Models Of Scales Research Articles

Design Method of a Novel Interface Connection Device for Multiscale Test Model Considering Multiparametric Similarity of Internal Forces

ABSTRACTThe multiscale model of building structures, as a balanced solution between accuracy and cost, has been widely used in the analysis of structural seismic performance. A reasonable interface connection method can accurately ensure load transfer and motion coordination between models of different scales. In this paper, a novel interface connection device and the corresponding design method for a multiscale test model of building structures were proposed, in which the upper structure with smaller sized components was replaced by a simplified story‐scale model, and the lower structure was adopted as a component‐scale model. The overall and local equations of motion for this multiscale model were established. For the interface connection between different scale models, a design method considering multiparametric similarity of shear force, axial force, and bending moment was proposed. In this method, the internal nodes at the interface of the component scale model were decomposed, and the coupling relationship of internal force between two adjacent nodes was established. The axial force of each node was decoupled into the interstory shear force and bending moment provided together. Additionally, the overturning moment is provided by adding the overlapping domain. According to the equilibrium relationships of the nodes at the interface, the corresponding transfer matrix was provided, and the design method of the interface connection device was proposed. The accuracy and feasibility of the method were validated by static and shaking table tests on a frame structure.

Retrieval In Decoder benefits generative models for explainable complex question answering

Large-scale Language Models (LLMs) utilizing the Chain-of-Thought prompting demonstrate exceptional performance in a variety of tasks. However, the persistence of factual hallucinations remains a significant challenge in practical applications. Prevailing retrieval-augmented methods treat the retriever and generator as separate components, which inadvertently restricts the generator’s capabilities to those of the retriever through intensive supervised training. In this work, we propose an unsupervised Retrieval In Decoder framework for multi-granularity decoding called RID, which integrates retrieval directly into the decoding process of generative models. It dynamically adjusts decoding granularity based on retrieval outcomes, and duly corrects the decoding direction through its direct impact on the next token. Moreover, we introduce a reinforcement learning-driven knowledge distillation method for adaptive explanation generation to better apply to Small-scale Language Models (SLMs). The experimental results across six public benchmarks surpass popular LLMs and existing retrieval-augmented methods, which demonstrates the effectiveness of RID in models of different scales and verifies its applicability and scalability.

Industrial‐Level Modulation of Catalyst‐Electrolyte Microenvironment for Electrocatalytic CO2 Reduction: Challenges and Advancements

AbstractCO2 reduction reaction (CO2RR), as a promising strategy for storing renewable energy and promoting carbon resource recycling, is critical for industrial development. Previous reports have extensively explored catalyst‐electrolyte microenvironmental modulation to elucidate coupling mechanisms and enhance catalytic conversion to multicarbon products. Currently, most reviews mainly focus on the impact of microenvironment modulation in low‐current systems on mechanism exploration and performance optimization, yet few of them can integrate macroscopic applications with microscopic investigations to explore the relevance between industrial development and catalyst‐electrolyte microenvironmental optimization. To address the gap, this review focuses on summarizing the challenges and advancements in microenvironment modulation for the development of high‐current devices. By introducing models of different scales sequentially, the connection between microenvironmental modulation and device performance is clarified. Then, various invalidation mechanisms and effective solutions are summarized to intuitively expound the impact of microenvironment modulation on high‐current stability. Meanwhile, as an intuitive measure of the rationality of microenvironment modulation, evaluation methods of device performance should be refined, which are also covered in further detail below. Finally, more valuable and challenging prospects are discussed for guiding the further industrial transformation of CO2RR.

Locating triarchic model constructs in the hierarchical structure of a comprehensive trait-based psychopathy measure: Implications for research and clinical assessment.

The triarchic model posits that distinct trait constructs of boldness, meanness, and disinhibition underlie psychopathy. The triarchic model traits are conceptualized as biobehavioral dimensions that can be assessed using different sets of indicators from alternative measurement modalities; as such, the triarchic model would hypothesize that these traits are not confined to any one item set. The present study tested whether the triarchic model dimensions would emerge from a hierarchical-structural analysis of the facet scales of the Elemental Psychopathy Assessment (EPA), an inventory designed to comprehensively index psychopathy according to the five-factor personality model. Study participants (Ns = 811, 170) completed the EPA and three different scale sets assessing the triarchic traits along with criterion measures of antisocial/externalizing behaviors. Bass-ackwards modeling of the EPA facet scales revealed a four-level structure, with factors at the third level appearing similar to the triarchic trait dimensions. An analysis in which scores for the Level-3 EPA factors were regressed onto corresponding latent-trait dimensions defined using the different triarchic scale sets revealed extremely high convergence (βs = .84-.91). The Level-3 EPA factors also evidenced validity in relation to relevant criteria, approximating and sometimes exceeding that evident for the Level-4 EPA factors. Together, these results indicate that the triarchic trait constructs are embedded in a psychopathy inventory designed to align with a general personality model and effectively predict pertinent external criteria. (PsycInfo Database Record (c) 2024 APA, all rights reserved).

Interface area density model for Large-Eddy Simulation of assisted atomization in fiber regime

In the context of simulating liquid rocket engines (LREs), accurately reproducing the complex combustion chamber environment poses a significant challenge. Recent works have shown that Direct Numerical Simulation has made good progress in achieving the simulation of the whole range of temporal and spatial two-phase flow scales involved. However, due to the computational costs, it is currently out of reach in an industrial context. Large-Eddy Simulations (LES) considerably reduce the computational cost but require modeling the two-phase flow at sub-grid level. A promising approach for such modeling is the interface area density (IAD) approach, which was developed initially for LES of flame fronts and subsequently adapted to Diesel jet atomization to recover spray droplet size distributions. In this method, a subgrid density of interface area is transported with an advection equation, while opportune source terms simulate its growth by breakup or its reduction by coalescence effects. In the context of the LREs assisted atomization, the complex mechanism of the assisted atomization in fiber regime must be taken into account by the subgrid model to recover the correct spray size distribution. The present work proposes a new interface area density evolution modeling for two-phase LES of coaxial assisted atomization in fiber regime. The proposed model focuses on recovering a realistic liquid-gas relative velocity at all the scales, enabling accurate modeling of sub-grid scales by using the IAD as a measure of liquid structure sizes.

A Systematic Review of the Potential Influence of Urbanization on the Regional Thunderstorm Process and Lightning Activity

In the context of global climate change, lightning disasters have emerged as a serious environmental factor that restricts the sustainable development of megacities. This paper provides a review of the research on the impact of urbanization on thunderstorm processes and lightning activity, exploring various aspects, such as aerosols, urban thermal effects, urban dynamic effects, and building morphology. Despite numerous significant achievements in the study of the impact of air pollutants on lightning activity, there is no consensus on whether aerosols serve to enhance or inhibit lightning activity. The temperature difference between the urban underlying surface and the natural underlying surface could sustain and promote the occurrence and development of convective systems, thus enhancing lightning activity. In terms of urban dynamics, the barrier effect has led to the maximum center of lightning appearing at the edge of a built-up area, which might be associated with factors, such as urban heat island (UHI) intensity, wind speed, synoptic background, and city size. Additionally, the size of a city and the height of the buildings was also an influencing factor on lightning activity. In summary, scholars have made progress in understanding the characteristics and drivers of urban lightning activity in recent years, but there are still some urgent problems that need to be solved: (1) How to analyze, comprehensively, the spatiotemporal patterns of urban lightning activity under different thunderstorm intensity backgrounds? (2) How to conduct analysis to investigate the influence of alterations in the boundary layer structure, water–heat energy balance, and water vapor circulation processes on urban lightning activity in the context of urbanization? (3) How to couple numerical models of different scales to enhance the understanding of the impact of complex underlying surfaces on urban lightning activity? Future studies could investigate the relationship between urbanization and thunderstorm/lightning activity using a combination of observational data, numerical modeling, and laboratory experiments, which holds promise for providing valuable theoretical insights and technical support to enhance the prediction, nowcasting, early warning, and risk assessment of thunderstorms and lightning in urban areas.

CPMI-ChatGLM: parameter-efficient fine-tuning ChatGLM with Chinese patent medicine instructions

Chinese patent medicine (CPM) is a typical type of traditional Chinese medicine (TCM) preparation that uses Chinese herbs as raw materials and is an important means of treating diseases in TCM. Chinese patent medicine instructions (CPMI) serve as a guide for patients to use drugs safely and effectively. In this study, we apply a pre-trained language model to the domain of CPM. We have meticulously assembled, processed, and released the first CPMI dataset and fine-tuned the ChatGLM-6B base model, resulting in the development of CPMI-ChatGLM. We employed consumer-grade graphics cards for parameter-efficient fine-tuning and investigated the impact of LoRA and P-Tuning v2, as well as different data scales and instruction data settings on model performance. We evaluated CPMI-ChatGLM using BLEU, ROUGE, and BARTScore metrics. Our model achieved scores of 0.7641, 0.8188, 0.7738, 0.8107, and − 2.4786 on the BLEU-4, ROUGE-1, ROUGE-2, ROUGE-L and BARTScore metrics, respectively. In comparison experiments and human evaluation with four large language models of similar parameter scales, CPMI-ChatGLM demonstrated state-of-the-art performance. CPMI-ChatGLM demonstrates commendable proficiency in CPM recommendations, making it a promising tool for auxiliary diagnosis and treatment. Furthermore, the various attributes in the CPMI dataset can be used for data mining and analysis, providing practical application value and research significance.

Modeling of the influence of local heat sources on a light gas stratification formation and erosion in a large-scale experimental facility using eddy resolving numerical approach

The main objective of this work is to evaluate the results of application of Implicit Large Eddy Simulation (ILES) turbulence modeling approach to simulate the complex hydrogen safety problem. The paper presents the results of numerical simulation of the OECD/NEA HYMERES (HYdrogen Mitigation Experiments for Reactor Safety) HP2 series tests in the PANDA facility, aimed at studying the thermal effect of passive autocatalytic recombiner (PAR) operation on mixing the containment atmosphere, using CABARET-SC1 code. The code is based on the eddy resolving CABARET technique, which allows implicit modeling of the subgrid turbulence scales without using tuning parameters (ILES approximation). The natural convection flow inside the PAR model was calculated explicitly using a simplified model of the heating section and the housing. Good agreement of calculated transient with experimental data was obtained, which indicates the effectiveness of such approach.

Psychometric Validation of the Translated and Adapted Bengali Version of the HLS-EU-Q16/Q6 for Adults.

Health Literacy (HL) emerges as a critical tool in addressing the escalating burden of noncommunicable diseases (NCDs) and their associated costs. Particularly in Bangladesh, where the inadequacy of HL presents multifaceted challenges, there is an urgent need to address this issue. This study aimed to translate and evaluate the Bengali versions of the European Health Literacy Questionnaire with 16 items (HLS-EU Q16) and its shorter 6-item version (HLS-EU Q6), as there is currently no validated Bengali tool for assessing HL. This article used a subset of data from a Bangladeshi national survey of Primary Healthcare (PHC) facilities. The study included adults seeking Non-Communicable Disease (NCD) services at PHCs. Validity and reliability testing succeeded in a detailed back-to-back translation. The statistics covered were descriptive, Cronbach's internal consistency, confirmatory factor analysis, and the chi-square test. Following the translation and preliminary testing, minor rephrasing and the insertion of Item-Relevant Stimulus Material were performed to ensure cultural equivalency. The Confirmatory Factor Analysis produced a 3-factor structure for the HLS-EU-Q16 that included a second-order general component, confirming the viability of using an HL total score. A 3-factor model based on a priori was determined to be suitable for the factor structure of the HLS-EU-Q6. The model fit indices (Chi-square/df, TLI, AGFI, CFI, GFI, SRMR, RMSEA, and PCLOSE) supported CFA models of both scales. The internal consistency of the translated and adapted instruments was α = .934 and .857, respectively. This study showed that the Bengali version of the HLS-EU-Q16 and HLS-EU-Q6 are psychometrically sound, have clear factor structures, and are equivalent to the original models. However, the HLS-EU_Q16 is recommended over the shorter version considering its better psychometric properties.

Multiscale modelling of hepatitis B virus at cell level of organization.

Multiscale modelling is a promising quantitative approach for studying infectious disease dynamics. This approach garners attention from both individuals who model diseases and those who plan for public health because it has great potential to contribute in expanding the understanding necessary for managing, reducing, and potentially exterminating infectious diseases. In this article, we developed a nested multiscale model of hepatitis B virus (HBV) that integrates the within-cell scale and the between-cell scale at cell level of organization of this disease system. The between-cell scale is linked to the within-cell scale by a once off inflow of initial viral infective inoculum dose from the between-cell scale to the within-cell scale through the process of infection; the within-cell scale is linked to the between-cell scale through the outflow of the virus from the within-cell scale to the between-cell scale through the process of viral shedding or excretion. The resulting multiple scales model is bidirectionally coupled in such a way that the within-cell scale and between-cell scale sub-models mutually affect each other, creating a reciprocal relationship. The computed reproductive number from the multiscale model confirms that the within-host scale and the between-host scale influence each other in a reciprocal manner. Numerical simulations are presented that also confirm the theoretical results and support the initial assumption that the within-cell scale and the between-cell scale influence each other in a reciprocal manner. This multiple scales modeling approach serves as a valuable tool for assessing the impact and success of health strategies aimed at controlling hepatitis B virus disease system.

A Generative Adversarial and Spatiotemporal Differential Fusion Method in Radar Echo Extrapolation

As an important part of remote sensing data, weather radar plays an important role in convective weather forecasts to reduce extreme precipitation disasters. The existing radar echo extrapolation methods do not utilize the local natural characteristics of the radar echo effectively but only roughly extract the whole characteristics of the radar echo. To address these challenges, we design a spatiotemporal difference and generative adversarial fusion model (STDGAN). Specifically, a spatiotemporal difference module (STD) is designed to extract local weather patterns and model them in detail. In our model, spatiotemporal difference information and spatiotemporal features captured by the model itself are fused together. In addition, our model is trained in a generative adversarial network (GAN) framework; it helps to generate a clearer map of future radar echoes at the image level. The discriminator consists of multi-scale feature extractors, which can simulate weather models of various scales more completely. Finally, extrapolation experiments were conducted using actual radar echo data from Shijiazhuang and Nanjing. The experiments have shown that our model has a more accurate prediction performance for predicting local weather patterns and overall echo change trajectories compared with previous research models. Our model achieved MSE, PSNE, and SSIM values of 132.22, 37.87, and 0.796, respectively, on the Shijiazhuang radar echo dataset. In addition, our model also showed better performance results on the Nanjing radar echo dataset. The results show that the MSE was 49.570, the PSNR was 0.714, and the SSIM was 30.633. The CC value was 0.855.

Structure-Aware Low-Rank Adaptation for Parameter-Efficient Fine-Tuning

With the growing scale of pre-trained language models (PLMs), full parameter fine-tuning becomes prohibitively expensive and practically infeasible. Therefore, parameter-efficient adaptation techniques for PLMs have been proposed to learn through incremental updates of pre-trained weights, such as in low-rank adaptation (LoRA). However, LoRA relies on heuristics to select the modules and layers to which it is applied, and assigns them the same rank. As a consequence, any fine-tuning that ignores the structural information between modules and layers is suboptimal. In this work, we propose structure-aware low-rank adaptation (SaLoRA), which adaptively learns the intrinsic rank of each incremental matrix by removing rank-0 components during training. We conduct comprehensive experiments using pre-trained models of different scales in both task-oriented (GLUE) and task-agnostic (Yelp and GYAFC) settings. The experimental results show that SaLoRA effectively captures the structure-aware intrinsic rank. Moreover, our method consistently outperforms LoRA without significantly compromising training efficiency.

Correlation between impact angle and corrosion-erosion damage behavior of ferritic/martensitic steel exposed to flowing oxygen-saturated lead-bismuth eutectic

Impact angle is considered to be one of the main factors that influence the damage behavior of materials exposed to flowing lead-bismuth eutectic (LBE) conditions in Gen IV LBE-cooled fast reactors, especially in the areas where the flow changes its direction sharply, such as sudden expansions or elbows. In the current study, the corrosion-erosion damage behaviors of T91 ferritic/martensitic steel under different impact angles were investigated in flowing (∼3 m/s) oxygen-saturated LBE at 500 °C. The results show that the damage behaviors of T91 steel are closely related to the impact angles. When the flowing direction is parallel to the specimen surface, i.e., 0° impact angle, the oxide scales show partially damaged features in the form of local buckling and exfoliation after exposure for 300 h. With continued exposure for 600 h and 1000 h, the specimen surface seems to be covered completely by the undamaged oxide scales again. In sharp contrast, when the flowing direction is perpendicular to the specimen surface, i.e., 90° impact angle, the oxide scales are almost totally delaminated after exposure for 600 h and 1000 h. As a consequence, the liquid LBE permeates into the T91 steel substrate along the boundaries of prior austenite to induce intergranular corrosion. Correspondingly, the possible reasons behind such phenomena are discussed. Meanwhile, the damage process models of oxide scales associated with 0° and 90° impact angles are established, respectively.

Weinberger Adjustment Inventory Portuguese Version: A Study of Cross-Cultural Adaptation and Measurement Invariance

We examine the psychometric properties of the Weinberger Adjustment Inventory (WAI) from a modern psychometric perspective, including second-order models, better estimation of reliability, and measurement invariance, among a sample of young adults (N = 610, M = 21.33 years, SD = 3.09 years) from Portugal. Results indicated that only the four-factor intercorrelated confirmatory models of the WAI Distress and Restraint scales, and the two-factor intercorrelated confirmatory models of the WAI Defensiveness scales showed adequate fits. The WAI showed good reliability and distinctive correlates with other psychometric measures and criterion variables. Cross-gender measurement invariance was established. Our findings support the WAI as a valid and reliable measure.

A deep reinforcement learning framework for dynamic optimization of numerical schemes for compressible flow simulations

Marginal or under-resolved simulations of compressible flow configurations that often occur in practical applications classically are enabled by administering sufficient numerical dissipation to keep the simulation stable. Such measures, however, often are physically inconsistent due to non-selectively altering of dynamics across scales. Sustaining physically consistent large scale dynamics requires the numerical solution to effectively model non-resolved small scale dynamics. In this work, we propose a general deep-reinforcement-learning framework for devising an agent to interact with high-resolution scheme in order to balance dissipation and dispersion such that physically consistent modeling of non-resolved scales is achieved. A densely distributed reward function without involving labeled data is defined. The agent is trained on low-resolution uniform grids that capture the dominant flow structures. We demonstrate that it can be applied directly to high-resolution simulations without the need for retraining or fine-tuning, thereby, demonstrating significantly improved modeling performance compared to empirically designed high-resolution schemes. The proposed methodology opens a new path for self-adaptive numerical solutions whose truncation errors act as physically consistent model for unresolved scales of widely differing flow configurations.

Origin identification of Polygonatum cyrtonema based on hyperspectral data

In this study, visual-near infrared(VNIR), short-wave infrared(SWIR), and VNIR + SWIR fusion hyperspectral data of Polygonatum cyrtonema from different geographical origins were collected and preprocessed by first derivative(FD), second derivative(SD), Savitzky-Golay smoothing(S-G), standard normalized variate(SNV), multiplicative scatter correction(MSC), FD+S-G, and SD+S-G. Three algorithms, namely random forest(RF), linear support vector classification(LinearSVC), and partial least squares discriminant analysis(PLS-DA), were used to establish the identification models of P. cyrtonema origin from three spatial scales, i.e., province, county, and township, respectively. Successive projection algorithm(SPA) and competitive adaptive reweighted sampling(CARS) were used to screen the characteristic bands, and the P. cyrtonema origin identification models were established according to the selected characteristic bands. The results showed that(1)after FD preprocessing of VNIR+SWIR fusion hyperspectral data, the accuracy of recognition models established using LinearSVC was the highest, reaching 99.97% and 99.82% in the province origin identification model, 100.00% and 99.46% in the county origin identification model, and 99.62% and 98.39% in the township origin identification model. The accuracy of province, county, and township origin identification models reached more than 98.00%.(2)Among the 26 characteristic bands selected by CARS, after FD pretreatment, the accuracy of origin identification models of different spatial scales was the highest using LinearSVC, reaching 98.59% and 97.05% in the province origin identification model, 97.79% and 94.75% in the county origin identification model, and 90.13% and 87.95% in the township origin identification model. The accuracy of identification models of different spatial scales established by 26 characteristic bands reached more than 87.00%. The results show that hyperspectral imaging technology can realize accurate identification of P. cyrtonema origin from different spatial scales.

Self-adaptive turbulence eddy simulation of a premixed jet combustor

The self-adaptive turbulence eddy simulation (SATES) is employed to investigate the lean premixed methane/air turbulent flame in a single-nozzle model gas turbine combustor, in which the high axial momentum jet issuing from an off-center nozzle facilitates the development of a large-scale, dominant lateral recirculation zone that stabilizes the flame. For turbulence modeling, the SATES method can dynamically adjust the proportion of resolving and modeling of turbulent scales according to the local grid scale and turbulence length scales, thus depressing the gird-sensitivity in large eddy simulation (LES) calculation. For combustion modeling, the thickened flame model with a reduced chemistry mechanism is integrated into the SATES turbulence modeling framework to capture the unsteady flame dynamics. The accuracy of SATES results is assessed against experimental data, as well as the ones from LES and the detached eddy simulation (DES) of this burner with the same combustion model and grids. The predicted length of large-scale recirculation of the flow field by SATES is significantly better than that by LES and DES with low mesh resolution. Detailed comparisons show that the SATES adaptively improves the modeling degree of near-wall turbulence to improve the prediction accuracy with the low grid resolution. Similar to LES, the SATES method does not cause serious delay of shear layer instability and weakening of flame wrinkle observed in DES. The study demonstrates the suitability and accuracy of the hybrid turbulence modeling methods of SATES for complex turbulent flame simulations coupled with suitable combustion model.

A multi-kernel and multi-scale learning based deep ensemble model for predicting recurrence of non-small cell lung cancer.

Predicting recurrence in patients with non-small cell lung cancer (NSCLC) before treatment is vital for guiding personalized medicine. Deep learning techniques have revolutionized the application of cancer informatics, including lung cancer time-to-event prediction. Most existing convolutional neural network (CNN) models are based on a single two-dimensional (2D) computational tomography (CT) image or three-dimensional (3D) CT volume. However, studies have shown that using multi-scale input and fusing multiple networks provide promising performance. This study proposes a deep learning-based ensemble network for recurrence prediction using a dataset of 530 patients with NSCLC. This network assembles 2D CNN models of various input slices, scales, and convolutional kernels, using a deep learning-based feature fusion model as an ensemble strategy. The proposed framework is uniquely designed to benefit from (i) multiple 2D in-plane slices to provide more information than a single central slice, (ii) multi-scale networks and multi-kernel networks to capture the local and peritumoral features, (iii) ensemble design to integrate features from various inputs and model architectures for final prediction. The ensemble of five 2D-CNN models, three slices, and two multi-kernel networks, using 5 × 5 and 6 × 6 convolutional kernels, achieved the best performance with an accuracy of 69.62%, area under the curve (AUC) of 72.5%, F1 score of 70.12%, and recall of 70.81%. Furthermore, the proposed method achieved competitive results compared with the 2D and 3D-CNN models for cancer outcome prediction in the benchmark studies. Our model is also a potential adjuvant treatment tool for identifying NSCLC patients with a high risk of recurrence.

Characteristics of dispersion and attenuation with a unified multiscale model in porous medium containing saturated fluid and multiscale fractures

To analyze seismic wave field characteristics and characterize fracture (or crack) reservoirs, it is essential to build proper wave-induced fluid flow (WIFF) models of various scales. There is much research related to WIFF fractured models that are mainly suitable only for different scales of fracture (crack) medium separately, such as microscopic cracks and mesoscopic fractures. Based on previous research, it is proposed a unified multiscale (mesoscopic and microscopic) dispersion and attenuation model for the medium with fractures, cracks, pores, and fluid. The formulation uses frequency-dependent fractured (or crack) parameters expressed as the form of multiplication of fracture parameters and relaxation function. The advantage of this method is that it is convenient to build fluid porous medium with different scale fractures (or cracks) and various fracture (or crack) configurations. The numerical simulation results prove the correctness and applicability of the proposed extended WIFF dispersion and attenuation model. Based on our proposed method, the characteristics of the dispersion and attenuation of the multiscale fractured model are analyzed. The results indicate that the characteristics of velocity dispersion and wave attenuation in the medium with different scales of fractures are similar. The velocity increases with frequency and finally tends to be stable. The main difference among the different scales of models is that the frequency bands of the dispersion and attenuation occurring are different. We also discover that for multiscale fractured medium, the characteristic frequency and attenuation peaks do not necessarily correspond to each other, which makes the analysis of the dispersion and attenuation more complicated.

An automated method to build 3D multi-scale geological models for engineering sedimentary layers with stratum lenses

Automatically building three-dimensional (3D) engineering geological models of sedimentary layer is still challenging. In particular, there is a lack of practical and scientific methods for constructing models of strata containing stratum lenses and pinch-out. This study develops an engineering geological structure modeling method, called the multi-scale drill control method, for strata containing stratum lenses and stratum pinch-out using drill data. Based on the stratigraphic code classification strategy, this approach first determines n levels of stratigraphic columns in the study area and the spatial distribution and occupation relationships of each stratum inside. Thereafter, geological surfaces are interpolated, intersected, corrected, and built by following the order of the stratigraphic column from top to bottom and from coarsest to finest level. Subsequently, surfaces with 0 thickness between the top and bottom surfaces are removed, and the vertical side surfaces are constructed. Finally, geo-bodies are built on seamless surfaces to conduct model creation. As required by modelers, models of different scales are applied, or multi-scaled models are mixed for use, thus providing support for later geological analysis. The case study with complicated geological conditions indicates that the method can accurately reproduce the geo-bodies' geometry and distribution, significantly enhancing geological model building efficiency.