillatory conditions of the coupled system or on eigenvalue analysis from frequency-domain sampling of the unsteady aerodynamics. The aerodynamic forcing is obtained with a harmonic-balanced solver developed for a general-purpose finite-volume fluid dynamics solver. This new implementation is first verified against experimental flutter results from the literature. A wing with relatively low bending stiffness is then used to explore the relative performance of each approach, both in terms of the numerical robustness and of their usability to support aeroelastic design.

FoVolNet: Fast Volume Rendering using Foveated Deep Neural Networks.

Volume data is found in many important scientific and engineering applications. Rendering this data for visualization at high quality and interactive rates for demanding applications such as virtual reality is still not easily achievable even using professional-grade hardware. We introduce FoVolNet-a method to significantly increase the performance of volume data visualization. We develop a cost-effective foveated rendering pipeline that sparsely samples a volume around a focal point and reconstructs the full-frame using a deep neural network. Foveated rendering is a technique that prioritizes rendering computations around the user's focal point. This approach leverages properties of the human visual system, thereby saving computational resources when rendering data in the periphery of the user's field of vision. Our reconstruction network combines direct and kernel prediction methods to produce fast, stable, and perceptually convincing output. With a slim design and the use of quantization, our method outperforms state-of-the-art neural reconstruction techniques in both end-to-end frame times and visual quality. We conduct extensive evaluations of the system's rendering performance, inference speed, and perceptual properties, and we provide comparisons to competing neural image reconstruction techniques. Our test results show that FoVolNet consistently achieves significant time saving over conventional rendering while preserving perceptual quality.

A Novel Predictive Position Control With Current and Speed Limits for PMSM Drives Based on Weighting Factors Elimination

Conventional predictive position control (PPC) method balances the position, speed and current of permanent magnet synchronous machine with different weighting factors. However, tedious tuning work of weighting factors is needed. To simplify the tuning process, a cost function that includes the system position, and current information is constructed based on deadbeat control method. A novel method based voltage boundary is proposed to limit the current and speed simultaneously, the voltage regions corresponding to current and speed constraints are analyzed and the voltage duty cycle is further modified according to the voltage regions, the complex weighting factor turning work is avoided, the current and speed limits are easy to implement with proposed method and less position error can be achieved. Finally, the proposed method is experimentally compared with a conventional finite-control-set PPC method and a direct speed model predictive control method.

Improving active canopy sensor-based in-season rice nitrogen status diagnosis and recommendation using multi-source data fusion with machine learning

Accurate and non-destructive diagnosis of crop nitrogen (N) surplus and deficit status based on N nutrition index (NNI) is crucially important for the success of precision N management to improve N use efficiency (NUE) and reduce negative environmental impacts. However, due to the variability of the reflectance data obtained from different active crop sensors and complexity of the environmental and management conditions for regional applications, accurate determination of crop N status and topdressing N rate only using active canopy sensor data is very challenging. The objectives of this study were to (1) develop an in-season N status diagnosis and recommendation model based on NNI prediction using multi-source data fusion with machine learning, and (2) evaluate the accuracy of N diagnosis and recommendation in terms of rice yield and NUE under diverse on-farm conditions. Thirty plot experiments and thirteen on-farm experiments were conducted in Qixing Farm, Jiansanjiang, Northeast China from 2008 to 2018, and the dataset was used for the model calibration, validation, and evaluation. Two indirect and one direct NNI prediction methods using simple regression, stepwise multiple linear regression (SMLR) and random forest regression (RFR) were compared for N diagnosis and then integrated into N recommendation model. The results indicated that combining environmental and agronomic variables with crop sensor data improved the SMLR and RFR model performance by 1–16% and 9–40% over the corresponding models only using crop sensor data, respectively. The direct NNI prediction approach achieved slightly better N status diagnostic accuracy (areal agreement = 84% and Kappa statistics = 0.71) than indirect NNI prediction strategies based on plant N uptake and ΔN estimation (areal agreement = 81% and Kappa statistics = 0.67) or aboveground biomass and plant N uptake estimation (areal agreement = 77% and Kappa statistics = 0.58) across plot experiments and diverse on-farm conditions, based on multi-source data fusion with random forest regression models. About 82% of recommended N rates by the developed integrated in-season rice N diagnosis and recommendation model were within ±10 kg ha−1 of the measured economic optimum N rate across different varieties, environmental conditions and transplanting densities. Precision rice management based on in-season N diagnosis and recommendation decreased N rates and increased N partial factor productivity (PFPN) by 23% over regional optimum rice management, and significantly increased yield (7–11%) and PFPN (33–77%) over farmer's management. More studies are needed to develop in-season N diagnosis and recommendation strategies for applications across different regions and combine them with integrated precision rice management strategies for food security and sustainable development.

Direct Prediction Method for Semi-Rigid Behavior of K-Joint in Transmission Towers Based on Surrogate Model

The assembled tube-gusset K-joint by bolts is a commonly used connection form in steel tubular transmission towers. At present, main existing research or design codes for steel tubular transmission towers regard this K-joint as either rigid or pinned connections, which do not consider the semi-rigid behavior of K-joint. In this paper, the semi-rigid behavior of K-joint in steel tubular transmission towers is investigated and a direct prediction (DP) method is proposed to evaluate the semi-rigid behavior of K-joints based on the support vector regression (SVR) model, especially to predict the moment–rotation curve of semi-rigid K-joints. First, the establishment and validation of the finite element (FE) model of semi-rigid K-joints are conducted. Second, a dataset of 144 samples generated by the FE model is used to train and test the SVR model. Finally, the accuracy assessment of the proposed DP method and comparison with other existing methods, including the Kishi–Chen model, EC3 model and ANN-based two-step prediction method, are presented. The accuracy assessment shows that predicted values of the proposed DP method based on the SVR model exhibit good agreement with the numerical analysis values, which indicates the quite high accuracy of this method. Additionally, the comparison reveals that the proposed DP method based on the SVR model for predicting moment–rotation curves is rather more accurate than other aforementioned methods. Therefore, the proposed DP method based on the SVR model is of high reliability in predicting the semi-rigid behavior of K-joints in steel tubular transmission towers, which affords an alternative way for further engineering analysis and initial design purposes.

A novel direct performance adaptive control of aero-engine using subspace-based improved model predictive control

The direct performance predictive control method can fully improve the engine's operating performance. Compared with the traditional control methods, it has a better control effect. However, it urgently needs to solve the problems of how to achieve the predictive control applicable for any operating point of the engine and how to further reduce the calculation cost of performance parameter estimation. Therefore, this paper first proposes a subspace-based improved model predictive control method. The subspace identification method is used to identify the past I/O data set of the engine offline to obtain the approximate predictive model, and then the predictive model is updated in real-time according to the currently collected I/O data online to adapt to the current actual operating point of the engine. Thus, the engine's direct performance adaptive control is successfully realized based on the predictive control using this obtained predictive model in full life cycle, full envelope, and full state in this paper. This method reduces the online computational burden through the Givens rotation transformation algorithm, and the design process is simple. Compared with the existing methods, it not only improves the robustness, adaptability, and independence of the engine model but also perfects the control performance. Secondly, aiming at the problems, such as complex design and a tremendous amount of calculation of the current direct performance estimator, a high-precision real-time estimation method of nonlinear onboard performance parameters based on spherical unscented Kalman filter is also proposed. Compared with the existing estimation methods, its computational complexity is reduced by one time while maintaining accuracy. Finally, a fully digital simulation system of aero-engine's direct performance adaptive predictive control is established. The simulation results show that the dynamic response time from idle to maximum state is 3.48 s, the steady-state error is ±0.2%, and there is no overshoot. Compared with the traditional control, the dynamic performance is improved by 32%.

Uncertainty quantification techniques for data-driven space weather modeling: thermospheric density application

Machine learning (ML) has been applied to space weather problems with increasing frequency in recent years, driven by an influx of in-situ measurements and a desire to improve modeling and forecasting capabilities throughout the field. Space weather originates from solar perturbations and is comprised of the resulting complex variations they cause within the numerous systems between the Sun and Earth. These systems are often tightly coupled and not well understood. This creates a need for skillful models with knowledge about the confidence of their predictions. One example of such a dynamical system highly impacted by space weather is the thermosphere, the neutral region of Earth’s upper atmosphere. Our inability to forecast it has severe repercussions in the context of satellite drag and computation of probability of collision between two space objects in low Earth orbit (LEO) for decision making in space operations. Even with (assumed) perfect forecast of model drivers, our incomplete knowledge of the system results in often inaccurate thermospheric neutral mass density predictions. Continuing efforts are being made to improve model accuracy, but density models rarely provide estimates of confidence in predictions. In this work, we propose two techniques to develop nonlinear ML regression models to predict thermospheric density while providing robust and reliable uncertainty estimates: Monte Carlo (MC) dropout and direct prediction of the probability distribution, both using the negative logarithm of predictive density (NLPD) loss function. We show the performance capabilities for models trained on both local and global datasets. We show that the NLPD loss provides similar results for both techniques but the direct probability distribution prediction method has a much lower computational cost. For the global model regressed on the Space Environment Technologies High Accuracy Satellite Drag Model (HASDM) density database, we achieve errors of approximately 11% on independent test data with well-calibrated uncertainty estimates. Using an in-situ CHAllenging Minisatellite Payload (CHAMP) density dataset, models developed using both techniques provide test error on the order of 13%. The CHAMP models—on validation and test data—are within 2% of perfect calibration for the twenty prediction intervals tested. We show that this model can also be used to obtain global density predictions with uncertainties at a given epoch.