CN101943324B - Weak signal detection device and method based on wavelets and RBF neural network - Google Patents

Weak signal detection device and method based on wavelets and RBF neural network Download PDF

Info

Publication number
CN101943324B
CN101943324B CN 201010209979 CN201010209979A CN101943324B CN 101943324 B CN101943324 B CN 101943324B CN 201010209979 CN201010209979 CN 201010209979 CN 201010209979 A CN201010209979 A CN 201010209979A CN 101943324 B CN101943324 B CN 101943324B
Authority
CN
Grant status
Grant
Patent type
Prior art keywords
detection
device
signal
memory
method
Prior art date
Application number
CN 201010209979
Other languages
Chinese (zh)
Other versions
CN101943324A (en )
Inventor
冯健
刘振伟
刘金海
张化光
董良
马大中
魏向向
鲁忠沂
Original Assignee
东北大学
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Grant date

Links

Abstract

A weak signal detection device and method based on wavelets and a radial basis function (RBF) neural network belong to the technical field of signal detection. The device comprises a fourth-order Butterworth low pass filter, an A/D converter, an ARM microprocessor, a synchronous dynamic random access memory (SDRAM), a Nor Flash memory and a N and Flash memory. The device is beneficial to inhibiting noises and restoring, enhancing and extracting useful signals. The method can realize detection of a few oil leakage accidents which can not be detected by the conventional leakage detection devices and detect fluctuation less than 3%.

Description

基于小波和RBF神经网络的微弱信号检测装置及方法 Weak signal detection apparatus and method for wavelet and RBF neural network

技术领域 FIELD

[0001] 本发明属于信号检测技术领域,特别涉及一种基于小波和RBF神经网络的微弱信号检测装置及方法。 [0001] The present invention belongs to the field of signal detection technologies, particularly to a weak signal detection apparatus and method based on wavelet and RBF neural network.

背景技术 Background technique

[0002] 目前管道运输业的发展极其迅速,管线越来越多,运输的距离也越来越远。 [0002] the current development pipeline transportation is extremely rapid, more and more pipelines, more and more distant from the transport. 但是,随着管道使用时间的增长,管道泄漏事故发生的概率也在增大。 However, with the growth of pipeline use of time, the probability of pipeline leakage accidents has also increased. 众所周知,管道内输送的流体具有危险性和污染性,比如石油、天然气,那么一旦发生泄漏事故将会造成巨大的生命及财产损失和环境污染。 As we all know, the pipeline transport of hazardous fluids and polluting, such as oil, natural gas, then once the accident occurred would cause huge losses of life and property and environmental pollution. 特别是在我国,油气管网已有相当一部分步入衰老期,并且近十余年来又遭到前所未有的人为破坏,因此由泄漏事故而造成的损失十分巨大,从而严重地影响了管道运输业的发展。 Especially in China, oil and gas pipelines has been a considerable part into the aging period, and the past decade has been unprecedented vandalism, and therefore the loss caused by the spill is huge, which seriously affected the transport pipeline development of.

[0003] 我国油气管道安装的实时泄漏检测系统主要依赖于管段两端采集到的瞬变信号(包括瞬时压力变化、瞬时流量变化等,其中最重要的是瞬时压力变化信号)而设计的。 [0003] Real-time installation of oil and gas pipeline leak detection system depends on the ends of the tube sections collected transients (including transient pressure changes, flow rate changes, the most important is the instantaneous pressure variation signal) and design. 这些方法的只能检测出管道出现大波动的时候,也就是管道出现大量的漏油情况,但是如果出现小的波动,这些方法就无能为力了。 These methods can only detect pipeline big fluctuations in time, that is a large number of pipeline oil spills, but if small fluctuations, these methods can not do anything. 例如:我们通过压力变送器采集到了I. 00-5. OOv的压力信号,以前的方法能处理相对噪声信号3%的电压波动信号,如果是比3%小的波动,则常规的泄漏检测装置将无法检测到该类泄漏。 For example: We collected by a pressure transducer to a pressure signal I. 00-5 OOv previous method can handle a signal voltage fluctuation of 3% relative to the noise signal, if it is less than 3% of the fluctuation of the conventional leak detection. such means can not detect a leak.

发明内容 SUMMARY

[0004] 为克服上述方法之不足,本发明提出一种基于小波和RBF神经网络的微弱信号检测装置及方法,以实现对微小波动的检测。 [0004] To overcome the deficiencies of the above-described method, the present invention provides a weak signal detection apparatus and method based wavelet and RBF neural network, to enable detection of fine ripples.

[0005] 本发明的技术方案是这样实现的:本发明基于小波和RBF神经网络的微弱信号检测装置,包括四阶巴特沃兹低通滤波器、A/D转换器、ARM微处理器、同步动态随机存储器SDRAM.Nor Flash存储器、Nand Flash存储器,四阶巴特沃兹低通滤波器的输出端连接A/D转换器的输入端,A/D转换器的输出端连接ARM微处理器的输入端,ARM微处理器的第一输入输出端连接第一同步动态随机存储器SDRAM的输入输出端,ARM微处理器的第二输入输出端连接第二同步动态随机存储器SDRAM的输入输出端,ARM微处理器的第三输入输出端连接Nor Flash存储器的输入输出端,ARM微处理器的第四输入输出端连接Nand Flash存储器的输入输出端; [0005] aspect of the present invention is implemented as follows: RBF neural network based on wavelet and weak signal detection apparatus according to the present invention, comprises a four-order Butterworth low-pass filter, A / D converter, the ARM microprocessor, synchronization DRAM SDRAM.Nor Flash memory, Nand Flash memory, fourth order Butterworth low-pass filter output terminal is connected to a / D converter input, a / D converter output terminal connected to an input ARM microprocessor end, a first input connected to the output terminal of ARM microprocessor input and output terminals of the first synchronous dynamic random access memory SDRAM, and a second input connected to the output terminal of ARM microprocessor input and output terminals of the second synchronous dynamic random access memory SDRAM of ARM micro a third input and output terminal is connected to the processor Nor Flash memory inputs and outputs, a fourth input connected to the output terminal of ARM microprocessor Nand Flash memory inputs and outputs;

[0006] 四阶巴特沃兹低通滤波器由2个结构相同的两阶巴特沃兹低通滤波器组成,其中第一两阶巴特沃兹低通滤波器包括由电阻、电容组成的低通滤波器和运算放大器,低通滤波器第一电阻的一端连接运算放大器的反相输入端、低通滤波器电容的一端,低通滤波器第一电阻的另一端连接低通滤波器第二电阻的一端,低通滤波器第二电阻的另一端连接低通滤波器电容的另一端、运算放大器的输出端,其中,低通滤波器第一电阻的一端作为第一两阶巴特沃兹低通滤波器的输入端,运算放大器的输出端作为第一两阶巴特沃兹低通滤波器的输出端,第一两阶巴特沃兹低通滤波器的输出端连接第二两阶巴特沃兹低通滤波器的输入端,第二两阶巴特沃兹低通滤波器的输出端作为四阶巴特沃兹低通滤波器的输出端。 [0006] IV-order Butterworth low-pass filter consists of two identical structure of the two-order Butterworth low-pass filter, wherein the two first-order Butterworth low-pass filter comprises a low pass by a resistor, the capacitors inverting input terminal of the operational amplifier and a filter, one end of the first resistor connected to the low pass filter operational amplifier, one end of the low-pass filter capacitor, the other end of the first resistor connected to the low pass filter low pass filter of the second resistor one end, the other end of the second resistor connected to the other end of the low-pass filter capacitor low-pass filter, the output of the operational amplifier, wherein an end of the first resistance of the first low-pass filter as the two-order Butterworth low-pass input of the filter, the output of the operational amplifier as a first output terminal of two-order Butterworth low-pass filter, the output of two first-order Butterworth low-pass filter is connected to two second-order Butterworth low input-pass filter, two second-order Butterworth low-pass filter output terminal as an output four-order Butterworth low-pass filter.

[0007] 基于小波和RBF神经网络的微弱信号检测方法包括以下步骤: [0007] comprising the steps of detecting weak signal wavelet and RBF neural network:

[0008] 步骤I :采集设于管道的压力变送器的瞬时压力信号; Instantaneous pressure signal acquisition is provided a pressure transmitter conduit;: [0008] Step I

[0009] 步骤2 :对瞬时压力信号进行消噪处理,过程如下: [0009] Step 2: instantaneous pressure signal denoising process is as follows:

[0010] 步骤2-1 :计算信号的小波分解层数,方法如下; [0010] Step 2-1: wavelet decomposition level signal is calculated, as follows;

[0011] I)设置粗调节分解层数为N,保留分解的NI次和N次小波系数w ; [0011] I) the coarse adjustment is provided for the decomposition level N, NI decomposition retention times and the N-th wavelet coefficient W;

[0012] 2)对NI层的信号进行重构,对小波系数进行重构,计算其模极大值CDn+同理求N层的模极大值CDn,如果CDim < CDn,则判定该信号为有用信号为主,跳到4),否则判定该信号为噪声信号为主,转到3); [0012] 2) layer NI signal is reconstructed reconstructed wavelet coefficients, calculated modulus maxima Similarly CDn + N modulo the maximum value CDn layer, if CDim <CDn, it is determined that the signal is mainly useful signal, jump to 4), or determines that the signal is a noise signal mainly to 3);

[0013] 3)取分解的层数为N = N+1,进行N层分解,转到2); [0013] 3) to take the decomposition layers N = N + 1, N layer be decomposed to 2);

[0014] 4)放弃最后一次分解结果,确定最终的分解层数为NI ; Last [0014] 4) abandon decomposition results, to determine the final decomposition level of the NI;

[0015] 步骤2-2 :对分解之后信号的小波系数进行阈值量化: [0015] Step 2-2: After the wavelet coefficients of decomposed signal quantization threshold:

[0016] I)利用阈值函数对的小波系数进行阈值量化,公式如下: [0016] I) using the threshold function of the wavelet coefficients of the quantization threshold value, the following formula:

Figure CN101943324BD00051

[0018] 式中,ff(j, k)经过阈值量化后的小波系数,X为未经过阈值量化的小波系数,Z = aplogen, O为噪声的标准差,j,k分别是伸缩因子和平移因子; [0018] wherein, ff (j, k) through wavelet coefficients threshold quantization, X is not subjected wavelet coefficient threshold quantized, Z = aplogen, O standard noise difference, j, k are telescopic and translation for factor;

[0019] 2)信号重构; [0019] 2) the reconstructed signal;

[0020] 3)提取经步骤2-2中2)处理后的信号幅度fi,计算平均值,公式如下: [0020] 3) extracted via step 2-2 2) the amplitude of the signal processed Fi, calculating an average value, the following formula:

[0021] [0021]

Figure CN101943324BD00052

[0022] 式中,f,是处理后信号的幅度,/是处理后信号幅度的平均值,n是采样点的个数;去掉幅度的最大值和最小值,将f"作为理想的滤波幅值,公式如下: [0022] In the formula, f, is the amplitude of the processed signal, / is the average of the signal amplitude measured after treatment, n being the number of sampling points; removing the maximum and minimum amplitude of the f "as the ideal filter web value, the following formula:

[0023] [0023]

Figure CN101943324BD00053

[0024] 若处理后的信号幅度& ^ f*,则跳到步骤3,否则在0〜I之间重新选择a值,重复步骤2-2的I)〜3);所述的a值一般初始值取0. 5 ; [0024] If the signal amplitude measured after treatment & ^ f *, skip to step 3, otherwise, the value of a reselection between 0~I, repeating the steps I 2-2) ~ 3); according to a general value The initial value takes 0.5;

[0025] 步骤3 :对消噪后的信号进行分解,过程如下: [0025] Step 3: denoising the signal after decomposition, as follows:

[0026] 对步骤2得到的信号进行N层多分辨率分解,得到近似空间和细节间:设原信号最高频率为fmax,则一级小波分解后在近似空间和细节空间分别得到在[0,fmax/2]和[fmax/2,f_]频段上的信号描述;经过第N级小波分解后,在近似空间和细节空间分别得到在[0,fmax/2n]和频段上的信号描述;提取每段子信号的一个特征向量,得到一组特征向量f' (l),f/⑵........f' (n); [0026] Step 2 of the signal obtained was N layer multiresolution decomposition, and approximate the space between the detail: of the original signal is a maximum frequency Fmax, the wavelet decomposition in an approximate spatial and space were obtained in detail [0, fmax / 2] and [fmax / 2, the signal described F_] band on; after the first N-level wavelet decomposition, the approximation space and details space were obtained in [0, fmax / 2n] signals described in the frequency band and; extracting a piece for each feature vector signal to obtain a set of feature vectors f '(l), f / ⑵ ........ f' (n);

[0027] 步骤4 :提取压力信号中的微弱信号,方法如下; [0027] Step 4: the pressure signal to extract a weak signal, as follows;

[0028] 采用负梯度结合最近邻聚类法计算微弱信号的过程如下: [0028] The negative gradient of the weak signal most recently computed binding neighbor clustering method as follows:

[0029] (I)设定一个径向基函数的宽度! [0029] (I) to set a width of the radial basis function! ■和一个误差数值e,设一组样本数据为p = ■ and an error value e, a set of sample data set p =

{f (i), f' (i+1)......f' (n-1)), t = {f (i+2), f' (i+3),.....f' (n)), i = I, {F (i), f '(i + 1) ...... f' (n-1)), t = {f (i + 2), f '(i + 3), .... .f '(n)), i = I,

2,3. .n,假设神经网络中心(Cl,c2,. . . cj,m = 1,2,3,•..卩,且卩< n,取神经网络中心Cni的初始值C1 = f' (I),输出节点的权值Wi的初始值为W1 = f' (3),扩展宽度5 i的初始值为3 i = r ; 2,3. .n, the neural network is assumed that the center (Cl, c2 ,... Cj, m = 1,2,3, • .. Jie and Jie <n, the neural network taking the center Cni initial value C1 = f 'right (I), the output node value Wi is the initial value of W1 = f' (3), the extension width of the initial value of 5 i 3 i = r;

[0030] (2)计算神经网络的输出y」,j = 1,2,. . . . n_l,求出样本数据到中心的距离的最小值d min = Il f' (i)_cm Il , (m = 1,2,3——p),利用式子 [0030] (2) calculation of the neural network output y ', j = 1,2 ,.... N_l, sample data is obtained from the center to the minimum value d min = Il f' (i) _cm Il, ( m = 1,2,3 - p), using the formula

Figure CN101943324BD00061

得到神经网 Get nerve network

络输出值与实际值的误差值E ;如果E > e,d min > r,跳到(3);如果E > e,d min < r,跳到(4);如果E<e,则跳至IJ (5); Envelope error value E and the actual value of the output; if E> e, d min> r, skip to (3); if E> e, d min <r, skip to (4); if E <e, the jumping to IJ (5);

[0031] (3)m = m+1, cm = f' (i), wm = f; (i+2), S m = r,跳到第(5)步; [0031] (3) m = m + 1, cm = f '(i), wm = f; (i + 2), S m = r, skip to step (5);

[0032] (4)对RBF神经网络的中心、权值和宽度进行调整,公式为: [0032] (4) on the center RBF neural network, adjusting the weights and widths, of the formula:

Figure CN101943324BD00062

[0036] (5)如果i = n时候,学习结束,确定了RBF神经网络的中心、宽度和权值;反之,i=i+1,跳到第⑵步; [0036] (5) if i = n when the end of the study, to determine the center, width, and RBF neural network weights; otherwise, i = i + 1, skip to step ⑵;

[0037] (6)输入实时数据f' (n+1), (n+2); [0037] (6) the real-time input data f '(n + 1), (n + 2);

[0038] (7)根据以上6步确定的神经网络模型,并且可以得到神经网络的输出值yn,E'=If' (n+2)-yn ; [0038] (7) According to the above neural network model determined in step 6, and the neural network can obtain an output value yn, E '= If' (n + 2) -yn;

[0039] (8)设T为微弱信号阈值,如果E' >1',转到第(9)步,反之,1 = 1+1,11 = 11+1转到⑴; [0039] (8) is provided for the weak signal threshold T, if E '> 1', Go to step (9), on the contrary, 1 + 1 = 11 + 1 = 1,11 Go ⑴;

[0040] (9)检测到微弱信号大小为E'。 [0040] (9) detects the size of weak signal E '.

[0041] 本发明优点:利用本发明装置,有利于抑制噪声、恢复、增强和提取有用信号;利用本发明方法,可以实现常规泄露检测装置无法实现的少量漏油事故的检测,检测到小于的3%的波动。 [0041] The advantages of the present invention: With the present invention, apparatus is advantageous in suppressing noise, restoration, enhancement and extract the useful signal; using the method of the present invention can realize a small oil spill detection conventional leak detection apparatus can not be achieved, the detected smaller than 3% of the fluctuations.

附图说明 BRIEF DESCRIPTION

[0042] 图I为本发明基于小波和RBF神经网络的微弱信号检测装置系统框图; [0042] I FIG weak signal detection apparatus based on a system block diagram of the wavelet and RBF neural network of the present invention;

[0043] 图2为本发明基于小波和RBF神经网络的微弱信号检测装置电路原理图; [0043] FIG. 2 weak signal detection means and circuit schematic diagram of wavelet RBF neural network of the present invention;

[0044] 图3为本发明基于小波和RBF神经网络的微弱信号检测方法小波消噪流程图; [0044] FIG. 3 is a weak signal detection Wavelet and RBF neural network wavelet denoising based on a flowchart of the invention;

[0045] 图4为本发明基于小波和RBF神经网络的微弱信号检测方法流程图负梯度下降法结合最近邻聚类法流程图; [0045] FIG. 4 Weak Signal Detecting and RBF neural network in conjunction with a flowchart negative gradient descent flowchart nearest neighbor clustering method of the present invention;

[0046] 图5(a)为未采用本发明方法所侧得的含有微弱信号的曲线示意图,图5(b)为未采用本发明方法检测微弱信号的曲线示意图; [0046] FIG. 5 (a) is not the method of the present invention contains a schematic side curve was weak signal, FIG. 5 (b) is a graphical diagram showing the weak signal is not detected by the method of the present invention;

[0047] 图6为本发明基于小波和RBF神经网络的微弱信号检测方法去噪后检测微弱信号不意图;其中,6(a)表不含有微弱信号的管道曲线;6(b)表不小波去噪后的曲线;6(c)表示微弱信号曲线。 [0047] FIG. 6 is not intended to detect weak signals based weak signal detection wavelet denoising and RBF neural network of the present invention; wherein, 6 (a) does not contain a table graph weak signal conduit; 6 (B) is not wavelet table curve after denoising; 6 (c) indicates a weak signal curve.

[0048] I四阶巴特沃兹低通滤波器,2保护电路,3A/D转换器,4ARM微处理器,5同步动态随机存储器SDRAM,6第二同步动态随机存储器SDRAM,7 Nor Flash存储器,8 Nand Flash存储器 [0048] I four-order Butterworth low-pass filter, the protection circuit 2, 3A / D converter, a microprocessor 4arm, 5 synchronous dynamic random access memory SDRAM, 6 second synchronous dynamic random access memory SDRAM, 7 Nor Flash memory, 8 Nand Flash memory

具体实施方式 detailed description

[0049] 下面结合附图和具体实施例对本发明作进一步详细说明。 [0049] The accompanying drawings and the following specific embodiments of the present invention is described in further detail.

[0050] 图I和图2为基于小波和RBF神经网络的微弱信号检测装置的框图和电路原理图,该装置包括16位A/D转换器、ARM微处理器、同步动态随机存储器SDRAM、Nor Flash存储器、Nand Flash存储器,16位A/D转换器的输出端连接ARM微处理器的输入端,ARM微处理器的第一输入输出端连接第一同步动态随机存储器SDRAM, ARM微处理器的第二输入输出端连接第二同步动态随机存储器SDRAM,ARM微处理器的第三输入输出端连接Nor Flash存储器,ARM微处理器的第四输入输出端连接Nand Flash存储器;此外,还包括四阶巴特沃兹低通滤波器和包括电路,四阶巴特沃兹低通滤波器的输出端连接保护电路的输入端,保护电路的输出端连接16位A/D转换器的输入端;压力变送器传过来的1-5V的电压信号经过四阶巴特沃兹低通滤波器滤波后,将大于IKHz的高频信号滤掉,同时增加一个保护电路使输入的信号 [0050] Figure I and Figure 2 is a block diagram and a circuit diagram of the wavelet weak signal detection means and RBF neural network, the apparatus comprising a 16-bit A / D converter, the ARM microprocessor, synchronous dynamic random access memory SDRAM, Nor Flash memory, Nand Flash memory, 16-bit a / D converter output terminal connected to an input terminal of the ARM microprocessor, a first input connected to the output terminal of the first ARM microprocessor synchronous dynamic random access memory the SDRAM, ARM microprocessor a second input connected to the output terminal of the second synchronous dynamic random access memory SDRAM, a third input connected to the output terminal of ARM microprocessor Nor Flash memory, a fourth input connected to the output terminal of ARM microprocessor Nand Flash memory; in addition, further comprising fourth order Butterworth low-pass filter and comprising a circuit, the output of four-order Butterworth low-pass filter connected to the input terminal of the protection circuit, the protection circuit is connected to the output terminal of the input 16 bit a / D converter; pressure transmitter voltage signals of 1-5V is transmitted over the low-pass filter after four order Butterworth filter, a high frequency signal greater than IKHz filtered out, while increasing a protection circuit causes the input signal 定在一定的范围内,使其幅值介于A/D转换器模拟信号输入电压的范围内; It is set within a certain range, so that the amplitude in the range A / D converter analog input voltage signal;

[0051] 四阶巴特沃兹低通滤波器由运算放大器、电阻和电容组成,运算放大器由第一运算放大器和第二运算放大器组成,电阻由第一电阻R10、第二电阻R11、第三电阻R12、第四电阻R13、第五电阻R14组成,电容由第一电容Cl、第二电容C2、第三电容C3、第四电容C4组成,第二电容C2的一端连接第一电阻RlO的一端、第二电阻Rll的一端,第一电阻RlO的另一端连接运算放大器的反相输入端、第一电容Cl的一端,运算放大器的正相输入端接地,第一电容Cl的另一端、第二电阻Rll的另一端连接运算放大器的输出端后连接第三电阻R12的一端,第三电阻R12的另一端连接第三电容C3的一端、第四电阻R13的一端、第五电阻R14的一端,第四电阻R13的另一端连接第四电容C4的一端、第二运算放大器的反相输入端,第四电容C4的另一端、第五电阻R14的另一端连接第二运算放大器的输出端, [0051] IV-order Butterworth low-pass filter formed by an operational amplifier, resistors and capacitors, the operational amplifier by the first operational amplifier and a second operational amplifier, the resistance of the first resistor R10, a second resistor R11, the third resistor R12, a fourth resistor R13, composed of a fifth resistor R14, the capacitor of the first capacitor Cl, a second capacitor C2, third capacitor C3, the fourth capacitor C4 form, one end of the second capacitor C2 is connected to one end of a first resistor RlO, positive input is connected to an inverting input end of the second resistor Rll, and the other end of the first resistor RlO is connected to the operational amplifier, the end of the first capacitor Cl, the operational amplifier, the other end of the first capacitor Cl, and a second resistance Rll other end connected to the output terminal of the operational amplifier is connected to one end of the third resistor R12, the other end of the third resistor R12 is connected to one end of the third capacitor C3, one end of the end of the fourth resistor R13, a fifth resistor R14, a fourth the other end of the resistor R13 is connected to one end of the fourth capacitor C4, the inverting input of the second operational amplifier, the other end of the fourth capacitor C4, the other end of the fifth resistor R14 is connected to the output terminal of the second operational amplifier, 二运算放大器的4端连接-15V电压,第二运算放大器的8端连接+15V电压,第二运算放大器的输出端作为四阶巴特沃兹低通滤波器的输出端;其中,运算放大器的型号为LTL355 ; 4 two ends -15V voltage operational amplifier, the operational amplifier 8 is connected to a second terminal voltage of + 15V, the output terminal of the second operational amplifier as a low-pass filter output of four order Butterworth; wherein the operational amplifier model as LTL355;

[0052] 保护电路由第一二极管Dl和第二二极管D2组成,第一二极管Dl的负极连接第二二极管D2的正极、保护电路的输出端,第一二极管Dl的正极连接+IV电压,第二二极管D2的负极连接+5V电压; [0052] The protection circuit of a first diode Dl, a second diode D2, the cathode of the first diode Dl is connected to the anode of the second diode D2, an output terminal of the protection circuit, a first diode Dl is connected to the positive voltage + IV, the anode of the second diode D2 is connected to + 5V;

[0053] 16位A/D转换芯片的33脚连接保护电路的输出端,16位AD转换芯片的I〜19脚依次连接ARM9处理器的VD23脚〜VD9脚、VDO脚、EINTO脚〜EINT3脚;A/D转换器由A/D转换芯片组成,其中A/D转换芯片的型号为AD7656,ARM处理器的型号为S3C2440 ; [0053] 16 pin 33 A / D converter chip is connected to the output terminal of the protection circuit, 16-bit AD converting chip is connected sequentially VD23 I~19 pin pin pin ~VD9 ARM9 processor, the VDO feet, feet ~EINT3 foot EINTO ; a / D converter by the a / D converter chips, wherein the a / D conversion chip model AD7656, ARM processor S3C2440 model;

[0054] SDRAM存储器由两个存储器芯片组成,分别是第一存储芯片U7和第二存储芯片U8,第一存储芯片U7的16脚〜19脚依次连接第二存储芯片的16脚〜19脚;第一存储器芯片U7的DQO脚〜DQ6脚、BAl脚、BAO脚、DQ7脚〜DQ15脚、UDQM脚、LDQM脚、SCLK脚、SCKE脚依次连接ARM9处理器的DATE16脚〜DATE22脚、ADDR25脚、ADDR24脚、DATE23脚〜DATE31脚、nWBE3脚、nWBE2脚、SCLK3脚、SCKE脚;第二存储器芯片U6的DQO脚〜DQ15脚、AO脚〜A12脚、BAO脚、BAl脚、LDQM脚、UDQM脚、SCKE脚、SCLK脚依次连接ARM9处理器的DATEO 脚〜DATE15 脚、ADDR2 脚〜ADDR14 脚、ADDR24 脚、ADDR25 脚、nffBEO 脚、nffBEl 脚、SCKE 脚、SCLKO 脚;其中SDRAM 的型号为K4S561632C-TC75 ; [0054] SDRAM memory consists of two memory chips, namely the 16-pin pin ~19 first memory chip and the second memory chip U7 U8, the first memory chip U7 pin 16 pin ~19 sequentially connected to the second memory chip; the first memory chip U7 the pin ~DQ6 DQO feet, feet BAl, BAO feet, feet ~DQ15 DQ7 pins, UDQM feet, feet LDQM, SCLK pin, pin SCKE sequentially connected ARM9 processor DATE16 ~DATE22 foot pin, ADDR25 foot, ADDR24 feet, DATE23 ~DATE31 foot pin, nWBE3 feet, nWBE2 feet, SCLK3 feet, SCKE foot; the second memory chip U6 ~DQ15 DQO foot pin, AO ~A12 foot pin, BAO feet, BAl feet, LDQM feet, UDQM feet , SCKE feet, SCLK pin ARM9 processor sequentially connected ~DATE15 DATEO foot pin, ADDR2 ~ADDR14 foot pin, ADDR24 feet, ADDR25 feet, nffBEO feet, nffBEl feet, SCKE feet, SCLKO feet; wherein the SDRAM model K4S561632C-TC75 ;

[0055] Nand Flash存储器的RE脚、WE脚、CE脚依次连接ARM9的nFRE脚、nFWE脚、nFCE脚;其中,Nand Flash存储器的型号为K9FXX08 ; [0055] RE foot Nand Flash memory, WE foot, CE pin of sequentially connected nFRE foot ARM9, nFWE feet, nFCE feet; wherein, Nand Flash memory model K9FXX08;

[0056] Nor Flash 存储器的ADDRl 脚〜ADDR20 脚、DO 脚〜D15 脚、nOE 脚、nWE 脚、nGSCS。 [0056] ADDRl foot ~ADDR20 feet Nor Flash memory, DO feet ~D15 feet, nOE feet, nWE feet, nGSCS. 脚依次连接ARM9处理器的Atl脚〜A19脚、DATEO脚〜DATE15脚、OE脚、WE脚、CE脚依次连接;Nor Flash存储器的型号为AM29LV800BB ; ARM9 processor sequentially connecting pin Atl ~A19 foot pin, DATEO ~DATE15 foot pin, OE pin, WE pin, CE pin sequentially connected; Nor Flash memory model AM29LV800BB;

[0057] Nand Flash存储器的1/00〜1/07依次连接NorFlash存储器的D7脚〜DO脚; [0057] 1 / 00~1 / 07 D7 are sequentially connected memory NorFlash ~DO foot pin Nand Flash memory;

[0058] 本实施例中基于小波和RBF神经网络的微弱信号检测方法,包括以下步骤: [0059] 步骤I :采集压力变送器的瞬时压力信号,压力变送器将该信号的幅值确定在IV〜5V之间; [0058] Weak Signal Detection Based on Wavelet and RBF neural network embodiment in the present embodiment, comprising the steps of: [0059] Step I: instantaneous pressure signal acquisition pressure transmitter, pressure transmitter determines the amplitude of the signal between IV~5V;

[0060] 步骤2 :对瞬时压力信号进行消噪处理,过程如下: [0060] Step 2: instantaneous pressure signal denoising process is as follows:

[0061 ] 步骤2-1 :根据管道实际情况,选择小波函数为db5,粗调节的分解层数设定为N =4,采用步骤2-1中2)-4)步的细调节确定最终的分解层数N = 6 ; [0061] Step 2-1: according to the actual pipe, wavelet function is selected DB5, the coarse adjustment decomposition level is set to N = 4, steps 2-1 using 2) -4) of the fine adjustment step to determine the final decomposition level N = 6;

[0062] 步骤2-2 :对分解之后信号的小波系数进行阈值量化: [0062] Step 2-2: After the wavelet coefficients of decomposed signal quantization threshold:

[0063]取 a = 0. 5,2 = ¢7-^2Ioge n ,其中n = 10 ; [0063] Take a 0. 5,2 = ¢ 7- ^ 2Ioge n =, where n = 10;

[0064] 步骤3 :对步骤2得到的信号进行5层多分辨率分解,经过5级小波分解后,在近似空间和细节空间分别得到在[0,fmax/2]、[fmax/22,f-/22_l]、[fmax/23,f-/23_l]、[fmax/24,f_/24-l]、[f_/25,f_/25-l]频段上的信号描述;提取每段子信号的一个特征向量,得到一组特征向量f' (l),f/ (2),f' (3),f' (4),f' (5),如图3 所示; [0064] Step 3: Step 2 of the signal layer 5 was subjected to multi-resolution decomposition. After wavelet decomposition 5, details of the approximation space and space were obtained in [0, fmax / 2], [fmax / 22, f - / 22_l], [fmax / 23, f- / 23_l], [fmax / 24, f_ / 24-l], [f_ / 25, f_ / 25-l signal description] frequency bands; extracting signals of each piece a feature vector, to obtain a set of feature vectors f '(l), f / (2), f' (3), f '(4), f' (5), shown in Figure 3;

[0065] 步骤4:提取压力信号中的微弱信号,如图4所示:采用3层网络结构的神经网络,输入节点数为3,神经元为2个,输出层节点数为I,即通过输入的3个节点,得到一个输出,这里离线学习数据为100个,在线学习的数据为50个,将阈值控制在T = I. 2 X IO-5,即可检测检测出所有微弱信号。 [0065] Step 4: extracting weak signals in the pressure signal, shown in Figure 4: a 3-layer neural network of the network structure, the input nodes is 3, two neurons, the output layer nodes is I, i.e., by three input nodes, to obtain an output, where the offline learning data 100, study data line 50 is the threshold control in T = I. 2 X IO-5, can be detected to detect all weak signals.

[0066] 从图5,图6可以看出,图5虽然能检测出信号,但是只能检测到大于0. 00015的信号,图6是本发明中小波和RBF结合,效果明显,可以检测的信号可以达到I. 2X 10_5。 [0066] From FIG. 5, it can be seen in FIG. 6, FIG. 5, although the signal can be detected, but only the detected signal is greater than 0.00015, the present invention FIG. 6 is a wavelet and RBF binding effect obviously, can be detected the signal can reach the I. 2X 10_5.

Claims (2)

  1. 1. 一种基于小波和RBF神经网络的微弱信号检测装置,其特征在于:该装置包括四阶巴特沃兹低通滤波器、A/D转换器、ARM微处理器、同步动态随机存储器SDRAM、Nor Flash存储器、Nand Flash存储器,四阶巴特沃兹低通滤波器的输出端连接A/D转换器的输入端,A/D转换器的输出端连接ARM微处理器的输入端,ARM微处理器的第一输入输出端连接第一同步动态随机存储器SDRAM的输入输出端,ARM微处理器的第二输入输出端连接第二同步动态随机存储器SDRAM的输入输出端,ARM微处理器的第三输入输出端连接Nor Flash存储器的输入输出端,ARM微处理器的第四输入输出端连接Nand Flash存储器的输入输出端; 所述的四阶巴特沃兹低通滤波器由2个结构相同的两阶巴特沃兹低通滤波器组成,其中第一两阶巴特沃兹低通滤波器包括由电阻、电容组成的低通滤波器和运算放大器,低通滤波器 A weak signal detection means Wavelet and RBF neural network, which is characterized in that: the apparatus includes a four-order Butterworth low-pass filter, A / D converter, the ARM microprocessor, synchronous dynamic random access memory SDRAM, Nor Flash memory, Nand Flash memory, fourth order Butterworth low-pass filter output terminal is connected to a / D converter input, a / D converter output terminal connected to an input of ARM microprocessor, ARM microprocessor a first input-output terminal is connected a first synchronous dynamic random access memory SDRAM inputs and outputs, a second input connected to the output terminal of ARM microprocessor input and output terminals of the second synchronous dynamic random access memory SDRAM, the third ARM microprocessor input-output terminal is connected Nor Flash memory inputs and outputs, a fourth input connected to the output terminal of ARM microprocessor Nand Flash memory inputs and outputs; the fourth order Butterworth low-pass filter composed of two identical structures 2 order Butterworth low-pass filter, wherein the two first-order Butterworth low-pass filter comprises a low pass filter and an operational amplifier by resistors, capacitors, and the low-pass filter 第一电阻的一端连接运算放大器的反相输入端、低通滤波器电容的一端,低通滤波器第一电阻的另一端连接低通滤波器第二电阻的一端,低通滤波器第二电阻的另一端连接低通滤波器电容的另一端、运算放大器的输出端,其中,低通滤波器第一电阻的一端作为第一两阶巴特沃兹低通滤波器的输入端,运算放大器的输出端作为第一两阶巴特沃兹低通滤波器的输出端,第一两阶巴特沃兹低通滤波器的输出端连接第二两阶巴特沃兹低通滤波器的输入端,第二两阶巴特沃兹低通滤波器的输出端作为四阶巴特沃兹低通滤波器的输出端。 One end of the inverting input terminal of the first resistor connected to one end of the operational amplifier, one end of the low-pass filter capacitor, the other end of the first resistor connected to the low pass filter low pass filter of the second resistor, a second low-pass filter resistor the other end of the output of the low pass filter connected to the other end of the capacitor, the output of the operational amplifier, wherein an end of the first resistance of the first low-pass filter as the two-order Butterworth low-pass filter input of the operational amplifier as the output of the first two side-order Butterworth low-pass filter, the output of two first-order Butterworth low-pass filter connected to the second input of the two-stage Butterworth low-pass filter, a second two an output terminal order Butterworth low-pass filter as an output four-order Butterworth low-pass filter.
  2. 2.采用权利要求I所述的基于小波和RBF神经网络的微弱信号检测装置的检测方法,其特征在于:包括以下步骤: 步骤I :采集设于管道的压力变送器的瞬时压力信号; 步骤2 :对瞬时压力信号进行消噪处理,过程如下: 步骤2-1 :计算信号的小波分解层数,方法如下; 1)设置粗调节分解层数为N,保留分解的NI次和N次小波系数w ; 2)对NI层的信号进行重构,对小波系数进行重构,计算其模极大值CDn+同理求N层的模极大值CDn,如果CDim < CDn,则判定该信号为有用信号为主,跳到4),否则判定该信号为噪声信号为主,转到3); 3)取分解的层数为N = N+1,进行N层分解,转到2); 4)放弃最后一次分解结果,确定最终的分解层数为NI ; 步骤2-2 :对分解之后信号的小波系数进行阈值量化: 1)利用阈值函数对的小波系数进行阈值量化,公式如下: 式中,W(j,k)经过阈值量化后的小 2. The use of claim weak signal detection means detecting method based on wavelet and RBF neural network of claim I, characterized by: comprising the following steps: Step I: instantaneous pressure signal acquisition is provided a pressure transmitter conduit; step 2: the instantaneous pressure signal denoising process is as follows: step 2-1: wavelet decomposition level signal is calculated, as follows; a) is provided as the coarse adjustment decomposition level N, NI decomposition retention times and the N-th wavelet coefficient w; 2) NI layer signal is reconstructed, reconstructed wavelet coefficients, calculated modulus maxima Similarly CDn + N modulo the maximum value CDn layer, if CDim <CDn, it is determined that the signal is mainly useful signal, jump to 4), or determines that the signal is a noise signal mainly to 3); 3) to take the decomposition layers N = N + 1, N layer be decomposed to 2); 4 ) last to abandon the decomposition, determining a final decomposition level of the NI; step 2-2: the wavelet coefficients after the signal decomposition threshold quantization: 1) the threshold quantization using the threshold function of the wavelet coefficient, the following formula: wherein , W (j, k) after quantization smaller threshold 波系数,X为未经过阈值量化的小波系数,A = (T^Hoge n, 0为噪声的标准差,j,k分别是伸缩因子和平移因子; 2)信号重构; 3)提取经步骤2-2中2)处理后的信号幅度fy计算平均值,公式如下:式中,A是处理后信号的幅度,>是处理后信号幅度的平均值,n是采样点的个数;去掉幅度的最大值和最小值,将f"作为理想的滤波幅值,公式如下: Wavelet coefficients, wavelet coefficients X has not passed the threshold quantized, A = (T ^ Hoge n, 0 is the noise standard deviation, j, k are telescopically factor and a translation factor; 2) signal reconstruction; 3) extracting by the step of 2-2 2) fy processed signal amplitude calculating an average value, the following formula: wherein, a is the amplitude of the processed signal,> is the average of the signal amplitude measured after treatment, n being the number of sampling points; remove amplitude the maximum and minimum values, the f "as the ideal filter amplitude, the following formula:
    Figure CN101943324BC00031
    若处理后的信号幅度A ^ 则跳到步骤3,否则在0〜I之间重新选择^值,重复步骤2-2的I)〜3);所述的a值一般初始值取0. 5 ; 步骤3 :对消噪后的信号进行分解,过程如下: 对步骤2得到的信号进行N层多分辨率分解,得到近似空间和细节间:设原信号最高频率为fmax,则一级小波分解后在近似空间和细节空间分别得到在[0,fmax/2]和[fmax/2,fmax]频段上的信号描述;经过第N级小波分解后,在近似空间和细节空间分别得到在[0,f_/2n]和[f_/2n,f_/2n-l]频段上的信号描述;提取每段子信号的一个特征向量,得到一组特征向量 If the processed signal amplitude A ^ skip to step 3, otherwise reselection ^ value, repeating the steps I 2-2) ~ 3) between 0~I; according to the general initial value takes a value 0.5 ; step 3: denoising the signal after decomposition, as follows: step 2 of the signal obtained was N layer multiresolution decomposition, the space between the approximate and detail: the highest frequency of the original signal is Fmax, the wavelet decomposition of a after the approximation space and details space were obtained in [0, fmax / 2] and [fmax / 2, fmax] signals described band on; after the first N-level wavelet decomposition, the approximation space and details space were obtained in [0 , f_ / 2n] and [f_ / 2n, the signal described f_ / 2n-l] frequency bands; extracting each piece of the one feature vector signal to obtain a set of feature vectors
    Figure CN101943324BC00032
    步骤4 :提取压力信号中的微弱信号,方法如下; 采用负梯度结合最近邻聚类法计算微弱信号的过程如下: (1)设定一个径向基函数的宽度r和一个误差数值e,设一组样本数据为P= (f' (i),(i+1).......f' (n-1)), t = (f' (i+2), f' (i+3),.....f' (n)), i = I, 2, 3. • n,假设神经网络中心(CpC2,. . . cm),m = I, 2, 3,. . . p,且p < n,取神经网络中心Cm的初始值C1 =f' (I),输出节点的权值Wi的初始值为W1 = f' (3),扩展宽度5 i的初始值为S工=r ; (2)计算神经网络的输出y」,j = 1,2,....n-1,求出样本数据到中心的距离的最小值dmin =IIf' (i)-cm| | (m = I, 2,. . . p),利用式子E = ^ ((- ~ L+)得到神经网络输出值与实际值的误差值E ;如果E > e, dmin > r,跳到(3);如果E > e, dmin < r,跳到(4);如果E<e,则跳到(5); (3)m = m+1, cm = f' (i), wm = f' (i+2), S m = r,跳到第(5)步; (4)对RBF神经网络的中心、权值和宽度进行调整,公式为: Step 4: the pressure signal to extract a weak signal, as follows; bonding process using the most recently calculated negative gradient o Clustering weak signals as follows: (1) setting a width of a radial basis function r and an error value E, provided a sample data set as P = (f '(i), (i + 1) ....... f' (n-1)), t = (f '(i + 2), f' (i +3), ..... f '(n)), i = I, 2, 3. • n, the neural network is assumed that the center (CpC2 ,... cm), m = I, 2, 3 ,.. . p, and p <n, the neural network taking the center Cm of the initial value of C1 = f '(I), the initial value of the output node weight values ​​Wi W1 = f' (3), extended the initial value of the width of 5 i workers S = r; (2) calculation of the neural network output y ', j = 1,2, .... n-1, the sample data is obtained from the center to a minimum value dmin = IIf' (i) -cm | | (m = I, 2 ,. p..), using the equation E = ^ ((- ~ L +) to give an error value E neural network output value and the actual value; if E> e, dmin> r, jumping to (3); if E> e, dmin <r, skip to (4); if E <e, skip to (5); (3) m = m + 1, cm = f '(i), wm = f '(i + 2), S m = r, skip to step (5); (4) the center of the RBF neural network, adjusting the weights and widths, of the formula: c(w + l) = c(w) + Acm=c(w) + ^—11)(/, (OQ) W-Iw(m +1) = w(m) + Awm = w(m)+^iG(|| f (/) - cm ||) r=lS(m + l)=S(m) + ASm = 11)11/々)_〜Il2 (5)如果i = n时候,学习结束,确定了RBF神经网络的中心、宽度和权值;反之,i =i+1,跳到第⑵步; (6)输入实时数据 f' (n+1),f' (n+2); (7)根据.以上6步确定的神经网络模型,并且可以得到神经网络的输出值yn,E'=f' (n+2)-yn ; (8)设T为微弱信号阈值,如果E' > T,转到第(9)步,反之,i = i+1, n = n+1转到(I); (9)检测到压力信号中的微弱信号大小为E'。 c (w + l) = c (w) + Acm = c (w) + ^ -11) (/, (OQ) W-Iw (m +1) = w (m) + Awm = w (m) + ^ iG (|| f (/) - cm ||) r = lS (m + l) = S (m) + ASm = 11) 11 / 々) _~Il2 (5) if i = n, when learning is completed determining the RBF center, width, and weights of the neural network; otherwise, i = i + 1, skip to step ⑵; (6) the real-time input data f '(n + 1), f' (n + 2); (7) a neural network model than 6 determined in step, and may obtain neural network output value yn, E '= f' (n + 2) -yn; (8) Let T be a weak signal threshold, if E ' > T, go to step (9), and vice versa, i = i + 1, n = n + 1 to (the I); (9) detects a weak signal in the magnitude of the pressure signal E '.
CN 201010209979 2010-06-28 2010-06-28 Weak signal detection device and method based on wavelets and RBF neural network CN101943324B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN 201010209979 CN101943324B (en) 2010-06-28 2010-06-28 Weak signal detection device and method based on wavelets and RBF neural network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN 201010209979 CN101943324B (en) 2010-06-28 2010-06-28 Weak signal detection device and method based on wavelets and RBF neural network

Publications (2)

Publication Number Publication Date
CN101943324A true CN101943324A (en) 2011-01-12
CN101943324B true CN101943324B (en) 2012-10-17

Family

ID=43435461

Family Applications (1)

Application Number Title Priority Date Filing Date
CN 201010209979 CN101943324B (en) 2010-06-28 2010-06-28 Weak signal detection device and method based on wavelets and RBF neural network

Country Status (1)

Country Link
CN (1) CN101943324B (en)

Families Citing this family (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102607055B (en) * 2012-02-29 2014-10-29 东南大学 A method of controlling the load change procedure boiler coal feed rate to prevent excessive adjustment
CN104270167A (en) * 2013-12-20 2015-01-07 张冬 Signal detection and estimation method based on multi-dimensional characteristic neural network
CN104089186B (en) * 2014-06-30 2016-09-28 东北大学 An abnormality diagnosis method based on a combination of filtering and dynamic line pressure threshold
CN104132250B (en) * 2014-07-14 2017-07-04 上海师范大学 Leakage one feature vector extraction duct improved wavelet packet based
CN105546361A (en) * 2016-03-08 2016-05-04 钱昊铖 Acoustic-wave-method gas pipeline leakage monitoring method based on ANN (Artificial Neural Network)
CN106015951A (en) * 2016-07-07 2016-10-12 北京信息科技大学 Gas transmission pipeline leakage detection system and method suitable for multiple state changes

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1799013A (en) 2003-06-05 2006-07-05 罗斯蒙德公司 Process device diagnostics using process variable sensor signal
CN1847867A (en) 2006-03-24 2006-10-18 西南交通大学 Post-wavelet analysis treating method and device for electric power transient signal
CN101246200A (en) 2008-03-10 2008-08-20 湖南大学 Analog PCB intelligent test system based on neural network
CN201508350U (en) 2009-07-14 2010-06-16 辽阳西姆莱斯石油专用管制造有限公司 Ultrasonic automatic defect detection device for petroleum pipes

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7949495B2 (en) * 1996-03-28 2011-05-24 Rosemount, Inc. Process variable transmitter with diagnostics

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1799013A (en) 2003-06-05 2006-07-05 罗斯蒙德公司 Process device diagnostics using process variable sensor signal
CN1847867A (en) 2006-03-24 2006-10-18 西南交通大学 Post-wavelet analysis treating method and device for electric power transient signal
CN101246200A (en) 2008-03-10 2008-08-20 湖南大学 Analog PCB intelligent test system based on neural network
CN201508350U (en) 2009-07-14 2010-06-16 辽阳西姆莱斯石油专用管制造有限公司 Ultrasonic automatic defect detection device for petroleum pipes

Also Published As

Publication number Publication date Type
CN101943324A (en) 2011-01-12 application

Similar Documents

Publication Publication Date Title
CN201508392U (en) Frequency-conversion grounding resistance measuring instrument
CN101162838A (en) Low current neutral grounding system fault route selecting method by wavelet package decompose and correlation analysis
Hausenblas Existence, uniqueness and regularity of parabolic SPDEs driven by Poisson random measure
Manikandan et al. Detection and classification of power quality disturbances using sparse signal decomposition on hybrid dictionaries
CN101261292A (en) Base wave and harmonic detection method based on fiver item Rife-Vincentú¿1ú®window double spectral line interpolation FFT
CN102495336A (en) Distributed single-phase earth fault ranging system and ranging method thereof
CN101154807A (en) Self-adaption route selection method for single-phase ground fault of power distribution network based on transient zero sequence current
CN102590716A (en) Power cable partial discharge signal denoising method based on wavelet analysis
CN102005739A (en) Method for identifying extra-high voltage direct current transmission line area inside and outside fault wavelet energy
CN102620928A (en) Wind-power gear box fault diagnosis method based on wavelet medium-soft threshold and electronic-magnetic diaphragm (EMD)
CN101630982A (en) Broadband cognitive radio frequency spectrum detection method based on parallelly compressed sensing
CN101793918A (en) Voltage sag detection method
CN101441233A (en) Base wave and harmonic detecting method based on Kaiser window double-line spectrum insert value FFT
CN103018627A (en) Adaptive fault type fault line detection method for non-effectively earthed system
Qin et al. Higher-density dyadic wavelet transform and its application
CN103529321A (en) Current transformer saturation detecting method
CN101666848A (en) Fault location method based on travelling wave
CN103116064A (en) Method and device for detecting voltage fluctuation and flicker based on energy operator and spectrum correction
CN101661074A (en) Judging method for transient and permanent faults of transmission line with parallel reactance
CN103308766A (en) Harmonic analysis method based on Kaiser self-convolution window dual-spectrum line interpolation FFT (Fast Fourier Transform) and device thereof
Sen et al. Application of wavelet soft-threshold de-noising technique to power quality detection [J]
CN102590703A (en) Single-phase grounding failure route selection method of resonance grounding system based on zero-sequence transient charge
CN101149391A (en) Differential capacitance type sensor detection circuit
CN101819253A (en) Probabilistic neural network-based tolerance-circuit fault diagnosis method
Shenghe A novel algorithm for embedding watermarks into digital audio signals [J]

Legal Events

Date Code Title Description
C06 Publication
C10 Entry into substantive examination
C14 Grant of patent or utility model
EXPY Termination of patent right or utility model