دانلود مقاله الگوریتم تشخیص آسیب مبتنی بر یادگیری ماشینی برای شناسایی میزان آسیب در دیوار برشی بتنی
چکیده
1. مقدمه
2. ساختمان های معیار
3. استخراج ویژگی
4. الگوریتم های یادگیری ماشین
5. اجرا
6. نتایج
7. نتیجه گیری
اعلامیه منافع رقابتی
ضمیمه.
منابع
Abstract
1. Introduction
2. Benchmark buildings
3. Feature extraction
4. Machine learning algorithms
5. Implementation
6. Results
7. Conclusion
Declaration of Competing Interest
Appendix.
References
چکیده
این مقاله یک چارچوب تشخیص آسیب مبتنی بر یادگیری ماشین سریع برای شناسایی میزان آسیب ساختمان‌های دیوار برشی بتنی ارائه می‌کند. برای این منظور، یک مطالعه پارامتری برای تعیین کارآمدترین الگوریتم یادگیری ماشین در طبقه‌بندی حالت‌های آسیب ساختمان انجام شد. بر اساس این مطالعه پارامتری، یادگیرنده K-نزدیکترین همسایه (KNN) به دلیل دقت بالاتری که توسط این الگوریتم به دست آمده است، به عنوان مدل پیش‌بینی مرجع انتخاب شد. برای تنظیم فراپارامترهای موثر بر دقت مدل از الگوریتم بهینه سازی بیزی (BO) استفاده شد. کارآمدترین ویژگی‌ها از مجموعه شاخص‌های آسیب از طریق الگوریتم BO برای آموزش مدل انتخاب شدند. سه ساختمان معیار مختلف، شامل ساختمان‌های دیوار برشی بتنی 7، 9، و 13 طبقه، برای ارزیابی استحکام چارچوب پیشنهادی استفاده شد. مجموعه ای از 111 حرکت جفت، که در ابتدا برای پروژه SAC توسعه یافته بود، برای ایجاد یک مجموعه داده تعمیم یافته استفاده شد. این حرکات به طور یکنواخت از 0.05 گرم تا 1.5 گرم مقیاس بندی شدند تا دامنه شدت رویدادها گسترش یابد. تمام سیگنال‌های شتاب با استفاده از سیگنال‌های گاوسی سفید برای شبیه‌سازی شرایط میدان به نویز 10 درصد آلوده شدند. نتایج نشان دهنده کارایی چارچوب پیشنهادی در شناسایی میزان آسیب در عناصر دیوار برشی بتنی ساختمان است. علاوه بر این، یک مطالعه پارامتری برای نشان دادن قابلیت اطمینان دو ویژگی متداول به نام سرعت مطلق تجمعی (CAV) و نسبت انرژی بین پاسخ شتاب و تحریک ورودی، در تعیین وضعیت آسیب دیوارهای برشی تحت حرکات لرزه ای انجام شد.
توجه! این متن ترجمه ماشینی بوده و توسط مترجمین ای ترجمه، ترجمه نشده است.
Abstract
This paper presents a rapid machine learning-based damage detection framework for identifying the damage extent of concrete shear wall buildings. For this purpose, a parametric study was carried out to determine the most efficient machine learning algorithm in classifying the damage states of the building. According to this parametric study, the K-Nearest Neighbor (KNN) learner was selected as the reference prediction model because of the higher accuracy achieved by this algorithm. Bayesian Optimization (BO) algorithm was used to tune the hyperparameters affecting the accuracy of the model. The most efficient attributes were selected from the set of damage indicators through the BO algorithm to train the model. Three different benchmark buildings, including 7-,9-, and 13-story concrete shear wall buildings, were used to evaluate the robustness of the proposed framework. A suite of 111 pair motions, originally developed for the SAC project, were employed to create a generalized dataset. These motions were uniformly scaled from 0.05 g to 1.5 g to expand the intensity range of the events. All the acceleration signals were polluted to 10% noise using white Gaussian signals to simulate the field condition. Results reveal the efficiency of the proposed framework in identifying the extent of damage in concrete shear wall elements of the building. In addition, a parametric study was conducted to illustrate the reliability of two commonly used features, called Cumulative Absolute Velocity (CAV) and the energy ratio between the acceleration response and the input excitation, in determining the damage states of the shear walls under seismic motions.
Introduction
Assessment of structural safety is essential for post-earthquake restoration. Generally, to evaluate the post-earthquake vitality of the exposed structures, a complete visual inspection is required [1]. Coordination and implementation of the manual visual inspection needs several dedicated teams and monetary resources. In this regard, considerable efforts have been carried out to automate the visual inspection process, e.g., image-based visual inspection [2]. However, such an engineering visual inspection is only able to detect the visible defects that occurred in the structures [3]. This means that some serious invisible damages may be left latent during the visual inspection. The process of identifying and tracking the structural damage is known as the Structural Health Monitoring (SHM) [4,5]. In SHM, damage detection is related to the methods developed for identifying the probable existence, severity, and location of the structural damage. Model-based and data-driven methods are two of the most commonly used strategies proposed for damage detection. Model-based methods generally involve a system identification algorithm paired with a finite element analysis to update the structural model [6]. The performance of the model-based approaches directly depends on the accuracy of the information about the physical properties of the under-study structure. In addition, updating the finite element model based on the physical properties of the structure is computationally expensive for large-scale structures, and rapid condition monitoring could be challenging in this condition [7,8]. On the other hand, data-driven methods apply statistical learning algorithms to the vibration data captured from the structure. This method uses learning algorithms to construct a classification or regression learner for predicting structural damage [9, 10]. Tsou and Shen [11] proposed the use of Neural Networks (NNs) for predicting the severity and location of the structural damage. They used the variations in the modal properties of the structure as the damage feature to identify the damage. Worden et al.
Conclusion
This paper presented a rapid algorithm for identifying the severity of local damage in the concrete shear wall buildings. A total number of 1884 nonlinear response history analyses were conducted for each building using the SAC motions. A suite of damage indicators was extracted from the acceleration signals to construct the prediction models. A parametric study was carried out to determine the most efficient learner for classifying the damage states of the buildings. The KNN classifier was selected to construct the predictive models because of its maximum accuracy compared to the other algorithms. The Bayesian optimization algorithm implemented to tune the hyperparameters of the classification learners. The main conclusions derived from the study are as below: