این مقاله انگلیسی ISI در نشریه الزویر در 13 صفحه در سال 2014 منتشر شده و ترجمه آن 31 صفحه میباشد. کیفیت ترجمه این مقاله رایگان – برنزی ⭐️ بوده و به صورت کامل ترجمه شده است.
دانلود رایگان مقاله انگلیسی + خرید ترجمه فارسی | |
عنوان فارسی مقاله: |
ساخت درخت تصمیم پیشرفته بر پایه انتخاب صفات و نمونه داده به منظور تشخیص نقص در ماشین آلات دوار |
عنوان انگلیسی مقاله: |
Improved decision tree construction based on attribute selection and data sampling for fault diagnosis in rotating machines |
|
مشخصات مقاله انگلیسی (PDF) | |
سال انتشار | 2014 |
تعداد صفحات مقاله انگلیسی | 13 صفحه با فرمت pdf |
رشته های مرتبط با این مقاله | مهندسی صنایع و مکانیک |
گرایش های مرتبط با این مقاله | برنامه ریزی و تحلیل سیستم ها، بهینه سازی سیستم ها و طراحی کاربردی |
چاپ شده در مجله (ژورنال) | کاربرد مهندسی هوش مصنوعی – Engineering Applications of Artificial Intelligence |
کلمات کلیدی | ساخت درخت تصمیم، هرس کردن، نمودار پژوهش، انتخاب صفات، نمونه داده |
رفرنس | دارد ✓ |
کد محصول | F1227 |
نشریه | الزویر – Elsevier |
مشخصات و وضعیت ترجمه فارسی این مقاله | |
وضعیت ترجمه | انجام شده و آماده دانلود |
تعداد صفحات ترجمه تایپ شده با فرمت ورد با قابلیت ویرایش | 31 صفحه با فونت 14 B Nazanin |
ترجمه عناوین تصاویر و جداول | ترجمه نشده است ☓ |
ترجمه متون داخل تصاویر | ترجمه نشده است ☓ |
ترجمه متون داخل جداول | ترجمه نشده است ☓ |
درج تصاویر در فایل ترجمه | درج شده است ✓ |
درج جداول در فایل ترجمه | درج شده است ✓ |
کیفیت ترجمه | کیفیت ترجمه این مقاله متوسط میباشد |
فهرست مطالب |
چکیده |
بخشی از ترجمه |
چکیده |
بخشی از مقاله انگلیسی |
Abstract This paper presents a new approach that avoids the over-fitting and complexity problems suffered in the construction of decision trees. Decision trees are an efficient means of building classification models, especially in industrial engineering. In their construction phase, the two main problems are choosing suitable attributes and database components. In the present work, a combination of attribute selection and data sampling is used to overcome these problems. To validate the proposed approach, several experiments are performed on 10 benchmark datasets, and the results are compared with those from classical approaches. Finally, we present an efficient application of the proposed approach in the construction of non-complex decision rules for fault diagnosis problems in rotating machines.. 1- Introduction In the industrial field, the risks of failure and disruption are increasing with the complexity of installed equipment. This phenomenon affects product quality, causes the immediate shutdown of a machine, and undermines the proper functioning of an entire production system. Rotating machines are a major class of mechanical equipment, and need the utmost care and continuous monitoring to ensure optimal operation. Traditionally, vibration analyses and many signal processing techniques have been used to extract useful information for monitoring the operating condition. Khelf et al. (2013) analysed the frequency domain to extract information and diagnose faults. Cepstral analysis has been used to construct a robust gear fault indicator (Badaoui et al., 2004), and a short-time Fourier transform representation was derived (Mosher et al., 2003). Other techniques have also been employed, such as the Wigner–Ville distribution (Baydar and Ball, 2001), continuous wavelet analysis (Kankar et al., 2011), and discrete wavelet analysis (Djebala et al., 2008). Classification algorithms can be used in the construction of condition-monitoring diagnostic systems. For example, neural networks (Chen and Chen, 2011), support vector machines (Deng et al., 2011), and Bayesian classifiers (Yang et al., 2005) have all been applied. However, decision tree techniques are still preferred in engineering applications, because they allow users to easily understand the behaviour of the built models against the abovementioned classifiers. Their use in such applications has been reported in numerous research papers, e.g. Sugumaran and Ramachandran (2007), Zhao and Zhang (2008), Sakthivel et al. (2010), and Sugumaran et al. (2007). The construction of a decision tree (DT) includes growing and pruning stages. In the growing phase, the training data (samples) are repeatedly split into two or more descendant subsets, according to certain split rules, until all instances of each subset wrap the same class (pure) or some stopping criterion has been reached. Generally, this growing phase outputs a large DT that includes the learning examples and considers many uncertainties in the data (particularity, noise and residual variation). Pruning approaches based on heuristics prevent the over-fitting problem by removing all sections of the DT that may be based on noisy and/or erroneous data. This reduces the complexity and size of the DT. The pruning phase can under-prune or over-prune the grown DT. Moreover, many existing heuristics are very challenging (Breiman et al., 1984; Niblett and Bratko, 1987; Quinlan, 1987), but, unfortunately, no single method outperforms the others (Mingers, 1989; Esposito et al., 1997). In terms of growing phase problems, there are two possible solutions: the first reduces DT complexity by reducing the number of learning data, simplifying the decision rules (Piramuthu, 2008). The second solution uses attribute selection to overcome overfitting problems (Yildiz and Alpaydin, 2005; Kohavi and John, 1997). To overcome both the DT size and over-fitting risks, we propose to combine attribute selection and data reduction to construct an Improved Unpruned Decision Tree IUDT . The optimal DT construction (DTC) problem will thus be converted into an exploration of the combinatorial graph research space problem. The key feature of this proposition is to encode each subset of attributes Ai and a samples subset Xj into a couple ðAi; XjÞ. All possible ðAi; XjÞ couples form the research space graph. The results show that the proposed schematic largely improves the tree performance compared to standard pruned DTs, as well as those based solely on attribute selection or data reduction. The rest of the paper is organized as follows: In Section 2, some previous studies on DTC are briefly discussed. Section 3 introduces the main notions used in this work. In Section 4, we describe our approach based on attribute selection and database sampling to outperform conventional DTC. Section 5 reports the experimental results using 10 benchmark datasets. In Section 6, IUDT is applied to the problem of fault diagnosis in rotating machines. Finally, Section 7 concludes the study. |