Traditional Modeling Methods Research Articles

A two-phase soft sensor modeling framework for quality prediction in industrial processes with missing data

Quality prediction plays an essential role in modern process industries to improve product quality, ensure production stability, and increase economic efficiency. The high-dimensional, nonlinear, and dynamic characteristics of process variables make it hard for the traditional soft sensor modeling methods. Meanwhile, high-quality data is particularly scarce due to the economic and technological constraints, making it easy to overfit and crash of the model training. In order to solve those problems, a two-phase soft sensor modeling framework based on time series generative adversarial network (TimeGAN) and minimal gated unit (MGU) is proposed. The first phase is performing data augmentation for small samples by TimeGAN to improve the generalization performance of model training, and the second phase uses MGU for soft sensor modeling with the purpose of quality prediction. Finally, two datasets with different thicknesses for the hot rolling process (HRP) are used to verify the accuracy and validity of the proposed scheme.

Tensorial-formulation-based method for modeling of flight dynamics with generalized coordinates

For the sake of minimizing fuel consumption, a rocket typically keeps flying in or near a plane specified by the Earth's center, launch pad, and orbit-insertion point, where the plane is called the Reference Flight Plane (RFP) here. To facilitate trajectory analysis, a series of generalized position and velocity coordinates are defined based on the RFP. To develop the Ordinary Differential Equations (ODEs) of the generalized coordinates, a novel dynamics modeling method based on tensor theory is proposed. As the first step of the new method, the rotating Earth is taken as the reference frame and a flight dynamics model described by orthogonal coordinates is developed. Then, according to the definitions of the generalized coordinates, two special curvilinear coordinate systems are put forward, and the relationships between the curvilinear and orthogonal coordinate systems are found. Finally, skillfully using the relationships as well as tensor theory, the dynamics model for the orthogonal coordinates are transformed into that for the generalized coordinates. The benefit of the new method is that it can derive the ODEs of multiple coordinates simultaneously because it is entirely based on matrix operations. By contrast, the traditional modeling methods from Newtonian mechanics and analytical mechanics need to use the infinitesimal approach to develop the equation of each generalized coordinate one by one, which may result in a large amount of repeated derivation. Therefore, the new method has higher efficiency in developing flight dynamics model described by generalized coordinates, and helps to prevent mistakes in derivation.

Adaptive Terminal Sliding Mode Fault-Tolerant Control of Spacecraft Based on the Left Attitude Error Function of SO(3)

For the problem of spacecraft attitude actuator failure, an adaptive terminal sliding mode fault-tolerant controller (ATSMFTC) based on the differential manifold SO(3) modelling is designed in this paper. First, SO(3) is used to provide a global and unique description of the spacecraft attitude dynamic model. This modelling method not only avoids the problems of singularity and unwinding that exist in traditional modelling methods but also the SO(3) modelling has a simple formulation of the dynamic equations. Then a left attitude error descriptor function is constructed on SO(3) to design an ATSMFTC. This controller is capable of fast and accurate tracking of the time-varying desired attitude. At the same time, it can react quickly to maintain system stability in case of spacecraft attitude actuator failure. The controller designed based on the left attitude error description system of SO(3) has the features of small computational effort and simple design process. Finally, the numerical simulation of the attitude tracking error verifies the feasibility and high efficiency of the controller designed in this paper.

Artificial neural network-based predictive model for supersonic ejector in refrigeration system

In this paper, a novel predictive model is proposed for the supersonic ejector in the refrigeration system based on the artificial neural network technique. A composite performance prediction framework for the supersonic ejector is developed by employing a back propagation neural network with particle swarm optimization (PSO-BPNN) algorithm and a dynamic error compensation method. Firstly, we study the model construction problem for the supersonic ejector based on the traditional BPNN modeling method and the PSO algorithm. And then, a dynamic compensation technique is given to improve the predictive accuracy of the ejector model based on the adaptive neural network method. Finally, both the simulation and experimental results are given to verify our method. The experimental validation results show that most prediction errors of our model are less than 4%. Therefore, compared with the existing traditional models, our model has a higher accuracy.

Agent-Based Modeling for Water–Energy–Food Nexus and Its Application in Ningdong Energy and Chemical Base

Water, Energy and Food (WEF) are coordinated and constrained by each other, constituting a multivariate coupled feed-forward dynamical system. Traditional modeling and simulation methods struggle to model and simulate complex interactions in the WEF nexus. Therefore, we proposed and developed an agent-based model, which is one of the most effective tools for simulating complex systems. It also has unique advantages in simulating WEF allocation, which is very helpful in improving regional WEF use efficiency. By taking Ningdong Energy and Chemical Base as the research area, an agent-based water–energy–food model based on MESA library was developed using Python 3.9 language, which includes six types of agents and can explore and simulate the complex dynamic interactions in the supply and demand process of WEF sectors. Different behavior rules were proposed to quantify the interactions between WEF sectors of Ningdong Energy and Chemical Base. Four different scenarios were set up, namely, the baseline scenario, the water conservation scenario, the new reservoir scenario and the integrated scenario, and the uncertain system evolution processes between departments and resources under the four different scenarios were analyzed in detail to quantitatively analyze the evolution of the water–energy–food complex system of Ningdong Energy and Chemical Base, which has proven the effectiveness of the proposed model. The results show that: water allocation, energy consumption and food consumption in the domestic sector have similar degrees of impact, because the natural population growth rate does not change under different scenarios; water allocation in the food sector shows a trend corresponding to changes in crop yields; water allocation in the energy management sector shows an upward trend, the water allocation in the actual years 2016–2020 is almost the same, and in the forecast years 2021–2025, the baseline scenario and the water conservation scenario can’t meet the demand volume of the energy management sector due to limited water sources, so the total allocated water is lower than that in the increased reservoir and comprehensive scenario; the water allocated to ecological sector has a balanced situation, and the annual growth of the ecological greening coverage area is also balanced; the total water allocation also shows a trend of annual growth; regarding the annual energy volume that can be delivered to the area outside the base, the curve first grows sharply with a growth rate of about 19.85%, and then becomes slowly with a growth rate of about 3.53%. The total volume is expected to increase to 4.96 × 107 tce by 2025; the total energy, consumed energy and output energy, in general, shows a growing trend, and with the development of the economy and technology, the total energy of the base will reach 7.96 × 107 tce by 2025.

Review of Challenges and Opportunities in Turbulence Modeling: A Comparative Analysis of Data-Driven Machine Learning Approaches

Engineering and scientific applications are frequently affected by turbulent phenomena, which are associated with a great deal of uncertainty and complexity. Therefore, proper modeling and simulation studies are required. Traditional modeling methods, however, pose certain difficulties. As computer technology continues to improve, machine learning has proven to be a useful solution to some of these problems. The purpose of this paper is to further promote the development of turbulence modeling using data-driven machine learning; it begins by reviewing the development of turbulence modeling techniques, as well as the development of turbulence modeling for machine learning applications using a time-tracking approach. Afterwards, it examines the application of different algorithms to turbulent flows. In addition, this paper discusses some methods for the assimilation of data. As a result of the review, analysis, and discussion presented in this paper, some limitations in the development process are identified, and related developments are suggested. There are some limitations identified and recommendations made in this paper, as well as development goals, which are useful for the development of this field to some extent. In some respects, this paper may serve as a guide for development.

Research on Seismic Exploration Velocity Modeling Based on Deep Learning

As the core parameter of seismic exploration, seismic wave propagation velocity runs through the seismic exploration process. The construction of reasonable and accurate velocity models is an important basis for exploration technologies such as high-precision seismic offset imaging, data processing and interpretation. The current velocity modeling strategy is limited by the target geological body, which has a long calculation period, high calculation cost, and is seriously affected by human subjective factors, and its adaptability to complex geological formations is poor. Based on this, this paper studies the accurate speed modeling method of universities. Firstly, the principle of earthquake detection is analyzed, and then the speed modeling based on improved fully convolutional neural network is studied. A velocity modeling fully convolutional neural network (VNB-FCN) is constructed to directly extract geological features from seismic data and establish a high-precision velocity model. Compared with the traditional velocity modeling method, the VMB-FCN model effectively avoids the problem of excessive dependence of neural networks on standard seismic datasets, and improves the accuracy and efficiency of seismic exploration velocity modeling.

A flexible and scalable single-level framework for OD matrix inference using IoT data

This study proposes a flexible and scalable single-level framework for origin–destination matrix (ODM) inference using data from IoT (Internet of Things) and other sources. The framework allows the analyst to integrate information from multiple data sources, while controlling for differences in data quality across sources. We assess the effectiveness of the framework through a real-world experiment in Greater Adelaide (GA), Australia. We infer car OD flows within the region using four separate data sources: site-level traffic counts from loop detectors, vehicle trajectories recorded by roadside Bluetooth sensors, partial OD flows based on data from in to vehicle navigation systems, and journey-to-work OD data collected by the Australian Census. We compare our OD inferences with those from the current version of the Strategic Adelaide Model (SAM), calibrated using data from traditional household travel surveys. We find remarkable consistency between our inferences and those from SAM, despite differences in input data and methodologies. For example, for the morning peak period, we predict the total number of trips made within GA to be equal to 556,000, while the corresponding prediction from SAM is equal to 484,000. The two predictions are within 15 per cent of each other. When we compare the spatial distribution of trips, in terms of origins and destinations, we find that our inferred OD matrix has an 86 per cent cosine similarity to the corresponding SAM matrix. Travel time predictions based on our OD inferences correspond closely to those predicted by the Google Directions API, lending further confidence in our inferences. In summary, our results show that the proposed framework can produce highly comparable ODM, trip production and trip attraction patterns to those inferred from traditional household travel survey-based transportation demand modelling methods.

Estimating hydrogeological parameters at groundwater level observation wells without pumping well information

Aquifer pumping is a common method of meeting human water needs, but it can also disrupt the natural flow of groundwater and cause fluctuations in groundwater levels. These fluctuations, however, can provide useful insights into hydrogeological properties when measured in an observation well. Traditional modeling methods are not well suited for analyzing such data because they lack essential details about the pumping wells, such as their number, distance from the observation well, and pumping rates. To address this issue, new impulse response function for transfer function modeling is developed that accurately describes the confined flow in such situations. The proposed function is based on the Theis solution, which uniformly distributes the effects of multiple pumping wells within a circular area around the observation well. This allows our function to accurately estimate hydrogeological parameters and pumping rates without knowing the number and specific location of pumping wells. The effects of well distance, spatial arrangement, and aquifer anisotropy on the accuracy of our function are also analyzed using synthetic experiments and Morris sensitivity analysis. The results show that the proposed function is effective in estimating essential hydrogeological parameters, such as the aquifer storativity, using only drawdown time series data from a observation well.

Unsupervised Noise Reductions for Gravitational Reference Sensors or Accelerometers Based on the Noise2Noise Method.

Onboard electrostatic suspension inertial sensors are important applications for gravity satellites and space gravitational-wave detection missions, and it is important to suppress noise in the measurement signal. Due to the complex coupling between the working space environment and the satellite platform, the process of noise generation is extremely complex, and traditional noise modeling and subtraction methods have certain limitations. With the development of deep learning, applying it to high-precision inertial sensors to improve the signal-to-noise ratio is a practically meaningful task. Since there is a single noise sample and unknown true value in the measured data in orbit, odd-even sub-samplers and periodic sub-samplers are designed to process general signals and periodic signals, and adds reconstruction layers consisting of fully connected layers to the model. Experimental analysis and comparison are conducted based on simulation data, GRACE-FO acceleration data, and Taiji-1 acceleration data. The results show that the deep learning method is superior to traditional data smoothing processing solutions.

Recent Approaches to Design and Analysis of Electrical Impedance Systems for Single Cells Using Machine Learning.

Individual cells have many unique properties that can be quantified to develop a holistic understanding of a population. This can include understanding population characteristics, identifying subpopulations, or elucidating outlier characteristics that may be indicators of disease. Electrical impedance measurements are rapid and label-free for the monitoring of single cells and generate large datasets of many cells at single or multiple frequencies. To increase the accuracy and sensitivity of measurements and define the relationships between impedance and biological features, many electrical measurement systems have incorporated machine learning (ML) paradigms for control and analysis. Considering the difficulty capturing complex relationships using traditional modelling and statistical methods due to population heterogeneity, ML offers an exciting approach to the systemic collection and analysis of electrical properties in a data-driven way. In this work, we discuss incorporation of ML to improve the field of electrical single cell analysis by addressing the design challenges to manipulate single cells and sophisticated analysis of electrical properties that distinguish cellular changes. Looking forward, we emphasize the opportunity to build on integrated systems to address common challenges in data quality and generalizability to save time and resources at every step in electrical measurement of single cells.

Research on High Precision Stiffness Modeling Method of Redundant Over-Constrained Parallel Mechanism.

Traditional stiffness modeling methods do not consider all factors comprehensively, and the modeling methods are not unified, lacking a global stiffness model. Based on screw theory, strain energy and the virtual work principle, a static stiffness modeling method for redundant over-constrained parallel mechanisms (PMs) with clearance was proposed that considers the driving stiffness, branch deformation, redundant driving, joint clearance and joint contact deformation. First, the driving stiffness and branch deformation were considered. According to the strain energy and Castiliano's second theorem, the global stiffness matrix of the ideal joint mechanism was obtained. The offset of the branch was analyzed according to the restraint force of each branch. The mathematical relationship between the joint clearance and joint contact deformation and the end deformation was established. Based on the probability statistical model, the uncertainty of the offset value of the clearance joint and the contact area of the joint caused by the coupling of the branch constraint force was solved. Finally, taking a 2UPR-RR-2RPU redundant PM as an example, a stiffness simulation of the mechanism was carried out using the finite element method. The research results show that the high-precision stiffness modeling method proposed in this paper is correct, and provides an effective method for evaluating the stiffness performance of the PM.

Accurate estimation of lake levels by the spatio-temporal modeling of satellite altimetry data

Satellite altimetry is currently one of the most widely used techniques for monitoring lake water levels. Raw altimetry measurements are generally delivered in an ellipsoid height system, while lake heights are finally required in a local level height system to ensure the lake surface being strictly flat under the physical constraints of both gravity and non-gravity factors. Therefore, there is always a need for height system transformation of altimetry data. Many of the previous studies have used orthometric heights in an approximate manner in the altimetric data processing. The geoid errors, together with other error sources, can introduce significant biases into water level estimations. In this paper, we propose a spatio-temporal regression method to calculate orthometric height to level height (O2L) differences, which can be used to yield accurate lake levels from satellite altimetry data. In the experiments conducted in this study, we chose ICESat-2 data to train the O2L bias correction model, and evaluated the algorithm performance on two large lakes in central Asian (Issyk-Kul and Baikal lakes) and one lake in North America (Lake Michigan). The experimental results indicate that the spatio-temporal modeling method performs better than the traditional spatial modeling method in terms of regression accuracy, and the water levels derived from the new method are much closer to in-situ gauging measurements. Influenced by O2L bias, there is often a jitter phenomenon in the time series of lake levels, and the results of different satellites are, accordingly, not mutually consistent. A visual analysis reveals that the spatio-temporal modeling method can greatly reduce the high-frequency bias.

Attitude control of 3D soft pneumatic actuators based on BP neural network

Soft pneumatic robotics have attracted considerable attention in recent years due to their deformation capabilities, which far exceed those of conventional robotics. However, precise control of soft pneumatic actuators remains a challenge due to the lack of model-based control techniques. This work aims to employ a high-precision and low-cost backpropagation (BP) neural network-based model method to control a 3D soft pneumatic actuator. Experiments show that this BP neural network-based model control method performs well in terms of precision, in which the errors of bending angle and deflection angle are within 0.8° and 1.2°, respectively, and the end point position error of the soft actuator is less than 2.5 mm, which is significantly better than traditional modeling methods, demonstrating the application potential of soft robots for high-precision operations.

Operational adjustment modeling approach based on Bayesian network transfer learning for new flotation process under scarce data

Establishing an intelligent operational adjustment model is significant for optimizing flotation performance and realizing the stable control of the flotation process. When sufficient historical data is available, the Bayesian network (BN) has been used to achieve this goal due to its powerful uncertainty modeling and reasoning capabilities. However, for a new flotation process in the early stage with scarce data, an accurate BN model can hardly be established to make reliable decisions. In this scenario, it is an effective way to transfer data or knowledge from other similar flotation processes to the new flotation process to help complete the target BN modeling task. Therefore, the operational adjustment modeling problem is transformed into the BN transfer learning problem. Moreover, this article proposes a new BN transfer learning approach, including structure and parameters transfer learning techniques. For structure transfer learning, a majority voting mechanism is proposed to get an integrated result of conditional independence tests and generates the network structure based on them. For parameters transfer learning, the sources are judged whether to be used to learn the parameters first, and then the suitable sources are used to learn the network parameters based on the weight determined by the similarity. The Asia network first assesses the proposed approach and then applies it to real-world flotation operational adjustment decisions. Experimental results on the Asia network show that the proposed approach can learn more accurate structure and parameters according to F1 score, Hamming distance, and KL divergence indicators compared to existing advanced methods. Furthermore, the experimental results of the flotation process show that the inference accuracy is improved from 57.1% to 85.7% compared with the traditional modeling method.

AI climate tipping‐point discovery (ACTD)

AI climate tipping‐point discovery (ACTD)

Urbanization Process: A Simulation Method of Urban Expansion Based on RF-SNSCNN-CA Model

One of the focal points in Geographic Information Science (GIS) is to uncover the spatial distribution patterns of geographical phenomena. In response to the insufficient spatial feature learning concerning neighborhoods in traditional machine learning-based Cellular Automata (CA) models for land use change, this study couples the Random Forest (RF) model and the Spatially Non-Stationary Convolutional Neural Network (SNSCNN) model to the CA model. The resulting RF-SNSCNN-CA model considers the issue of spatial non-stationarity by incorporating attention mechanisms. Using observed urban land change data from 2010 to 2017 in the 21 districts of Chongqing’s main city as an example, two sets of experiments comprising eight scenarios were designed to verify the neighborhood effects. The results demonstrate that the proposed RF-SNSCNN-CA model achieves an Overall Accuracy (OA) of 97.82%, Kappa of 0.7683, and Figure of Merit (FoM) of 0.3836. The study reveals the following findings. Firstly, the RF-SNSCNN-CA model integrates the dual advantages of traditional machine learning and deep learning models, in which SNSCNN improves by the combined effect of channel and spatial attention mechanisms improves the learning of neighborhood features; secondly, the machine learning-like urban sprawl CA modeling process, regardless of the approach taken to obtain development suitability, cannot completely replace the learning of the neighborhood part; lastly, the use of traditional neighborhood modeling methods may produce suppression of simulation results and make the model inadequately learn spatial features.

Learning dispatching rules via novel genetic programming with feature selection in energy-aware dynamic job-shop scheduling

The incorporation of energy conservation measures into production efficiency is widely recognized as a crucial aspect of contemporary industry. This study aims to develop interpretable and high-quality dispatching rules for energy-aware dynamic job shop scheduling (EDJSS). In comparison to the traditional modeling methods, this paper proposes a novel genetic programming with online feature selection mechanism to learn dispatching rules automatically. The idea of the novel GP method is to achieve a progressive transition from exploration to exploitation by relating the level of population diversity to the stopping criteria and elapsed duration. We hypothesize that diverse and promising individuals obtained from the novel GP method can guide the feature selection to design competitive rules. The proposed approach is compared with three GP-based algorithms and 20 benchmark rules in the different job shop conditions and scheduling objectives considered energy consumption. Experiments show that the proposed approach greatly outperforms the compared methods in generating more interpretable and effective rules. Overall, the average improvement over the best-evolved rules by the other three GP-based algorithms is 12.67%, 15.38%, and 11.59% in the meakspan with energy consumption (EMS), mean weighted tardiness with energy consumption (EMWT), and mean flow time with energy consumption (EMFT) scenarios, respectively.

Forecasting education expenditure with a generalized conformable fractional-order nonlinear grey system model

As an important human capital investment, education is an effective means to improve the comprehensive quality of people. Education expenditure is an important material guarantee for the development of educational undertakings. Education expenditure data is highly susceptible to numerous economic and social factors that complicate its nonlinear structure. In order to model the complex nonlinear problems of the system, this paper proposes a generalized conformable fractional-order nonlinear grey prediction model for the first time by analyzing the traditional time series-based modeling method in a nonlinear grey domain. The proposed model expands on the classical grey Bernoulli model by introducing the generalized conformable fractional accumulation as a new accumulation generator and utilizes error minimization principles in the modeling process. By altering the optimal order of the model and the cumulative generation operator, this model can adapt to various time series and reduce errors. Finally, the model is applied to education expenditure forecasting, and it is proved that the proposed model achieved good results and has higher accuracy than other models.

Auto-MatRegressor: liberating machine learning alchemists

Machine learning (ML) is widely used to uncover structure–property relationships of materials due to its ability to quickly find potential data patterns and make accurate predictions. However, like alchemists, materials scientists are plagued by time-consuming and labor-intensive experiments to build high-accuracy ML models. Here, we propose an automatic modeling method based on meta-learning for materials property prediction named Auto-MatRegressor, which automates algorithm selection and hyperparameter optimization by learning from previous modeling experience, i.e., meta-data on historical datasets. The meta-data used in this work consists of 27 meta-features that characterize the datasets and the prediction performances of 18 algorithms commonly used in materials science. To recommend optimal algorithms, a collaborative meta-learning method embedded with domain knowledge quantified by a materials categories tree is designed. Experiments on 60 datasets show that compared with the traditional modeling method from scratch, Auto-MatRegressor automatically selects appropriate algorithms at lower computational cost, which accelerates constructing ML models with good prediction accuracy. Auto-MatRegressor supports dynamic expansion of meta-data with the increase of the number of materials datasets and other required algorithms and can be applied to any ML materials discovery and design task.