作者机构：
[Dongping Xiong; Lijun Ouyang] School of Computing/Software, University of South China, Hengyang, Hunan, China [email protected];[Yudan Li; Xiaozhi Zhang] School of Electrical Engineering, University of South China, Hengyang, Hunan, China [email protected]
会议名称：
ISAIMS '24: Proceedings of the 2024 5th International Symposium on Artificial Intelligence for Medicine Science
会议论文集名称：
Artificial Intelligence for Medicine Science
摘要：
Accurately segmenting brain tumors from multimodal MRI sequences is a key prerequisite for brain tumor diagnosis, prognosis assessment, and surgical treatment. However, in practical applications, one or more modal data is often missing due to image corruption, different acquisition protocols, artifacts, contrast agent allergies, or cost considerations. To address the challenges of brain tumor segmentation under modality loss, this paper proposes an innovative tumor feature perception strategy. The core of this strategy is to introduce a Mamba-based Encoder (MBE) architecture, which effectively improves the feature expression ability of each modality encoder under limited data conditions. In view of the irregularity of tumor morphology, a Modulation and Demodulation Fusion Block (MDFB) is designed to accurately capture the semantic features of the tumor from the missing multimodal image data, providing strong guidance for the network to locate the tumor area. Experimental results on the widely used BraTS2020 dataset demonstrate the effectiveness of MMITS, and the brain tumor segmentation effect under various incomplete modalities is better than the state-of-the-art methods.
作者机构：
[Xig, Jiaojiao; Li, Wenjun; Ma, Wanjun; Peng, Huan] Changsha Univ Sci & Technol, Hunan Prov Key Lab Intelligent Proc Big Data Tran, Changsha, Hunan, Peoples R China.;[Liang, Weijun] Univ South China, Affiliated Changsha Cent Hosp, Hengyang Med Sch, Changsha, Hunan, Peoples R China.
会议名称：
7th Chinese Conference on Pattern Recognition and Computer Vision
会议时间：
OCT 18-20, 2024
会议地点：
Urumqi, PEOPLES R CHINA
会议主办单位：
[Li, Wenjun;Xig, Jiaojiao;Peng, Huan;Ma, Wanjun] Changsha Univ Sci & Technol, Hunan Prov Key Lab Intelligent Proc Big Data Tran, Changsha, Hunan, Peoples R China.^[Liang, Weijun] Univ South China, Affiliated Changsha Cent Hosp, Hengyang Med Sch, Changsha, Hunan, Peoples R China.
会议论文集名称：
Lecture Notes in Computer Science
关键词：
Embedded Deep Learning;Rifampicin;Drug Resistance;Tuberculosis;CT Images;Diagnostic Application
摘要：
In the treatment of tuberculosis (TB), drug-resistant tuberculosis arises when Mycobacterium tuberculosis undergoes genetic mutations or acquires resistance through horizontal gene transfer. Identifying the treatment response of TB patients to Rifampicin, a principal medication for TB treatment, is essential for healthcare professionals to make timely and accurate diagnoses. Not only can this approach save on the costs and duration of TB treatment, but it also helps prevent the disease's spread and fatalities. Traditional methods for diagnosing Rifampicin-resistant TB involve molecular biology tests and drug susceptibility testing, which are time-consuming, expensive, and labor-intensive. To assist physicians in diagnosing the treatment response of TB patients to Rifampicin more rapidly and efficiently, this study introduces a computer-aided diagnostic algorithm based on Embedded Deep Learning (EDL). Initially, CT images from target patients at two imaging centers were collected. The classifier model used in this research combines image preprocessing techniques, three convolutional neural networks, and decision fusion technology to enhance the model's classification efficiency and reduce overfitting. Additionally, the Grad-CAM model was utilized for visualizing the areas of lesions. In the test sets from both centers, the Embedded Deep Learning Model (EDL Model) demonstrated superior performance over other models by combining hard voting or soft voting mechanisms, with an average accuracy improvement of 3.16-16.87%, AUC increase of 3.05-12.66%, and F1-score enhancement of 6.38-22.49%. The diagnostic tool developed in this research for assisting in the diagnosis of TB patients' response to Rifampicin treatment has significant clinical potential, particularly in settings lacking specialized radiological expertise.
作者机构：
[Zhicheng Bi; Jinfeng Xiao; Chaofeng Wang] School of Electrical Engineering, University of South China, Hengyang, Hunan, China [email protected]
会议名称：
AISNS '24: Proceedings of the 2024 2nd International Conference on Artificial Intelligence, Systems and Network Security
摘要：
Underwater acoustic communication networks can satisfy the need for long-distance reliable communication for underwater node, serving as a key technology for the informatization and intelligentization of such communication. However, due to the use of sound waves as a carrier for information, underwater acoustic networks face complex challenges such as long propagation delays, severe channel attenuation, multipath effects, and environmental noise. To enhance the communication efficiency and reduce network congestion of underwater acoustic networks, this study proposes a MAC protocol based on deep reinforcement learning to adaptively adjust the transmission strategy of network nodes. Each network node collaboratively decides the optimal transmission slot based on its own data buffer status and the history of transmission conflicts, to suppress transmission conflicts and thereby maximize transmission rates. Simulation results show that the designed protocol significantly improves the network throughput while ensuring the fairness of transmission among nodes, providing an important foundation for the rapid exchange of information between underwater nodes.
会议名称：
13th International Conference on Natural Language Processing and Chinese Computing
会议时间：
NOV 01-03, 2024
会议地点：
Westlake University, Hangzhou, PEOPLES R CHINA
会议主办单位：
Westlake University
会议论文集名称：
Lecture Notes in Artificial Intelligence
关键词：
large language models;group process;natural language process
摘要：
Exploring the alignment of multi-agent systems with human values during group process is an essential step towards the development of artificial general intelligence. In this work, we present a novel approach to systematically evaluate factors that influence the value orientation of large language models (LLMs) in simulating human group process. Our proposed framework, which requires neither fine-tuning nor pre-training, enables LLMs to simulate debates among personas on a given topic, autonomously assess response confidence, and retrieve external information to enhance low-confidence responses. We conduct comparative experiments using the same framework with human participants. Additionally, we introduce a comprehensive set of evaluation metrics to reveal discrepancies in value orientation alignment between LLM systems and human systems in group process.
摘要：
Contemporary makeup approaches primarily hinge on unpaired learning paradigms, yet they grapple with the challenges of inaccurate supervision (e.g., face misalignment) and sophisticated facial prompts (including face parsing, and landmark detection). These challenges prohibit low-cost deployment of facial makeup models, especially on mobile devices. To solve above problems, we propose a brand-new learning paradigm, termed "Data Amplify Learning (DAL)," alongside a compact makeup model named "TinyBeauty." The core idea of DAL lies in employing a Diffusion-based Data Amplifier (DDA) to "amplify" limited images for the model training, thereby enabling accurate pixel-to-pixel supervision with merely a handful of annotations. Two pivotal innovations in DDA facilitate the above training approach: (1) A Residual Diffusion Model (RDM) is designed to generate high-fidelity detail and circumvent the detail vanishing problem in the vanilla diffusion models; (2) A Fine-Grained Makeup Module (FGMM) is proposed to achieve precise makeup control and combination while retaining face identity. Coupled with DAL, TinyBeauty necessitates merely 80K parameters to achieve a state-of-the-art performance without intricate face prompts. Meanwhile, TinyBeauty achieves a remarkable inference speed of up to 460 fps on the iPhone 13. Extensive experiments show that DAL can produce highly competitive makeup models using only 5 image pairs. Please visit https://github.com/TinyBeauty for code and demos.
作者机构：
[Wang, Jincan; Song, Yucheng; Lan, Peng; Ge, Yifan; Liao, Zhifang] Cent South Univ, Sch Comp Sci & Engn, Changsha, Peoples R China.;[Guo, Jia] Cent South Univ, Xiangya Sch Nursing, Changsha, Peoples R China.;[Li, Lifeng] Univ South China, Affiliated Changsha Cent Hosp, Hengyang Med Sch, Dept Radiol, Changsha, Hunan, Peoples R China.;[Li, Lifeng] Nanchang Univ, Med Imaging Ctr, Affiliated Hosp 1, Nanchang, Jiangxi, Peoples R China.
会议名称：
7th Chinese Conference on Pattern Recognition and Computer Vision
会议时间：
OCT 18-20, 2024
会议地点：
Urumqi, PEOPLES R CHINA
会议主办单位：
[Song, Yucheng;Wang, Jincan;Ge, Yifan;Liao, Zhifang;Lan, Peng] Cent South Univ, Sch Comp Sci & Engn, Changsha, Peoples R China.^[Guo, Jia] Cent South Univ, Xiangya Sch Nursing, Changsha, Peoples R China.^[Li, Lifeng] Univ South China, Affiliated Changsha Cent Hosp, Hengyang Med Sch, Dept Radiol, Changsha, Hunan, Peoples R China.^[Li, Lifeng] Nanchang Univ, Med Imaging Ctr, Affiliated Hosp 1, Nanchang, Jiangxi, Peoples R China.
会议论文集名称：
Lecture Notes in Computer Science
关键词：
Medical image classification;Noisy label;Knowledge distillation with multiple teachers;Deep learning;Point-of-care
摘要：
In recent years, the development of medical imaging technology has transformed imaging solutions from laboratory-based to point-of-care imaging with real-time capabilities. However, these point-of-care devices are often constrained by environmental factors such as ambient light and noise, leading to poor image quality and consequently affecting the diagnostic accuracy of point-of-care devices. Furthermore, due to the need for lightweight models in point-of-care devices, traditional models fail to meet requirements in terms of computational resources, model parameters, and inference time. Therefore, to address the aforementioned issues, this paper proposes an optimized lightweight student model that focuses on residual information. A lightweight structure based on Shift MLP is designed on the residual branch of the model to enhance the model's capability to acquire spatial feature information at multiple scales. Simultaneously, we propose a multi-teacher distillation strategy to improve the accuracy and noise-resistance of the student model. Firstly, we introduce an adaptive learning approach based on auxiliary teachers, leveraging unlabeled and noisy data for adaptive learning to enhance the model's robustness. Then, we design a global teacher model to enhance the accuracy of the student model and indirectly improve the teaching ability of auxiliary teacher model, thereby achieving knowledge transfer at a global level. We evaluate our approach on two public medical image classification datasets, and the results demonstrate that while almost maintaining accuracy, we reduce the number of parameters by 38 times, decrease computational complexity by 11 times, and achieve an inference time of only 18.94ms on CPU.
摘要：
The popularity of Connected Autonomous Vehicles (CAVs) has led to improvements in the efficiency of the transportation system. Controller Area Network (CAN) is the standard communication protocol used in CAVs. However, the absence of effective security measures within CAN has resulted in vulnerabilities that can be exploited by attackers. To address this issue, we propose an Intrusion Detection System (IDS) using Time series Imaging and Deep Learning called TIDL-IDS. First, the CAN ID in the CAN frame is encoded as the Markov Transition Field (MTF) images to take into account the temporal characteristics of the CAN time series. Given the limited resources in the vehicle network environment, a simple four-layer deep convolutional neural network is designed to classify the converted images. A comprehensive evaluation of TIDL-IDS on real datasets demonstrates that the proposed method outperforms the other two baseline methods in terms of F1 score and accuracy. Furthermore, the model parameters are also superior to those of the other methods.
作者机构：
[Yehong Zhao; Jinmei Li] College of Electrical and Information Engineering, Hunan Institute of Traffic Engineering, Hengyang, Hunan, China [email protected];[Junwen Deng; Mingyue Zhang; Huan Liu; Mingtao Liu] School of Computing / Software, University of South China, Hengyang, Hunan, China [email protected]
会议名称：
EITCE '24: Proceedings of the 2024 8th International Conference on Electronic Information Technology and Computer Engineering
会议论文集名称：
Electronic Information Technology and Computer Engineering
摘要：
The aging population has led to an increased need for reliable fall detection systems that can provide timely assistance to individuals who have fallen. This paper presents a multi-sensor fusion system designed to detect falls and issue early warnings, overcoming the limitations of camera-based systems, particularly in areas where camera surveillance is not feasible. We utilize a combination of an MPU6050 six-axis pose sensor and a MAX30102 heart rate and blood oxygen sensor to achieve high accuracy in fall detection. The system is designed to differentiate between falling and normal physical activities, reducing false positives. When a fall is detected, the system triggers an alert and sends real-time health data to a TCP server for monitoring on a mobile terminal. Our approach addresses camera blind spots and provides a comprehensive solution for fall detection in non-visual environments.
会议名称：
13th International Conference on Natural Language Processing and Chinese Computing
会议时间：
NOV 01-03, 2024
会议地点：
Westlake University, Hangzhou, PEOPLES R CHINA
会议主办单位：
Westlake University
会议论文集名称：
Lecture Notes in Artificial Intelligence
关键词：
Biomedical Relation Extraction;Natural Language Inference;Counterfactual Inference;Debias
摘要：
Biomedical relation extraction is a core problem in biomedical natural language processing, whose goal is to classify the relations between entity mentions within a classified given text, being modeled as a classification method. There has been some recent work on converting biomedical RE to other auxiliary tasks to deal with it. However, they have all neglected the fact that this form of RE is subject to the inherent bias of the auxiliary task, resulting in the inability to make truly reasonable predictions. In this paper, we propose the CouBRE method, which excludes the direct influence of premise-only and hypothesis-only branches in the NLI task when converting biomedical RE to NLI task from a causal perspective on the RE results, allowing the NLI model to make valid predictions based on the combined information of premise and hypothesis rather than based on shortcut paths. Also by removing sample selection bias and label bias in the dataset, CouBRE can make unbiased prediction results more clearly. Comprehensive experiments demonstrate the effectiveness of our approach in low-resource scenarios, outperforming previous state-of-the-art models and remaining competitive even in full-shot scenarios.
作者机构：
[Liu, Jie; Dai, Miaoxin] Univ South China, Hengyang 421200, Hunan, Peoples R China.;[Chen, Zhi; Jiang, Wei; Yan, Hao; Chen, Z] Nucl Power Inst China, Sci & Technol Reactor Syst Design Technol Lab, Chengdu 610213, Sichuan, Peoples R China.
会议名称：
International Conference on Power Electronics and Artificial Intelligence (PEAI)
会议时间：
JAN 19-21, 2024
会议地点：
Xiamen, PEOPLES R CHINA
会议主办单位：
[Dai, Miaoxin;Liu, Jie] Univ South China, Hengyang 421200, Hunan, Peoples R China.^[Chen, Zhi;Jiang, Wei;Yan, Hao] Nucl Power Inst China, Sci & Technol Reactor Syst Design Technol Lab, Chengdu 610213, Sichuan, Peoples R China.
会议论文集名称：
Power Electronics and Artificial Intelligence
关键词：
Stacked denoising self-coding network;Feature extraction;Multi-kernel relevance vector machine;Safety class signal conditioning module;Fault prediction
摘要：
To improve the reliability and maintainability of the nuclear safety class DCS system, this paper conducts a study on fault prediction of key components in the output circuit of the nuclear safety class signal conditioning module. To address the issues of insufficient feature extraction for minor offset fault characteristics and low accuracy of fault prediction, we proposed a prediction model based on stacked denoising autoencoder (SDAE) feature extraction and improved multi-kernel relevance vector machine (MKRVM) model. Therefore, fault simulation modeling is performed for key components of the signal output circuit to obtain fault datasets of key components, and the SDAE model is used to extract fault features. And the fault degree index is obtained by calculating the Pearson correlation coefficient between these fault features.The fault prediction model based on MKRVM is established, and the combination coefficients of the kernel functions in the MKRVM model are optimized using adaptive gray wolf optimization algorithm (AGWO). The prediction performance evaluation indicators are used to evaluate the prediction results of the AGWO-MKRVM model, RVM model. The results show that the MKRVM model optimized by AGWO has better prediction accuracy for the faults of the circuit critical components, moreover, can accurately and stably predict the fault trend of the circuit.
作者机构：
[Honghao Luo; Yifa Sheng] School of Electrical Engineering, University of South China, Hengyang, China
会议名称：
2024 4th International Conference on Energy, Power and Electrical Engineering (EPEE)
会议时间：
20 September 2024
会议地点：
Wuhan, China
会议论文集名称：
2024 4th International Conference on Energy, Power and Electrical Engineering (EPEE)
关键词：
Local discharge;prpd mapping;convolutional ne ural network;support vector machine
摘要：
With the wide application of the deep learning algorithm in the field of PD pattern recognition, it makes up for the disadvantages of the traditional method, at the same time, it can solve the problems of traditional machine learning methods such as poor performance and under-fitting when the training sample size is large.By training the convolutional neural network to recognize the PRPD maps, not only can we avoid the complex feature extraction of statistical data, but also can greatly improve the work efficiency and the accuracy of localization type recognition. For this reason, this paper is designed based on the comparative analysis of convolutional neural network and support vector machine. First, an experimental model of four typical partial discharge defects in GIS was designed and constructed, and experimental data were collected. The PRPD spectrogram training set is then augmented with data using conditional generative adversarial networks. Finally, the PRPD spectra of each type of defects are identified and classified using convolutional networks and conventional machine learning support vector machines, respectively, and analyzed for comparison.
摘要：
The point cloud can be de-entangled into sharply varying frame components and slowly varying plane components, which complement each other to construct complete point cloud geometric information. Geometric de-entanglement attention network (GDAnet) dynamically de-entangles the point cloud into the contour and plane portions of a 3D object, which are represented by sharply varying and slowly varying components, respectively, and supplements the localized information by capturing and refining the overall and complementary 3D geometric semantic meanings. In this paper, the geometric disentanglement part and the feature fusion part of GDAnet are improved, and the metric formula is improved in the disentanglement part to make the disentanglement better, while the attention calculation and fusion and splicing operation of the components reduces some of the redundancy information, fully explores the relationship between the points, and incorporates the remote context into the local information, which makes the improved network faster and more accurate in the ModelNet40 and SharpNet datasets are significantly improved.
会议名称：
62nd Annual Meeting of the Association-for-Computational-Linguistics (ACL) / Student Research Workshop (SRW)
会议时间：
AUG 11-16, 2024
会议地点：
Bangkok, THAILAND
会议主办单位：
[Ying, Jiahao;Cao, Yixin] Singapore Management Univ, Singapore, Singapore.^[Xiong, Kai] Harbin Inst Technol, Harbin, Peoples R China.^[He, Yidong;Cui, Long;Liu, Yongbin] Univ South China, Guangzhou, Peoples R China.
摘要：
This study investigates the behaviors of Large Language Models (LLMs) when faced with conflicting prompts versus their internal memory. This will not only help to understand LLMs' decision mechanism but also benefit real-world applications, such as retrieval-augmented generation (RAG). Drawing on cognitive theory, we target the first scenario of decision-making styles where there is no superiority in the conflict and categorize LLMs' preference into dependent, intuitive, and rational/irrational styles. Another scenario of factual robustness considers the correctness of prompt and memory in knowledge-intensive tasks, which can also distinguish if LLMs behave rationally or irrationally in the first scenario. To quantify them, we establish a complete benchmarking framework including a dataset, a robustness evaluation pipeline, and corresponding metrics. Extensive experiments with seven LLMs reveal their varying behaviors. And, with role play intervention, we can change the styles, but different models present distinct adaptivity and upper-bound. One of our key takeaways is to optimize models or the prompts according to the identified style. For instance, RAG models with high role play adaptability may dynamically adjust the interventions according to the quality of retrieval results - being dependent to better leverage informative context; and, being intuitive when the external prompt is noisy. Our dataset can be found at https://github.com/yingjiahao14/KRE.
作者机构：
[Youwei He; Linzhi Liu; Jinliang Luo] School of Mechanical Engineering, University of South China, Hengyang, People's Republic of China
会议名称：
6th Chinese International Turbomachinery Conference (CITC 2024)
会议时间：
2024
会议地点：
Sanya, China
会议论文集名称：
6th Chinese International Turbomachinery Conference (CITC 2024)
摘要：
Recently, multi-fidelity multi-output Kriging have been proposed to take the advantages of both the multi-fidelity and multi-output modelling methods. The predictive performance of multi-fidelity multi-output Kriging model has been evaluated with the modelling efficiency being neglected. In this paper, a multi-fidelity multi-output Kriging model is developed and compared with Kriging, multi-fidelity Kriging, and multi-output Kriging w.r.t. both the modelling accuracy and efficiency. Furthermore, the efficient modelling method for high-dimensional problems is incorporated into the various Kriging models. A 144-dimension engineering problem is established and collected with analytic problems to investigate the modelling accuracy and efficiency. Results show that the multi-output Kriging or multi-fidelity multi-output Kriging model can provide more accurate predictions than the single- or multi-fidelity single-output models on problems with low-dimensional inputs. While, for problems with high-dimensional inputs, the predictive capability of the multi-output Kriging or multi-fidelity multi-output Kriging degenerates significantly with the explosive growth of the time for model construction. With the efficient modelling method, the predictive capability and the modelling efficiency of the multi-output Kriging or multi-fidelity multi-output Kriging model are improved remarkably. However, their performance is still outperformed by single-output multi-fidelity Kriging model throughout the empirical experiments for problems with more than 10 inputs.
作者机构：
[Yankai Zhang; Youwei He] School of Mechanical Engineering, University of South China, Hunan, Hengyang, China
会议名称：
2024 4th International Conference on Energy Engineering and Power Systems (EEPS)
会议时间：
09 August 2024
会议地点：
Hangzhou, China
会议论文集名称：
2024 4th International Conference on Energy Engineering and Power Systems (EEPS)
关键词：
Thermal management;Phase-change direct cooling system;Thermal-fluid-solid coupling simulation
摘要：
Cooling high heat flux electronic devices has always been a significant challenge in the field of electronics. With the continuous increase in laser power density, the demand for effective cooling has become increasingly urgent. This study focuses on a phase-change direct cooling system designed for a 2kW laser. To analyze the heat dissipation performance of the cold plate, the key component of the cooling system, the thermal-fluid-structural coupled simulations were conducted using Fluent software. The simulations covered the temperature distribution of the cold plate under design conditions and phase-change analysis within the flow channels. The highest temperature on the cold plate is 36.69°C. Additionally, simulations under non-design conditions were performed to obtain data such as the maximum temperature of the cold plate. Under 50% load, the highest temperature decreases by 5°C compared to the design conditions.
摘要：
目的:探讨电针预处理尺泽、足三里调控线粒体自噬减轻急性肺损伤(ALI)的作用机制。方法:3月龄雄性大鼠24只随机分为空白组、ALI组和电针预处理组,每组8只。造模方法:参照尾静脉注射LPS制作ALI模型。电针预处理方法:穴位选择双侧"尺泽""足三里"穴,电针波形为疏密波,疏波频率3 Hz,密波频率15 Hz,刺激强度以局部肌肉轻微收缩为度(约1.0 m A),30 min/次,干预5 d。电针预处理组预处理5d后,电针预处理组、ALI组开始造模,空白组正常喂养、不造模。造模6 h后处死动物,苏木精-伊红(hematine-eosine,HE)染色法评估肺组织损伤程度;逆转录聚合酶链反应(reverse transcription-polymerase chain reaction,RT-PCR)和蛋白免疫印迹法(western blotting,WB)分别检测BNIP3L m RNA和蛋白表达。结果:与空白组相比,ALI组病理结果见大鼠肺组织结构明显受损,肺损伤评分增高(P<0.01);BNIP3L m RNA及蛋白表达显著增高(P<0.01)。与ALI组比较,电针预处理组损伤程度明显减轻,损伤评分降低(P<0.01);BNIP3L m RNA及蛋白表达较ALI组降低(P<0.01)。结论:电针预处理改善ALI,可能与调控BNIP3L介导的线粒体自噬有关。
作者机构：
[Guocheng Wang] Yijie Machinery Leasing Co., Ltd of Hengyang, Hengyang, China;[Kuan Tan; Youwei He] College of Mechanical Engineering, University of South China, Hengyang, China
会议名称：
2024 3rd International Conference on Energy, Power and Electrical Technology (ICEPET)
会议时间：
17 May 2024
会议地点：
Chengdu, China
会议论文集名称：
2024 3rd International Conference on Energy, Power and Electrical Technology (ICEPET)
关键词：
the axial fan;surrogate-based optimization;data mining
摘要：
The efficiency of the axial fan has a direct impact on the cooling performance of refrigerators, so this paper proposes a design optimization and data mining method based on the surrogate model to improve the efficiency of the axial fan and understand the optimization principle behind it. In this paper, the optimization algorithm based on the Kriging model is applied to optimize the axial fan. The candidate points for each evaluation are set to 5 to improve the optimization efficiency, and the final axial fan efficiency is improved by 2.58%. Then, through data mining technology of the Analysis of Variance (ANOVA), self-organizing map (SOM), and parallel axis analyses of the design space, it is found that the increase of the thickness of the pressure surface of the third flow surface and the decrease of the thickness of the suction surface can make the axial flow fan obtain better aerodynamic performance.
