• <tr id="yyy80"></tr>
  • <sup id="yyy80"></sup>
  • <tfoot id="yyy80"><noscript id="yyy80"></noscript></tfoot>
  • 99热精品在线国产_美女午夜性视频免费_国产精品国产高清国产av_av欧美777_自拍偷自拍亚洲精品老妇_亚洲熟女精品中文字幕_www日本黄色视频网_国产精品野战在线观看 ?

    Gross Error Detection and Identification Based on Parameter Estimation for Dynamic Systems*

    2009-05-14 03:04:50JIANGChunyang姜春陽QIUTong邱彤ZHAOJinsong趙勁松andCHENBingzhen陳丙珍
    關(guān)鍵詞:勁松

    JIANG Chunyang (姜春陽), QIU Tong (邱彤)**, ZHAO Jinsong (趙勁松) and CHEN Bingzhen (陳丙珍)

    ?

    Gross Error Detection and Identification Based on Parameter Estimation for Dynamic Systems*

    JIANG Chunyang (姜春陽), QIU Tong (邱彤)**, ZHAO Jinsong (趙勁松) and CHEN Bingzhen (陳丙珍)

    Department of Chemical Engineering, Tsinghua University, Beijing 100084, China

    The detection and identification of gross errors, especially measurement bias, plays a vital role in data reconciliation for nonlinear dynamic systems. Although parameter estimation method has been proved to be a powerful tool for bias identification, without a reliable and efficient bias detection strategy, the method is limited in efficiency and cannot be applied widely. In this paper, a new bias detection strategy is constructed to detect the presence of measurement bias and its occurrence time. With the help of this strategy, the number of parameters to be estimated is greatly reduced, and sequential detections and iterations are also avoided. In addition, the number of decision variables of the optimization model is reduced, through which the influence of the parameters estimated is reduced. By incorporating the strategy into the parameter estimation model, a new methodology named IPEBD (Improved Parameter Estimation method with Bias Detection strategy) is constructed. Simulation studies on a continuous stirred tank reactor (CSTR) and the Tennessee Eastman (TE) problem show that IPEBD is efficient for eliminating random errors, measurement biases and outliers contained in dynamic process data.

    gross error detection, data reconciliation, parameter estimation

    1 Introduction

    In the chemical industry, accurate and real-time process data is the foundation of on-line fault diagnosis, real-time monitoring, advanced control and dynamic optimization. Data reconciliation can effectively eliminate errors, typically for the normally distributed random errors, so that the quantities derived from the process data obey the first principles. In fact the industrial measurements are, however, usually contaminated by gross errors, which can be subdivided into measurement bias and outliers. Measurement bias refers to the situation in which measurement devices provide consistently erroneous values, either high or low. Outliers can be considered as abnormal behavior of the measurements, process peaks or unmeasured disturbances [1]. With the presence of gross errors, the reconciled measurements may greatly deviate from their true values. It’s quite necessary to carry out gross error detection and identification, which can locate and eliminate gross errors before data reconciliation.

    For outliers there have been relatively effective methods, which can be divided into two categories. One is related to the process model and utilized to identify outliers and eliminate or reduce their influences on the performance of the conventional data reconciliation by making use of cluster analysis [2], quantile probability plots [3] or robust estimation [4-6]. The other one has no relation with the process model but utilizes the multi-scale property of wavelets for data processing and outlier identification [7].

    For measurement bias detection and identification, current methods are limited in accuracy and efficiency, and cannot meet the demand for on-line application.

    Narasimhan and Mah extended the generalized likelihood ratio (GLR) method for identifying gross errors and used a simple chi-square test to estimate the occurrence time of the gross errors [8]. Kao and Tamhane made use of the cumulative sum (CUSUM) test to detect the presence of gross errors and applied GLR method to identify and estimate their magnitudes [9]. Rollins. presented unbiased estimation technique (UBET) to identify measurement bias and process leaks [10, 11]. Devanathan. proposed a gross error identification approach based on the detection of bias changes [12]. Bagajewicz and Jiang presented a dynamic integral measurement test (DIMT) to identify hold-up measurements as suspects of gross errors [13]. But the above methods are all designed only for linear dynamic systems, which usually operate around a pseudo steady state.

    Karjala and Himmelblau proposed an approach to identify measurement bias or measurement errors, which involved the usage of recurrent neural networks (RNN) and the extended Kalman filter (EKF) [14]. However, it is probable for RNN to encounter combinatory explosion problems. Vachhani. viewed the biased parameter identification as a diagnostic problem and brought in methods from fault diagnosis literature to improve the performance [15]. However, this approach is not adoptable for the situation in which multiple gross errors may exist.

    Liebman. developed nonlinear dynamic data reconciliation (NDDR) method, which estimated measurement bias as a parameter [16]. This method could be applied to strongly nonlinear dynamic systems with multiple gross errors present. However, no strategy is included in this method to detect the presence of gross errors. Instead, each variable of the process is assumed to be biased and added a parameter for estimation, which increases the complexity of NDDR and limits its efficiency in on-line applications.

    Based on Liebman’s parameter estimation method, McBrayer and Edgar proposed a new approach which examined the residuals and carried out summation test and regression test for bias detection [17]. This approach can serially identify and compensate measurement bias and avoid unnecessary parameter estimation, but involves too many unnecessary reconciliations. Besides, the base case required in this approach is ideal and difficult to acquire. Abu-el-zeet. simplified McBrayer’s strategy and presented a new methodology, which combined bias and outlier identification in dynamic data reconciliation [18]. Although base case is removed and some unnecessary calculations are avoided, it is still a serial detection and compensation procedure and not favorable for online applications. Kong. proposed the concept of gross error identifiability and derived the identifiable condition of a process. They also addressed an identification approach on the basis of parameter estimation [19]. But all these methods did not overcome the weakness of parameter estimation and could not meet the needs for high efficiency.

    In this article, based on parameter estimation, a new bias detection strategy is proposed, which can simply detect the presence of bias and its occurring and ending time. As a result, the number of estimated parameters is remarkably reduced and the performance of parameter estimation is greatly improved. Furthermore, the strategy is combined with present outlier identification method. It is shown through simulations on the continuous stirred tank reactor (CSTR) test case and Tennessee Eastman (TE) Challenge Process that the integrated approach simultaneously and efficiently identifies bias and outliers for on-line data reconciliation.

    2 The original parameter estimation method

    In the absence of gross errors the measurement model can be written as:

    The data reconciliation problem can be stated as a constrained nonlinear programming problem [16]:

    The model above has been proved powerful for nonlinear dynamic systems with random errors [16]. However, if the assumption that systematic bias is absent is not hold, the measurement model can be written as:

    The original parameter estimation method is applicable and effective for nonlinear dynamic process with multiple gross errors present. But it does not include any bias detection strategy so as to be limited in efficiency, especially when only a small proportion of the measured variables contains gross errors, and present improved strategies bring in other iteration problems while reducing the number of the estimated parameters [17, 18], as mentioned above.

    3 The improved parameter estimation method with bias detection strategy

    3.1 Bias detection criterion

    The CSTR test case published by Liebman. [16] was studied and the reconciled results which were acquired using model (2) were analyzed. It shows that the reconciled results differ a lot with or without the presence of bias. If the measurements only contain random errors in a moving horizon, there will be little deviation between the mean of estimated values and the mean of measured values [see Fig. 1 (a)]. If bias exists, the reconciled results will smear a lot from the measurements [see Fig. 1 (b)].

    Figure 1 Reconciled results (a) without measurement bias and (b) with measurement bias ——?true values; ×?measured values;●?estimated values

    If the smearing could be detected in time using a simple criterion, measurement bias would be identified, including the contaminated variable and the occurring and ending time of the bias. However, it is not easy to tell whether the current deviation is caused by bias or not, because measurement bias is consistently erroneous, as is mentioned above. As a result, detection should be carried out on the basis of statistical analysis of measurements in a moving horizon. The straightforward idea is to analyze the difference between the mean of estimated values and the mean of measured values. Thus, a new criterion for bias detection is proposed as the following:

    3.2 Improved data reconciliation model

    Based on the criterion, the original parameter estimation model, as Eq. (5), is amended as the following:

    whereis the aggregate of subscript of the variables which might contain bias,is a vector to record the bias detection results.

    It should be noted that a positive detection of criterion (6) does not mean that the current measured variable is positively biased. It should be determined eventually by whether the estimated value ofδis over the threshold.

    3.3 Bias detection and identification strategy

    Bias detection and identification strategy of the improved parameter estimation method is quite straightforward as the following:

    1) For the first moving horizon, there is no historical measured data for reference. It is assumed that all the variables are not free of bias, sois set to unit vector. Then data reconciliation is carried out using model (7).

    2) Save the reconciled results of the former moving horizon, and check whether criterion (6) holds. Record the detection result in the vector.

    3) Reconcile the measurements in the current horizon. If the bias estimation of some variable is over the threshold, the presence of bias is insured and recorded.

    4) Return to Step 2).

    In this strategy unnecessary parameter estimations and iterations are avoided, so the complexity of data reconciliation is greatly reduced. Besides, the strategy is convenient to be integrated with the original parameter estimation method. Therefore, the improved parameter estimation method with bias detection strategy (IPEBD) can eliminate measurement bias and random errors simultaneously and efficiently.

    4 Simulation case study

    The IPEBD approach was successfully applied to the CSTR test case and the TE Challenge Processsimulation, which were both based on Matlab 7.3.0 platform on a PC (P4 3.0G, 1G ram).

    4.1 CSTR test case

    4.1.1

    For comparison, IPEBD was tested on the same CSTR test case used by Liebman[16]. The scaled dynamic model can be written as:

    4.1.2

    4.1.3

    Bias detection result is shown in Fig. 2. The vertical axis represents the judgment of the detection strategy, where 1 stands for positive detection and 0 for negative detection. As a result of the assumption that all the variables in the first moving horizon are not free of bias, it is not detection fault that the first stage of all the plots is 1. Since the strategy uses reconciled results in the former horizon for bias detection, there is a short delay in detection results, which is about 0.5-1 times the size of a horizon. There is only one detection fault at time steps 150-180. It is mainly caused by the step change in the input concentration, which is reasonable and acceptable.

    The complexity and time consumption are greatly reduced by narrowing the search region. Table 1 shows that the efficiency of IPEBD is greatly improved whether measurement bias present or not. For this test case, in which approximate 50% of the measured data is contaminated by bias, the reconciliation efficiency is increased by about 30%. Since most measured data is free of bias in chemical industry, the efficiency increased will be up to 40%.

    4.2 TE challenge process

    The advantages of IPEBD have been proved in the CSTR test case where the measured data is contaminated by random errors and measurement bias. But real industrial process is usually more complex than a single CSTR, and the measured data usually contains outliers. In this section, IPEBD will be tested in a complex process named TE Challenge Process with random errors, bias and outliers present.

    Table 1 Comparison of average time consumption

    4.2.1

    The TE Challenge Process—hereafter called the TE problem—was proposed by Downs and Vogel at an AIChE meeting in 1990 [20]. It involves coordination of four unit operations: an exothermic two-phase reactor, a flash separator, a reboiled stripper and a recycle compressor. The process produces two products from four exothermic, irreversible reactions. There are a total of eight components labeled as A-H, with component B an inert, components G and H the primary products, and component F a byproduct. The chemical reactions provided are written as:

    4.2.2

    Downs and Vogel omit details of the kinetics, phase equilibria and process dynamics, and no plant model is provided. Fortunately, some mechanistic nonlinear model has been developed and validated with data taken from the original Downs and Vogel work and their accompanying Fortran programs [21-23]. In this article, all the simulations of the TE problem are based on Ricker’s work [21, 22]. The mechanistic nonlinear model includes 76 algebraic equations, 26 differential equations and a total of 112 variables including 67 measured variables. More details could be found in the Refs. [21, 22].

    The TE problem is tested through computer simulation with a sampled time of 1 min, and each moving horizon includes 7 measurements. Random errors in normal distribution, outliers of magnitude 5-10and measurement bias of magnitude 4-6are introduced randomly.

    Since outliers are present, the IPEBD approach needs to be improved. With the object function replaced by a robust estimator [24, 25], IPEBD will be capable to deal with random errors, measurement bias and outliers simultaneously. The robust estimator adopted in the simulation is first published by the author, which is stated as the following [26]:

    where,,are parameters to be tuned.

    4.2.3

    Since there are too many measured variables in the TE problem, only part of the reconciled results will be shown in the following text.

    As a result of the introduction of the robust estimator, outliers are successfully identified and eliminated by IPEBD as shown in Fig. 5.

    Figure 5 Results of outlier detection and identification——?true values; × measured values; ●?estimated values; ○?outlier errors

    There are totally 4 measured variables contaminated by measurement bias, all of which are successfully detected by IPEBD (see Fig. 6). Bias detection only decides which variable might contain bias and the final judgments are made by the results of parameter estimation. As the CSTR test case, there are a short delay and some faults in bias detection, but bias identification is not affected. All the 4 biases are identified and reconciled eventually as shown in Fig. 7.

    Because of the accurate bias detection of IPEBD, most parameters in the original parameter estimation method are avoided, which greatly improves the efficiency of data reconciliation. With the help of bias detection strategy, the average time consumption is reduced by about 30%. Since most measurements are free of bias in the industrial process, time consumption will be further reduced by about 38% (see Table 2). Besides, for the measurements in a single time step, the maximum time consumption of IPEBD is less than the sampled time step, typically 60 seconds in this case, which means IPEBD can be used for online application.

    Table 2 Comparison of time consumption

    Figure 7 Results of bias identification——?true values; × measured values; ●?estimated values

    5 Conclusions

    Parameter estimation method has been proved to be a powerful tool for bias identification. However, without a reliable and efficient bias detection strategy, data reconciliation based on parameter estimation is limited in efficiency and cannot be applied widely. Against the shortcomings of parameter estimation method, a new bias detection strategy is successfully constructed and has been shown to work efficiently in detecting the presence of measurement bias and recognizing its occurring time. With the strategy present, the number of parameters to be estimated is greatly reduced, and serially detection and iterations are avoided. Besides, the number of decision variables of optimization is reduced, through which influences between parameters estimated are decreased and the accuracy of parameter estimation is improved. By incorporating the strategy into the data reconciliation model, a new methodology named IPEBD is constructed based on parameter estimation method, and it works pretty well in CSTR and TE testing cases.

    NOMENCLATURE

    feasible region of the original method

    feasible region of the improved method

    aggregate of subscript of the variables which might contain bias

    6reactor feed, kmol·h-1

    7reactor outlet, kmol·h-1

    differential equation constraints

    algebraic inequality constraints

    width of the history horizon

    algebraic equality constraints

    number of the measured variables

    vector of time

    ccurrent time

    Δsample time step

    vector of the unmeasured variables

    variance-covariance matrix of the measurements

    vector of the reconciled measurements

    r(G) G in reactor, % (by mol)

    5(E) E in top outlet of stripper, % (by mol)

    6(H) H in reactor feed, % (by mol)

    7(E) E in reactor outlet, % (by mol)

    8(B) B in recycle stream, % (by mol)

    8(F) F in recycle stream, % (by mol)

    vector to record the bias detection results

    vector of the measurement bias

    δtheth assumed parameter to be estimated

    vector of the random errors

    ζthreshold of the criterion

    penalty weight for theth variable

    1 Yuan, Y.G., Li, H.S., Data Rectification of Process Measurements, China PetroChemical Press, Beijing (1996).

    2 Chen, J., Romagnoli, J.A., “A strategy for simultaneous dynamic data reconciliation and outlier detection”,..., 24 (4/5), 559-562 (1998).

    3 Chen, J., Bandoni, A., Romagnoli, J.A., “Outlier detection in process plant data”,..., 22 (4/5), 641-646 (1998).

    4 Zhou, L.K., Su, H.Y., Chu, J., “A new method to solve robust data reconciliation in nonlinear process”,...., 14 (3), 357-363 (2006).

    5 ?zyurt, D.B., Pike, R.W., “Theory and practice of simultaneous data reconciliation and gross error detection for chemical processes”,..., 28, 381-402 (2004).

    6 Faber, R., Li, B., Li, P., W?zny, G., “Data reconciliation for real-time optimization of an industrial coke-oven-gas purification process”,, 14, 1121-1134 (2006).

    7 Jiang, T.W., Chen, B.Z., He, X.R., “Industrial application of wavelet transform to the on-line prediction of side draw qualities of crude unit”,..., 24, 507-512 (2000).

    8 Narasimhan, S., Mah, R.S.H., “Generalized likelihood ratios for gross error identification in dynamic processes”,., 34, 1321-1331 (1988).

    9 Kao, C.S., Tamhane, A.C., Mah, R.S.H., “Gross error detection in serially correlated process data (2) Dynamic systems”,...., 31, 254-262 (1992).

    10 Rollins, D.K., Davis, J.F., “Unbiased estimation of gross errors in process measurements”,., 38 (4), 563-572 (1992).

    11 Rollins, D.K., Devanathan, S., “Unbiased estimation in dynamic data reconciliation”,., 39, 1330-1334 (1993).

    12 Devanathan, S., Rollins, D.K., Vardeman, S.B., “A new approach for improved identification of measurement bias”,..., 24, 2755-2764 2000).

    13 Bagajewicz, M. J., Jiang, Q., “Gross error modeling and detection in plant linear dynamic reconciliation”,..., 22, 1789-1809 (1997).

    14 Karjala, T.W., Himmelblau, D.M., “Dynamic rectification of datarecurrent neural nets and extended Kalman filter”,., 42, 2225-2239 (1996).

    15 Vachhani, P., Rengaswamy, R., Venkatasubramanian, V., “A framework for integrating diagnositic knowledge with nonlinear optimization for data reconciliation and parameter estimation in dynamic system”,..., 56, 2133-2148 (2001).

    16 Liebman, M.J., Edgar, T.F., Lasdon, L.S., “Efficient data reconciliation and estimation for dynamic processes using nonlinear programming techniques”,..., 16 (10/11), 963-986 (1992).

    17 McBrayer, K.F., Edgar, T.F., “Bias detection and estimation in dynamic data reconciliation”,., 5 (4), 285-289 (1995).

    18 Abu-el-zeet, Z.H., Becerra, V.M., Roberts, P.D., “Combined bias and outlier identification in dynamic data reconciliation”,..., 16, 921-935 (2002).

    19 Kong, M.F., Chen, B.Z., He, X.R., Hu, S.Y., “Gross error identification for dynamic system”,..., 29, 191-197 (2004).

    20 Downs, J.J., Vogel, E.F., “A plant-wide industrial process control problem”,..., 17 (3), 245-255 (1993).

    21 Ricker, N.L., “Optimal steady-state operation of the Tennessee Eastman challenge process”,..., 19 (9), 949-959 (1995).

    22 Ricker, N.L., Lee, J.H., “Nonlinear modeling and state estimation for the Tennessee Eastman challenge process”,..., 19 (9), 983-1005 (1995).

    23 Tian, Z.H., Hoo, K.A., “Multiple model-based control of the Tennessee- Eastman process”,...., 44, 3187-3202 (2005).

    24 Huber, P.J., Robust Statistics, Wiley, New York (1981).

    25 Hampel, F.R., Ronchetti, E.M., Rousseeuw, P.J., Stahel, W.A., Robust Statics: the Approach Based on Influence Functions, John Wiley & Sons, New York (1985).

    26 Jiang, C.Y., Qiu, T., Chen, B.Z., Zhao, J.S., “An improved robust dynamic data reconciliation approach”,..., 24 (10), 1297-1301 (2007).

    2008-09-23,

    2009-03-11.

    the National High Technology Research and Development Program of China (2006AA04Z176).

    ** To whom correspondence should be addressed. E-mail: qiutong@tsinghua.edu.cn

    猜你喜歡
    勁松
    顧勁松
    藝術(shù)家(2024年2期)2024-04-15 08:19:20
    一種分步組合的部分模糊度固定方法
    莊勁松美術(shù)作品
    “貓”忍不住
    故事會(2022年7期)2022-04-03 19:26:31
    貓忍不住
    小小說月刊(2022年1期)2022-02-07 10:44:22
    王勁松:行走在邊緣的一顆恒星
    時代郵刊(2019年20期)2019-07-30 08:05:38
    Simulation and experimental research of digital valve control servo system based on CMAC-PID control method①
    Isolated attosecond pulse generation with few-cycle two-color counter-rotating circularly polarized laser pulses?
    余勁松
    閱讀理解精練
    国产麻豆69| 国产欧美日韩精品亚洲av| 在线看a的网站| 黄色片一级片一级黄色片| 多毛熟女@视频| 国产成+人综合+亚洲专区| 丝瓜视频免费看黄片| 国产在线视频一区二区| 免费在线观看黄色视频的| 国产精品一区二区在线不卡| 一级a爱视频在线免费观看| 美女午夜性视频免费| videosex国产| 日本撒尿小便嘘嘘汇集6| xxxhd国产人妻xxx| 亚洲国产成人一精品久久久| 女人久久www免费人成看片| 亚洲色图 男人天堂 中文字幕| 欧美日韩亚洲国产一区二区在线观看 | 夜夜夜夜夜久久久久| 精品国内亚洲2022精品成人 | av线在线观看网站| 女人久久www免费人成看片| 69精品国产乱码久久久| 99国产精品99久久久久| 精品亚洲乱码少妇综合久久| 中文字幕另类日韩欧美亚洲嫩草| 狂野欧美激情性xxxx| 午夜91福利影院| 欧美+亚洲+日韩+国产| 国产在线视频一区二区| 亚洲国产av新网站| 一本久久精品| 啦啦啦免费观看视频1| 国产伦人伦偷精品视频| 2018国产大陆天天弄谢| 老熟妇乱子伦视频在线观看| 美女午夜性视频免费| 亚洲av第一区精品v没综合| 久久久久久亚洲精品国产蜜桃av| 国产精品熟女久久久久浪| 波多野结衣av一区二区av| 国产精品av久久久久免费| 日本wwww免费看| 免费一级毛片在线播放高清视频 | 性少妇av在线| 超碰成人久久| 成年动漫av网址| 99国产精品一区二区蜜桃av | 夜夜夜夜夜久久久久| 色尼玛亚洲综合影院| 日本av手机在线免费观看| 热re99久久国产66热| 两性午夜刺激爽爽歪歪视频在线观看 | 久久久久久久大尺度免费视频| 在线观看人妻少妇| 国产av一区二区精品久久| 精品人妻熟女毛片av久久网站| 俄罗斯特黄特色一大片| 久久精品成人免费网站| 午夜老司机福利片| 性少妇av在线| 99热国产这里只有精品6| 日韩成人在线观看一区二区三区| 久久婷婷成人综合色麻豆| 亚洲国产欧美日韩在线播放| 国产又色又爽无遮挡免费看| 亚洲成av片中文字幕在线观看| 在线看a的网站| 色婷婷av一区二区三区视频| 2018国产大陆天天弄谢| 久久国产精品人妻蜜桃| 热re99久久国产66热| 香蕉久久夜色| 岛国在线观看网站| 国产精品99久久99久久久不卡| 黑人巨大精品欧美一区二区mp4| 黑人巨大精品欧美一区二区蜜桃| 久久精品国产亚洲av香蕉五月 | 久久久欧美国产精品| 久久国产精品大桥未久av| 一级,二级,三级黄色视频| 日本一区二区免费在线视频| 亚洲avbb在线观看| 啦啦啦在线免费观看视频4| 岛国在线观看网站| 嫩草影视91久久| 免费在线观看视频国产中文字幕亚洲| 国产成人欧美在线观看 | 久久免费观看电影| 黑丝袜美女国产一区| 中文字幕人妻丝袜一区二区| 亚洲 国产 在线| 国产成人系列免费观看| 日韩欧美一区视频在线观看| 性高湖久久久久久久久免费观看| 中国美女看黄片| 国产成人免费观看mmmm| 91大片在线观看| 一级毛片精品| 亚洲一区中文字幕在线| 在线观看www视频免费| 亚洲美女黄片视频| 日韩成人在线观看一区二区三区| 欧美老熟妇乱子伦牲交| 亚洲av电影在线进入| 日韩人妻精品一区2区三区| 国产欧美日韩综合在线一区二区| 夫妻午夜视频| 91九色精品人成在线观看| 黄网站色视频无遮挡免费观看| 久久久国产欧美日韩av| www.999成人在线观看| 久久亚洲真实| 操美女的视频在线观看| 国产aⅴ精品一区二区三区波| 欧美人与性动交α欧美软件| 激情视频va一区二区三区| 十八禁高潮呻吟视频| 日本欧美视频一区| 久久久国产精品麻豆| 国产成人影院久久av| 18禁裸乳无遮挡动漫免费视频| 欧美日韩国产mv在线观看视频| 亚洲 国产 在线| 人人妻人人澡人人爽人人夜夜| 欧美日韩一级在线毛片| av国产精品久久久久影院| 久久婷婷成人综合色麻豆| 在线天堂中文资源库| 亚洲人成电影免费在线| 高潮久久久久久久久久久不卡| 亚洲三区欧美一区| 亚洲男人天堂网一区| 最近最新免费中文字幕在线| 国产区一区二久久| 我的亚洲天堂| 国产精品久久久久久精品电影小说| 日韩欧美一区二区三区在线观看 | 老汉色av国产亚洲站长工具| 精品一区二区三区四区五区乱码| 一二三四在线观看免费中文在| 欧美黑人欧美精品刺激| 在线av久久热| 成人特级黄色片久久久久久久 | 成人18禁高潮啪啪吃奶动态图| 国产成人欧美| 69av精品久久久久久 | 欧美激情极品国产一区二区三区| 婷婷丁香在线五月| 午夜两性在线视频| 黄色视频,在线免费观看| 亚洲国产欧美在线一区| 老司机靠b影院| 久久国产精品大桥未久av| 色综合欧美亚洲国产小说| 一级黄色大片毛片| 天堂动漫精品| 2018国产大陆天天弄谢| 女人被躁到高潮嗷嗷叫费观| 啪啪无遮挡十八禁网站| 中文字幕高清在线视频| 国产欧美日韩一区二区三| 国产熟女午夜一区二区三区| 色婷婷久久久亚洲欧美| 99在线人妻在线中文字幕 | 无限看片的www在线观看| 在线观看免费视频网站a站| 丝袜人妻中文字幕| 亚洲精品av麻豆狂野| 欧美av亚洲av综合av国产av| 香蕉国产在线看| 热re99久久国产66热| 巨乳人妻的诱惑在线观看| 高清毛片免费观看视频网站 | 极品人妻少妇av视频| 曰老女人黄片| 欧美激情极品国产一区二区三区| www.精华液| 国产精品久久久久久精品古装| 午夜福利在线免费观看网站| 伦理电影免费视频| 极品少妇高潮喷水抽搐| 欧美亚洲 丝袜 人妻 在线| 啦啦啦在线免费观看视频4| 欧美精品高潮呻吟av久久| 欧美日韩成人在线一区二区| 国产成人精品久久二区二区免费| 久久天堂一区二区三区四区| 成人国语在线视频| 欧美成人午夜精品| 成人亚洲精品一区在线观看| 亚洲国产中文字幕在线视频| 欧美激情高清一区二区三区| 国产在线精品亚洲第一网站| 日本欧美视频一区| 国产精品美女特级片免费视频播放器 | 日韩欧美一区二区三区在线观看 | 免费久久久久久久精品成人欧美视频| 国产深夜福利视频在线观看| av超薄肉色丝袜交足视频| av国产精品久久久久影院| 在线观看一区二区三区激情| 免费在线观看黄色视频的| www日本在线高清视频| 91精品三级在线观看| 中文字幕色久视频| 亚洲 国产 在线| 免费看a级黄色片| 久久中文字幕一级| 亚洲色图综合在线观看| 国产一卡二卡三卡精品| av又黄又爽大尺度在线免费看| 不卡一级毛片| 色精品久久人妻99蜜桃| 一本久久精品| 人人妻人人爽人人添夜夜欢视频| 精品久久蜜臀av无| 欧美+亚洲+日韩+国产| 老鸭窝网址在线观看| 国产av国产精品国产| 国产不卡一卡二| 免费观看a级毛片全部| 久久精品亚洲精品国产色婷小说| 欧美激情极品国产一区二区三区| 欧美大码av| 99国产极品粉嫩在线观看| 中国美女看黄片| 女人爽到高潮嗷嗷叫在线视频| kizo精华| 婷婷丁香在线五月| 97在线人人人人妻| 日韩有码中文字幕| 又紧又爽又黄一区二区| 免费女性裸体啪啪无遮挡网站| 青青草视频在线视频观看| tube8黄色片| 两性夫妻黄色片| 欧美精品啪啪一区二区三区| 午夜福利在线免费观看网站| 精品国内亚洲2022精品成人 | 蜜桃在线观看..| 亚洲伊人久久精品综合| 18禁观看日本| 成年人黄色毛片网站| 欧美 亚洲 国产 日韩一| 美女高潮喷水抽搐中文字幕| 欧美国产精品一级二级三级| 亚洲国产毛片av蜜桃av| 亚洲国产av新网站| 少妇粗大呻吟视频| 久久精品亚洲av国产电影网| 侵犯人妻中文字幕一二三四区| 国产精品av久久久久免费| 青青草视频在线视频观看| 日韩 欧美 亚洲 中文字幕| 99香蕉大伊视频| 水蜜桃什么品种好| videosex国产| 老司机午夜十八禁免费视频| 国产成+人综合+亚洲专区| 免费日韩欧美在线观看| 国产一区二区 视频在线| 一边摸一边做爽爽视频免费| 天天影视国产精品| 国产av精品麻豆| 亚洲人成伊人成综合网2020| 9色porny在线观看| 亚洲五月色婷婷综合| 久久精品aⅴ一区二区三区四区| av网站免费在线观看视频| svipshipincom国产片| 亚洲一区中文字幕在线| 少妇被粗大的猛进出69影院| 久久精品成人免费网站| 91大片在线观看| 日韩免费av在线播放| www.999成人在线观看| 男人操女人黄网站| 久久久久久久久免费视频了| 伊人久久大香线蕉亚洲五| 中文字幕人妻丝袜一区二区| 欧美日韩一级在线毛片| 9色porny在线观看| 国产精品亚洲一级av第二区| 欧美午夜高清在线| 欧美人与性动交α欧美精品济南到| 日本av手机在线免费观看| 国产在线一区二区三区精| 久久中文字幕一级| 精品午夜福利视频在线观看一区 | 日韩中文字幕欧美一区二区| 精品人妻熟女毛片av久久网站| 三级毛片av免费| 成人永久免费在线观看视频 | 丰满迷人的少妇在线观看| 精品人妻1区二区| 亚洲国产精品一区二区三区在线| www.精华液| av电影中文网址| 精品午夜福利视频在线观看一区 | 国产精品免费视频内射| 国产精品国产av在线观看| 一进一出抽搐动态| 高潮久久久久久久久久久不卡| 99热国产这里只有精品6| 久久影院123| 国产精品免费视频内射| 考比视频在线观看| 亚洲av国产av综合av卡| 91老司机精品| www.自偷自拍.com| 国产男女超爽视频在线观看| 欧美激情高清一区二区三区| 精品午夜福利视频在线观看一区 | 午夜91福利影院| 女人高潮潮喷娇喘18禁视频| 亚洲精品中文字幕一二三四区 | 啦啦啦免费观看视频1| 黄片大片在线免费观看| 91精品国产国语对白视频| 欧美av亚洲av综合av国产av| 中国美女看黄片| 一级毛片女人18水好多| 日韩成人在线观看一区二区三区| 亚洲黑人精品在线| 黄色丝袜av网址大全| 18禁观看日本| 亚洲黑人精品在线| 9色porny在线观看| 涩涩av久久男人的天堂| 欧美成狂野欧美在线观看| 91九色精品人成在线观看| 久久亚洲精品不卡| bbb黄色大片| 欧美日韩av久久| 深夜精品福利| 狂野欧美激情性xxxx| 老汉色∧v一级毛片| 中文字幕人妻丝袜制服| e午夜精品久久久久久久| 国产在视频线精品| 精品人妻在线不人妻| 两人在一起打扑克的视频| 成年版毛片免费区| av欧美777| 久久午夜亚洲精品久久| 一二三四社区在线视频社区8| 成人18禁高潮啪啪吃奶动态图| www日本在线高清视频| 少妇被粗大的猛进出69影院| 欧美日韩视频精品一区| 天天躁日日躁夜夜躁夜夜| 久久久精品免费免费高清| 女警被强在线播放| 久久精品亚洲精品国产色婷小说| tube8黄色片| 欧美精品一区二区大全| 国产高清videossex| 欧美 日韩 精品 国产| 免费观看av网站的网址| 黑人猛操日本美女一级片| 亚洲国产成人一精品久久久| 啪啪无遮挡十八禁网站| 国产成人av激情在线播放| 国产深夜福利视频在线观看| 国产成人精品无人区| 黄色怎么调成土黄色| 亚洲欧美日韩另类电影网站| 五月天丁香电影| 精品一品国产午夜福利视频| 后天国语完整版免费观看| 香蕉久久夜色| 日韩成人在线观看一区二区三区| 欧美精品一区二区大全| 青青草视频在线视频观看| 三上悠亚av全集在线观看| 亚洲国产精品一区二区三区在线| 亚洲欧洲精品一区二区精品久久久| 999久久久国产精品视频| 国产精品一区二区在线观看99| 美女扒开内裤让男人捅视频| 97在线人人人人妻| 99riav亚洲国产免费| 亚洲午夜理论影院| 极品少妇高潮喷水抽搐| 欧美日韩成人在线一区二区| 岛国在线观看网站| 黄片小视频在线播放| 日韩大片免费观看网站| 12—13女人毛片做爰片一| 国产免费福利视频在线观看| 黄色a级毛片大全视频| 极品少妇高潮喷水抽搐| 如日韩欧美国产精品一区二区三区| 色播在线永久视频| 国产不卡一卡二| av福利片在线| www日本在线高清视频| 99精品欧美一区二区三区四区| 日本撒尿小便嘘嘘汇集6| 在线观看免费视频日本深夜| 欧美日韩中文字幕国产精品一区二区三区 | 国产在线观看jvid| 国产精品偷伦视频观看了| 窝窝影院91人妻| 考比视频在线观看| 国产成人欧美在线观看 | 国产野战对白在线观看| 国产黄频视频在线观看| 757午夜福利合集在线观看| 国产欧美亚洲国产| 欧美国产精品一级二级三级| 男人舔女人的私密视频| 久久精品91无色码中文字幕| 怎么达到女性高潮| 久久狼人影院| 亚洲精品乱久久久久久| 亚洲欧美日韩另类电影网站| 中文字幕另类日韩欧美亚洲嫩草| 在线观看免费高清a一片| 亚洲精品一二三| 亚洲成国产人片在线观看| 亚洲欧美激情在线| 亚洲九九香蕉| 淫妇啪啪啪对白视频| 日韩 欧美 亚洲 中文字幕| 国产欧美日韩一区二区精品| 国产精品99久久99久久久不卡| 亚洲欧美一区二区三区黑人| 亚洲国产中文字幕在线视频| 午夜精品国产一区二区电影| 国产成人精品无人区| 色综合婷婷激情| 国产午夜精品久久久久久| 国产精品免费大片| 欧美 日韩 精品 国产| 69av精品久久久久久 | videosex国产| 老司机福利观看| 成人国产一区最新在线观看| 久久天躁狠狠躁夜夜2o2o| 老司机深夜福利视频在线观看| 午夜福利视频在线观看免费| 两性夫妻黄色片| 少妇 在线观看| 久久久久久久精品吃奶| 成人精品一区二区免费| 亚洲男人天堂网一区| 欧美国产精品一级二级三级| 亚洲美女黄片视频| 精品国产一区二区久久| 一边摸一边做爽爽视频免费| 亚洲国产精品一区二区三区在线| 色综合婷婷激情| 国产欧美日韩一区二区三| 亚洲综合色网址| 精品亚洲成a人片在线观看| 69精品国产乱码久久久| 久久婷婷成人综合色麻豆| 国产色视频综合| 日韩一区二区三区影片| 男女边摸边吃奶| 一边摸一边抽搐一进一出视频| 亚洲av电影在线进入| 国产1区2区3区精品| www.999成人在线观看| 亚洲欧洲精品一区二区精品久久久| 欧美精品高潮呻吟av久久| bbb黄色大片| 少妇粗大呻吟视频| 制服诱惑二区| 麻豆乱淫一区二区| 欧美精品人与动牲交sv欧美| 久久精品亚洲精品国产色婷小说| 90打野战视频偷拍视频| 啦啦啦中文免费视频观看日本| 国产欧美日韩一区二区三| 777米奇影视久久| 亚洲全国av大片| 亚洲人成电影观看| a级毛片黄视频| 国内毛片毛片毛片毛片毛片| 啦啦啦在线免费观看视频4| 亚洲,欧美精品.| 亚洲欧洲精品一区二区精品久久久| 欧美午夜高清在线| av又黄又爽大尺度在线免费看| 精品人妻1区二区| 亚洲人成电影免费在线| 大型黄色视频在线免费观看| 精品熟女少妇八av免费久了| 国产高清激情床上av| 国产成人免费观看mmmm| 久久久久久人人人人人| 成人手机av| 亚洲成人免费电影在线观看| 久久精品91无色码中文字幕| 老司机亚洲免费影院| 亚洲九九香蕉| 国产亚洲精品一区二区www | 51午夜福利影视在线观看| 成人永久免费在线观看视频 | 亚洲专区字幕在线| 黄色视频,在线免费观看| 王馨瑶露胸无遮挡在线观看| 国产免费现黄频在线看| 精品少妇内射三级| 久久久国产精品麻豆| 精品卡一卡二卡四卡免费| 国产精品自产拍在线观看55亚洲 | 激情在线观看视频在线高清 | 黄片播放在线免费| 极品教师在线免费播放| 又紧又爽又黄一区二区| 精品国产一区二区久久| 欧美黑人精品巨大| 黄色毛片三级朝国网站| 久久亚洲真实| 18禁观看日本| 国产精品欧美亚洲77777| 啦啦啦 在线观看视频| 欧美国产精品va在线观看不卡| 香蕉国产在线看| 午夜福利视频在线观看免费| 成人18禁高潮啪啪吃奶动态图| 亚洲性夜色夜夜综合| 一级黄色大片毛片| 日本撒尿小便嘘嘘汇集6| 少妇被粗大的猛进出69影院| 亚洲av片天天在线观看| 色精品久久人妻99蜜桃| 日韩欧美国产一区二区入口| 国产人伦9x9x在线观看| 国产精品九九99| 欧美日韩精品网址| 在线观看免费午夜福利视频| 成年动漫av网址| 一级毛片精品| 久久亚洲真实| av线在线观看网站| 亚洲欧洲精品一区二区精品久久久| 午夜福利在线免费观看网站| 精品少妇一区二区三区视频日本电影| 亚洲人成伊人成综合网2020| 亚洲精品一二三| 精品国产国语对白av| 99riav亚洲国产免费| 亚洲少妇的诱惑av| 最新美女视频免费是黄的| 香蕉丝袜av| 丰满饥渴人妻一区二区三| e午夜精品久久久久久久| 亚洲精品一二三| 搡老熟女国产l中国老女人| 亚洲一区二区三区欧美精品| 不卡一级毛片| 成人三级做爰电影| 捣出白浆h1v1| 99九九在线精品视频| 亚洲人成电影观看| 免费黄频网站在线观看国产| 天堂8中文在线网| 亚洲精品中文字幕一二三四区 | 777米奇影视久久| 亚洲熟妇熟女久久| 欧美性长视频在线观看| 欧美人与性动交α欧美精品济南到| 国产不卡一卡二| 少妇猛男粗大的猛烈进出视频| 一级片免费观看大全| av线在线观看网站| 久久热在线av| 亚洲精品中文字幕一二三四区 | 精品欧美一区二区三区在线| 欧美日本中文国产一区发布| 亚洲午夜精品一区,二区,三区| 免费在线观看黄色视频的| 国产精品久久久人人做人人爽| 亚洲欧美一区二区三区久久| 久久精品亚洲av国产电影网| 国产xxxxx性猛交| 成人18禁高潮啪啪吃奶动态图| 18禁美女被吸乳视频| 午夜视频精品福利| 国产亚洲午夜精品一区二区久久| 大陆偷拍与自拍| 不卡一级毛片| 亚洲欧洲日产国产| 黑丝袜美女国产一区| 国产无遮挡羞羞视频在线观看| 婷婷丁香在线五月| 90打野战视频偷拍视频| 一区二区三区激情视频| 自拍欧美九色日韩亚洲蝌蚪91| 久久精品人人爽人人爽视色| 中文亚洲av片在线观看爽 | 9热在线视频观看99| 一进一出抽搐动态| 国产一卡二卡三卡精品| 老司机亚洲免费影院| 18禁黄网站禁片午夜丰满| 国产一卡二卡三卡精品| 一区二区三区乱码不卡18| 久久香蕉激情| 另类亚洲欧美激情| 女人爽到高潮嗷嗷叫在线视频| 欧美人与性动交α欧美精品济南到| 日韩欧美一区视频在线观看| 午夜福利影视在线免费观看| 久久久国产成人免费| 他把我摸到了高潮在线观看 | 成人手机av| 在线观看免费视频网站a站| 欧美日韩国产mv在线观看视频|