201143313 六、發明說明： 【發明所屬之技術領域】 本發明係關於一種光纖判斷方法，尤指一種利用光時 域反射儀（OTDR)配合一經驗模態分離法檢測光纖纜線何處 發生異常之方法。 【先前技術】 請參考圖12所示，光時域反射儀（〇TDR)10是一種可 • 用來檢測光纖纜線21 ~24之精密測量儀器，可測出光纖雙 線21〜24的衰減和接頭的衰減，以及可檢測出光纖雙線 21〜24發生中斷點的所在位置，其中一種使用的測量方法 稱為背向散射法（Back Scattering)，當光束在光纖鐵線 21〜24中傳輸時，由於光纖纜線21〜24的密度及組成不均 勻’會造成折射率的微小變動，以致產生雷利散射（Rayieigh Scattering)。在光纖纜線21 ~24中’大部分的能量還是沿 著纖核前向傳輸，只有少部分的能量會朝光發射端向後傳 • 播。 因此’若由一雷射光源1 1將光信號入射至光纖纜線 21〜24後，光時域反射儀1〇可用一高敏感度的檢光器監 測該光纖纜線中的背向散射光線的強度變化，獲得如圖，3 所示沿光纖規線長度分佈的軌跡信號（trace)。此技術對於 探測光纖之散射係數、損耗、連接點或中斷點等情況均相 當有用。 但利用光時域反射儀1 0進行監測時，因為是基於測 量背向散射的光信號而據以判斷，因此當進行長距離的監 201143313 測時’因信號微弱，難以精準的定位出光纖镜線2彳~24在 何處發生異常事件。 一般的信號利用波形圖表示時，是以時間作為橫軸。 在光學系統下由於週期過於短暫’又以巨觀的觀點輸出信 號，所以軌跡信號的橫軸是以距離作為表示單位。但軌跡 信號難以被線性化、週期性低，幾乎無週期性可言，因此 在解析軌跡信號時更提高其作業難度，不易精準定位出事 件發生的所在位置。 【發明内容】 有鑑於以光時域反射儀判斷光纖纜線事件時，隨著光 纖纜線的距離延長而使信號更加微弱，導致判斷結果產生 較大誤差且軌跡信號之解析作業不易，本發明之主要目的 係提供一種光纖事件判斷方法，以更精確的判斷出發生事 件的所在位置。 為達成前述目的’本發明之方法包含有： 自一光時域反射儀取得一光纖纜線之軌跡信號，該軌 跡信號係作為原始資料： 判斷疋否濾波，係設定一能量臨界值，根據原始資料 其能量高低判斷是否需要進行濾波，低於該能量臨界值的 部分係需要渡波’反之不需濾波； 以經驗模態分離法濾波，經判斷能量較低而需濾波的 原始資料，係以一經驗模態分離法加以濾波； 執行微分運算，係將判斷為不需濾波的原始資料以及 已利用該經驗模態分離法完成濾波的資料，對其執行一次 201143313 微分運算而獲得微分資料； 執行動態篩選，係濾除不必要之微分資料，僅保留所 需的微分資料； 重建事件’係根據該經過筛選後之微分資料重新建構 出事件群’得知光纖纜線發生事件之所在位置。 利用上述步驟，本發明可針對原始軌跡信號中能量較 弱的部分先進行濾波以濾除雜訊’而仍可保留原始軌跡信 號的特性；而判斷不須濾波以及完成濾波作業之軌跡信 遽，經由一次微分後可清楚獲得特徵表現，根據經過動態 篩選後之微分資料係可重新建構出事件群，故針對不具週 期性變化之軌跡信號，本發明係可判斷出光纖纜線發生事 件的所在位置。 其中，該經驗模態分離法係具有兩邊界條件，分別為： ^ a.極值點的個數總和必須與零交越點個數相等，或者 最多相差一個； b.在數線上，由數據的區域極大值所定義出的上包絡 次/、區域極小值所定義出的下包絡線，兩包絡線之平均值 為一斜率單調收斂至零點的任意曲線。 【實施方式】 响參考圖1所不，首先由光時域反射儀取得一 •'纖境’線的執跡k號（trace)作為原始資肖X⑴，軌跡信號 、圖2所示，虽中可能具有待分析的事件資訊a〜D，經由 、下步驟可刀析出光纖镜線所發生之事件及不同事件 A〜D所對應的位置，本發明的主要步驟包含： 201143313 —判：是否遽波101，由於原始資料x(t)係包含有不同 篁的身料’ &先設定—能量臨界值，根據能量高低判斷 是否需要進行遽波’針對低於該能量臨界值的部分係判斷 為需要濾波，反之則不用濾波。 以經驗模態分離法遽波102，經判斷能量較低而需遽 波的資料，本發明細放寬邊界條件的經驗模態分離法加 以遽波，其詳細作法將在後面詳述。201143313 VI. Description of the Invention: [Technical Field] The present invention relates to a fiber judging method, and more particularly to an optical time domain reflectometer (OTDR) combined with an empirical mode separation method for detecting an abnormality of an optical fiber cable. method. [Prior Art] Referring to FIG. 12, the optical time domain reflectometer (〇TDR) 10 is a precision measuring instrument for detecting the optical fiber cables 21 to 24, and can measure the attenuation of the optical fiber double wires 21 to 24. And the attenuation of the connector, and the position where the break point of the optical fiber two wires 21 to 24 can be detected. One of the measurement methods used is called Back Scattering, and the light beam is transmitted in the fiber wires 21 to 24. At the time, the density and composition of the optical fiber cables 21 to 24 are uneven, which causes a slight variation in the refractive index, so that Rayieigh Scattering is generated. In the fiber optic cables 21 ~ 24, most of the energy is still transmitted along the fiber core, and only a small amount of energy will be transmitted backwards toward the light emitting end. Therefore, if an optical signal is incident on the optical fiber cables 21 to 24 by a laser light source 11, the optical time domain reflectometer 1 can monitor the backscattered light in the optical fiber cable with a high sensitivity detector. The intensity change is obtained as shown in Fig. 3, and the trace signal (trace) distributed along the length of the fiber gauge line. This technique is useful for detecting the scattering coefficient, loss, junction or break point of an optical fiber. However, when the optical time domain reflectometer 10 is used for monitoring, it is judged based on the optical signal for measuring the backscattering. Therefore, when the long-distance monitoring 201143313 is performed, it is difficult to accurately locate the fiberscope due to the weak signal. Where is the line 2彳~24 where an abnormal event occurs. When a general signal is represented by a waveform diagram, time is taken as the horizontal axis. In the optical system, since the period is too short, and the signal is output from a giant view, the horizontal axis of the trajectory signal is represented by a distance. However, the trajectory signal is difficult to be linearized, the periodicity is low, and there is almost no periodicity. Therefore, when the trajectory signal is analyzed, the difficulty of the operation is further improved, and it is difficult to accurately locate the location where the event occurs. SUMMARY OF THE INVENTION In view of judging an optical fiber cable event by an optical time domain reflectometer, the signal is made weaker as the distance of the optical fiber cable is extended, resulting in a large error in the judgment result and difficulty in analyzing the trajectory signal. The main purpose is to provide a fiber optic event judging method to more accurately determine the location of the event. In order to achieve the foregoing objectives, the method of the present invention comprises: obtaining a track signal of a fiber optic cable from an optical time domain reflectometer, the track signal being used as the original data: determining whether the filter is filtered, setting an energy threshold according to the original The energy of the data is judged whether it needs to be filtered. The part below the energy threshold needs to cross the wave 'there is no need to filter; the empirical mode separation method filters the raw data that needs to be filtered after the lower energy is judged. The empirical mode separation method is used for filtering; the differential operation is performed, and the original data that is determined not to be filtered and the data that has been filtered by the empirical mode separation method are performed, and a differential operation is performed on the 201143313 differential operation to obtain differential data; Screening is to filter out unnecessary differential data and only retain the required differential data; the reconstruction event is based on the filtered differential data to reconstruct the event group to know where the fiber cable is. By using the above steps, the present invention can filter the weaker portion of the original track signal to filter out the noise and still retain the characteristics of the original track signal; and determine the track signal without filtering and completing the filtering operation. The feature representation can be clearly obtained after one differentiation, and the event group can be reconstructed according to the dynamically selected differential data system. Therefore, the present invention can determine the location of the fiber cable occurrence event for the track signal without periodic variation. . Wherein, the empirical mode separation method has two boundary conditions, respectively: ^ a. The sum of the extreme points must be equal to the number of zero crossing points, or at most one difference; b. on the number line, by the data The upper envelope defined by the upper limit of the area/the minimum value of the area is defined by the minimum value of the area. The average of the two envelopes is an arbitrary curve whose slope monotonically converges to zero. [Embodiment] Referring to Figure 1, the first time is obtained by the optical time domain reflectometer. The trace of the ''texture' line is taken as the original capital X(1). The track signal is shown in Figure 2. There may be event information a to D to be analyzed, and the events occurring in the fiber mirror line and the positions corresponding to the different events A to D can be analyzed through the following steps. The main steps of the present invention include: 201143313 - Judgment: Whether or not chopping 101, since the original data x(t) contains different body materials ' & first set - energy threshold value, according to the energy level to determine whether it is necessary to perform chopping 'for the part below the energy threshold value is judged as needed Filtering, otherwise, no filtering. The empirical mode separation method chopper 102, which is determined by the lower energy and the chopping data, is subjected to the empirical mode separation method of the finely widened boundary condition of the present invention, and the detailed method will be described in detail later.
執行微分運算1〇3,請參考圖3所示，若將已知的確 切光纖事件進行-次微分後，可發現該事件經—次微分後 會具有一定的特徵表現，本發明即以該特徵為預設參考資 料，提供作為與未知事件比對參考；目此，經前述步驟判 斷不需濾波的資料以及利用「經驗模態分離法」完成濾波 的資料，係對其執行一次微分的運算，以獲得資料的特徵 表現，圖3中的各實心圓點代表軌跡信號經過一次微分後 所獲得的資料。 執行動態篩選1 04，請參閱圖4所示，針對微分運算 所產生的資料，進一步進行篩選以保留適當的數據。在執 •行篩選時’可根據實務經驗的合理性與以往歷史數據統士十 之事件在軌跡信號中所佔據的百分比，擬定一合理的門# 值。例如圖4所示的同心圓係具有不同的半徑「〜 1 1 3 代表 不同的篩選範圍，各同心圓可定義出一組上、下限，在此 上、下限範圍之内的微分資料將會被剔除，該圓心的半# Γ可用以下公式表示： 其中 n=0, 1, 2,... r=,T」一次微分後的最大値I nX 历割精度 201143313 言奢參考圖5所示，完成動態筛後所保留的微分資料如 圖所示。 重建事件105,係將該篩選完成後之微分資料重新建 構為事件群，得知光纖窥線於何處發生事件。在進行重建 時’首先如圖6所示’將保留之微分資料從值域反推回其 所在之空間域，得到複數個事件群（事件群一至事件群六厂 再進-步如圖7所示，依據光時域反射儀（〇TDR)辨識盲區 的定義將前述事件群重組，建立事件區塊群（事件區塊群一 鲁至事件區塊群四），最後，請參考圖8所示，係根據事件區 塊群一至四對應至原始的軌跡信號中，在原始的軌跡信號 中便可獲付事件位置《至四。 本發明可進一步包含—輸出事件106之步驟，係將已 判斷出之事件輸出’供檢測者判斷參考 比對前述圖2及圖8後可知，根據前述步驟，本發明 確實可在執跡信號十尋找出光纖發生事件之所在位置。 前述用以進行濾波之「經驗模態分離法（Empirica| 鼸Mode Dec〇mposition, EMD)」，係可將該原㈣軌跡信號 分解成數個固有模態函數（丨ntrinsic M〇de Functi〇ns, IMF)，而每個固有模態函數具有良好的希爾伯特轉換 (Hilbert transform)特性。 首先，固有模態函數其傳統定義係根據以下兩個邊界 條件判斷，若資料可滿足這兩個條件，即可稱作固有模態 函數： 1 ·極值點（極大值、極小值）的個數總和必須與零交越 點（zero crossing)個數相等，或者最多相差一個。 201143313 2.在任思時間點’由數據的區域極大值（Local Max) 所定義出的上包絡線（upper envelop)與區域極小值（L〇ca| Min.)所定義出的下包絡線，取兩包絡線之平均值要為零。 根據上述兩邊界條件，原始資料x(t)可利用經驗模態 分離法轉換為固有模態函數，其步驟為： 建立上、下包絡線，係在原始資料x(t)中辨識出其局 部極大值與局部極小值，並分別對這兩組局部極值以三次Perform the differential operation 1〇3, please refer to FIG. 3. If the known exact fiber event is subjected to the -differential, it can be found that the event has a certain characteristic after the time-differentiation, and the present invention adopts the feature. For the preset reference data, provide a reference for comparison with the unknown event; for this reason, the data that is not filtered by the foregoing steps and the data that is filtered by the "empirical mode separation method" are subjected to a differential operation. To obtain the characteristic performance of the data, the solid dots in Fig. 3 represent the data obtained after the trajectory signal is differentiated once. Perform dynamic screening 1 04, see Figure 4, for further analysis of the data generated by the differential operations to preserve the appropriate data. In the implementation of the screening process, a reasonable threshold value can be drawn based on the reasonableness of the practical experience and the percentage of the previous historical data of the official data in the trajectory signal. For example, the concentric circles shown in Fig. 4 have different radii "~ 1 1 3 represents different screening ranges, and each concentric circle can define a set of upper and lower limits, and the differential data within the upper and lower limits will be Exclusion, the half of the center of the circle can be expressed by the following formula: where n=0, 1, 2,... r=, T" the maximum 値I nX after one differentiation. The accuracy of the cut is 201143313. The differential data retained after the dynamic sieve is completed is shown in the figure. The reconstruction event 105 reconstructs the differential data after the screening is completed into an event group to know where the fiber optic line is. In the reconstruction, 'first as shown in Figure 6', the retained differential data is pushed back from the value domain to the space domain in which it is located, and a plurality of event groups are obtained (the event group one to the event group six factories are further advanced). The event group is reorganized according to the definition of the blind zone of the optical time domain reflectometer (〇TDR), and the event block group (the event block group Lulu to the event block group 4) is established. Finally, please refer to FIG. According to the event block group one to four corresponding to the original track signal, the event position "to four" can be obtained in the original track signal. The invention may further include - the step of outputting the event 106, which will be judged The event output 'for the tester to judge the reference ratio is as shown in the foregoing FIG. 2 and FIG. 8. According to the foregoing steps, the present invention can indeed find the position of the fiber occurrence event in the trace signal ten. The aforementioned "experience for filtering" The modal separation method (Empirica| 鼸Mode Dec〇mposition, EMD) can decompose the original (four) trajectory signal into several intrinsic modal functions (丨ntrinsic M〇de Functi〇ns, IMF), and each eigenmode The function has a good Hilbert transform property. First, the traditional definition of the intrinsic mode function is judged according to the following two boundary conditions. If the data can satisfy these two conditions, it can be called the intrinsic mode function. : 1 · The sum of the extreme points (maximum value, minimum value) must be equal to the number of zero crossings, or at most one. 201143313 2. At the time point of the 'thinking' (Local Max) The upper envelope defined by the upper envelope (upper envelop) and the region minimum value (L〇ca| Min.), the average of the two envelopes is zero. According to the above two boundaries Condition, the original data x(t) can be converted into the intrinsic mode function by the empirical mode separation method. The steps are as follows: The upper and lower envelopes are established, and the local maximum and local are identified in the original data x(t). a minimum value and three times for each of the two sets of local extremum
樣條曲線（cubic sp丨jne curve)内插而得到兩條曲線，分別 作為歷時資料x(t)之上包絡線及下包絡線。 计异平均值，將該原始資料χ⑴之平均值曲線即為上、 下包絡線之平均，以m t表示。 執行綿選，在第一次篩選中，係將原始的原始資料χ⑴ =第-次計算出的平均值％⑴獲得第—個分量信號 单 完成一次的篩選動作’藉由不斷地篩選動作，可使 點沾值⑺逐漸趙於平緩，極值的個數也會逐漸等於零交越 點的個數’筛選過程可表示如下： 交越 x(t)-m1 = h1 h1-m2 = h2 匈斷篩選後的分量信號β 程中，妝— 置1〇唬疋否滿足邊界條件，在 將母一次得到的分量 而綷八二* 置彳與前述兩邊界條件 叩得合兩邊界條件者， 以h >主 刀量信號即為一固有模態 k — h表示，至此穿 筛選過程（sifting proces; 201143313 分離固有模態函數， 在獲侍固有模態函數後，將其白 原始資料x(t)中分離出也l 使肝再自 / iH n ，此時會得到一殘餘值 Π (residual)，此殘餘值包含 期較長、頻率較低之信號， 將此殘餘ϋ為新的原始㈣，重複前述步驟數次 =得到頻率由高至低的數個固有模態函數，其分離過 程如下· x(t)-c1=r1 「1 ·〇2 =「2The cubic sp丨jne curve is interpolated to obtain two curves, which are respectively the envelope and the lower envelope above the chronological data x(t). Taking the average value, the average curve of the raw data χ(1) is the average of the upper and lower envelopes, expressed as m t . In the first screening, the original raw data χ(1) = the average value %(1) calculated by the first time is obtained to obtain the screening action of the first component signal list once by 'continuous screening action. Make the point value (7) gradually flatten, the number of extreme values will gradually equal the number of zero crossing points. The screening process can be expressed as follows: Crossover x(t)-m1 = h1 h1-m2 = h2 In the filtered component signal β process, if the makeup condition is set to 1〇唬疋, the boundary condition is satisfied, and the component obtained at the time of the mother is set to 綷 二 * 彳 彳 彳 彳 彳 彳 彳 彳 彳 彳 彳 彳> The main tool quantity signal is an intrinsic mode k - h, so that the screening process (singing proces; 201143313 separates the intrinsic mode function, after obtaining the intrinsic mode function, the white source data x(t) In the isolation, the liver is again self-iH n , and a residual value Π (residual) is obtained. The residual value includes a signal with a longer period and a lower frequency, and the residue is reduced to a new original (four), repeated. The foregoing steps several times = get several natural modes from high to low frequency The function, the separation process is as follows: x(t)-c1=r1 "1 ·〇2 = "2
rk-i’Ck = rk ^ X{t)~Y,cj^rn Μ j述篩選過程即提供k號濾波效果，但是當篩選次數 過多時，因為信號的波形漸趨對稱、上下包絡線之平均值 漸趨於零，便會破壞原來較具物理意義的特徵值，故必項rk-i'Ck = rk ^ X{t)~Y,cj^rn Μ j The filtering process provides the k-filter effect, but when the number of screenings is too large, the waveform of the signal becomes symmetrical and the average of the upper and lower envelopes When the value is gradually zero, it will destroy the original more eigenvalues, so it must be
設定-停止㈣使制有模態函數仍保有具物理意義的特 徵值’在此該停止門檻SD係定為： T ®=Σ Λ=1 0.2 〜0.3 ~~κ7(〇~~ 然而原有經驗模態分離法其邊界定條件過於嚴格，難 以符合現實需求’因此’在本發明中係將帛2個邊界條件 在數線上，由數據的區域極大值（L〇Ca| Max)所定 義出的上包絡線（upper envelop)與區域極小值（L〇ca丨Mjn ) 所定義出的下包絡線，兩包絡線之平均值為一斜率單調收 斂至零點的任意曲線。」 201143313 由於經驗模態分離法所期待的結果是一條最終為單調 函數之曲線’且包絡線之平均值漸趨於零，又因為其所使 用之篩選方式，由‘‘ Weierstrass M-test定理”，可證得數 據之組成必須為一 Cauchy數列；所以，可將經驗模態分 離法之邊界條件放寬為Cauchy數列的特性—存在一固定 常數〇〇，使得數據X⑴在t2>T(T為一固定值）時， < Mn ’ Mn為一固定值，η為自然數，且 ;又由“極值定理”可知Μη必須為收斂序列；換言之， 馨 在數線上’由數據的區域極大值（Loca丨Max.)所定義出 的上包絡線（upper envelop)與區域極小值（Loca丨Min)所 疋義出的下包絡線’兩包絡線之平均值為一斜率單調收斂 至零點的任意曲線。」 4參考圖9所示’圖中的波形代表原始軌跡信號中須 進行濾波的部分，其上包絡線L1及下包絡線|_2係顯示在 圖1 〇 ’而圖1 1則表示經濾波後所獲得的資訊，其中可看 _ 出原始波形的特性被留下，雜訊的部分則可被降低。 【圖式簡單說明】 圖1 :本發明之方法流程圖。 圖2 :待分析之軌跡信號波形圖。 圖3 :待分析之軌跡信號經一次微分之對照示意圖。 圖4 :本發明動態篩選之示意圖。 _ 5 :本發明動態篩選完成之示意圖。 圖6:本發明將篩選完成之值域微分資料反算回空間 10 201143313 域之示意圖。 圖7 :本發明依據〇tdr辨識盲區的定義，將事件群 重組為事件區塊群之示意圖。 圖8 :本發明根據重建之事件區塊群於原始軌跡信號 中尋找事件位置之示意圖。 圖9 :本發明經判斷須執行濾波之原始軌跡資料示意 圖 參 圖 圖本發明建立上、下包絡線之示意圖。 1 1 .本發明原始資料經遽波處理後之示意圖。 圖1 2 .習用以光時域反射儀量測光纖斷點之架構示意 圖13:習用光時域反射儀所測得之軌跡資料示意圖。 【主要元件符號說明】 10光時域反射儀 11雷射光源 2144光纖纜線Set-stop (4) to make the modal function still retain the physical characteristic value'. Here, the stop threshold SD is defined as: T ®=Σ Λ=1 0.2 ~0.3 ~~κ7 (〇~~ However, the original experience The modal separation method has too strict boundary conditions and is difficult to meet the actual demand. Therefore, in the present invention, the two boundary conditions are on the number line, defined by the region maximum value (L〇Ca| Max) of the data. The lower envelope defined by the upper envelop and the region minimum value (L〇ca丨Mjn), the average of the two envelopes is an arbitrary curve whose slope monotonically converges to zero.” 201143313 Due to empirical mode separation The result expected by the law is a curve that is ultimately a monotonic function' and the mean value of the envelope gradually becomes zero, and because of the screening method used, the ''Weierstrass M-test theorem' can be used to prove the composition of the data. Must be a Cauchy sequence; therefore, the boundary condition of the empirical mode separation method can be relaxed to the characteristic of the Cauchy sequence—there is a fixed constant 〇〇 such that the data X(1) is at t2>T (T is a fixed value), <; Mn ' Mn is one Fixed value, η is a natural number, and; by the "extreme value theorem", it can be seen that Μη must be a convergent sequence; in other words, Xin is on the number line 'the upper envelope defined by the region maximum value of the data (Loca丨Max.) (upper envelop) and the region minimum value (Loca丨Min) the lower envelope 'the average of the two envelopes is an arbitrary curve whose slope monotonically converges to zero." 4 Refer to the figure in Figure 9 The waveform represents the portion of the original track signal that needs to be filtered. The upper envelope L1 and the lower envelope |_2 are shown in Figure 1 〇 ' and Figure 11 shows the filtered information. The characteristics of the noise are left behind, and the part of the noise can be reduced. [Simplified illustration of the drawing] Figure 1: Flow chart of the method of the present invention Figure 2: Waveform waveform of the track signal to be analyzed Figure 3: Track signal to be analyzed Figure 4: Schematic diagram of the dynamic screening of the present invention. _ 5: Schematic diagram of the dynamic screening of the present invention. Figure 6: The present invention reverses the calculation of the ranged differential data back to the space 10 201143313 . Figure 7 is a schematic diagram of the present invention for reorganizing an event group into an event block group according to the definition of the 〇tdr identification dead zone. Figure 8 is a schematic diagram of the present invention for finding an event location in the original trajectory signal according to the reconstructed event block group. The schematic diagram of the original trajectory data which is judged to be subjected to filtering according to the present invention is a schematic diagram of the upper and lower envelopes of the present invention. 1 1. The schematic diagram of the original data of the present invention after chopping processing. Schematic diagram of the structure of the fiber optic breakpoint measured by the domain reflectometer 13: Schematic diagram of the trajectory data measured by the conventional optical time domain reflectometer. [Main component symbol description] 10 optical time domain reflectometer 11 laser light source 2144 optical fiber cable