2024年1月9日发(作者：)
附件 外文翻译
Optic fiber-based dynamic pressure sensor for WIM system
ShenfangYuana, ,
, FahardAnsarib, XiaohuiLiua and Yang Zhaob
aThe Aeronautical Key Laboratory for Smart Materials and Structures, Nanjing
University of Aeronautics and Astronautics, 29 YuDao Street, Nanjing 210016, China
bDepartment of Civil and Materials Engineering, University of Illinois at
Chicago, Illinois, IL 60607, USA
Received 16 August 2004;
accepted 10 November 2004.
Available online 15 December 2004.
Abstract
An optic fiber-based dynamic pressure sensor is described here to measure
weight-in-motion of vehicles. In the research reported herein, a Michelson
interferometer with specially designed hardware and software were developed and
experimentally subjected to dynamic compressive loads of different magnitudes, and
loading rates. Experiments showed that both output fringe number and fringe period
could be used to indicate the dynamic load. A calibration technique has been put
forward to calibrate the sensor. Both the dynamic weight and static weight of the
vehicle passed can be obtained. The findings that resulted from these studies
developed an understanding for the behavior of interferometer sensor under dynamic
compressive states of stress and are fundamental to the application of fiber optic
sensors for the monitoring of truck vehicle weights while in motion.
Keywords: Optic fiber sensor; Dynamic pressure; Weight-in-motion; Hardware
and software
Article Outline
1. Introduction
2. The sensor design
2.1. Sensor setup
2.2. Sensor principle
3. Experimental procedures and results
3.1. Experimental setup
3.2. Experimental data
3.3. Repeatability of the sensor
3.4. Calibration of the sensor
3.4.1. Calibration of the static weight
3.4.2. Calibration of the dynamic weight
4. Conclusion
References
Vitae
1. Introduction
The need to weigh vehicles in motion, applied especially to traffic control, has
grown substantially in the past decades. Several techniques for weighting vehicles
in-motion are now used including piezoelectric cables, capacitive mats, hydraulic and
bending-plate load cells [1]. Hydraulic and bending-plate load cells offer high
accuracy (1–5%) and dynamic range, yet suffer from high installation costs and size
constraints. The piezoelectric and capacitive mat techniques are substantially lower in
cost, yet are less accurate (5–15%) and do not function properly at speeds lower than
20 km/h [2] and [3]. To offer the required accuracy at reduced installation and
maintenance costs, optic fiber-based WIM sensors are now being developed to
improve, complement or even replace the ones currently in use.
Based on the effect of polarization coupling between two orthogonally polarized
eigenmodes of polarization-maintaining fiber, Ansari et al. report on using highly
birefringence polarization-maintaining (HiBi) fiber for dynamic measurement of
pressure with practical ramifications to the determination of weigh-in-motion of
trucks [3]. Navarrete and Bernabeu report a multiple fiber-optic interferometer
consisting of a Mach-Zehnder interferometer configuration with one of its arms
replaced by another Mach-Zehnder interferometer [4]. Cosentino and Grossman
developed a dynamic sensor using the microbend theory to test weight-in-motion [5].
The present work describes the development of a dynamic pressure sensor based
on the Michelson interferometer, which has simple structure, is cost effective and can
potentially offer the high accuracy required for many applications. Special hardware
and system software based on Labview WINDOWS/CVI are designed to implement
the sensor functions, such as eliminating environmental noise, self-triggering of the
test procedure and the fringe number and fringe period simultaneous count. Responses
of the dynamic sensor are studied when subjected to dynamic compressive loads with
different magnitudes and loading rates. Data calibration method is also researched to
calibrate the sensor.
2. The sensor design
2.1. Sensor setup
Fig. 1 illustrates schematically the proposed dynamic pressure sensor system.
Single-mode optical fiber is used as a sensing element to form a Michelson
interferometer. The optoelectronics components of the interferometer consist of a
laser operating at wavelength of 1550 nm, a laser isolator and a photodiode. The
sensor is made of communication grade optical fiber (Corning SMF28). The output
signal from the detector-amplifier is first fed to a special hardware circuits including a
two-order high pass filter, a zero-point detection circuit and a Schmitt Trigger circuit.
The hardware circuits are designed to implement the following functions: (1)
self-diagnose the arrival time of the vehicle to self-trigger the measurement process;
(2) provide function to eliminate the low frequency disturbances, such as temperature
influences and slow changes of the element's performances; (3) provide function to
reduce the noise to a frequency band similar to the useful fringe output of the
Michelson interferometer. One possible source of this noise is caused by vehicles
passing in a near-by lane. Since the output of the Michelson interferometer under
pressure is fringe which can be considered as high frequency signal comparing to the
noise caused by temperature changes, laser and diode performance change and other
low frequency environmental influences, a two-order high pass filter was adopted to
eliminate those low frequency components. A zero-point detection circuit is designed
to change the sine-form fringe to pulse signal for the counter in the computer data
acquisition system to count the fringe number and measure the fringe period. The
self-trigger function is accomplished by the Schmitt circuit. The threshold voltage of
the Schmitt circuit is set according to experiments to distinguish the real fringe signal
caused by vehicles and the pseudo fringe signal caused by small vibration in the test
environment. In practical application, this could be caused by the passing vehicles in
adjoining lanes. System software based on Labview Windows/CVI is designed to set
measurement parameters, control the test procedure and display results.
Full-size image (14K)
Fig. 1. Dynamic fiber optic pressure sensor setup.
2.2. Sensor principle
One arm of the Michelson interferometer is subjected to a distributing dynamic
load Ld(t). The generalized stress–optic relationship between the optical path change
Δl and the strain induced over the gauge length can be derived as Eq. (1)[6]:
(1)
and
(2)
whereP11 and P12 are the Pockels constant; l the length(gauge length) of the
optical fiber within the pressure field; tx, ty and tz correspond to the mechanical and
geometrical property of the optical fiber and the host material-epoxy.
By measuring the deformation of the fiber, the strain in the host material can be
measured. The strain is linearly proportional to the external applied pressure p.
Consider α as a constant of proportionality between the pressure and the change in
length of the fiber Δl, then
(3)
and
(4)
Thus, the output fringe number of the Michelson interferometer is
(5)
whereNf is the output fringe number; λ the wavelength of the laser light.
The fringe period can be deduced by Eq. (6):
(6)
whereTf is the fringe period; t the loading time, approximately equal to the
duration time of the optic fiber fringe.
3. Experimental procedures and results
3.1. Experimental setup
The experimental setup is shown in Fig. 2. A steel chamber is designed to
contain the optical fiber. The optic fiber is sandwiched between two 1 cm thick stiff
rubber pads and glued to one of the two pieces. Rubber pads are necessary for the
protection of fiber from damage. The cross-sectional area of the loading surface for
the steel chamber is 247.1 mm × 24 mm. The gauge length of the optical fiber
pressure sensor is 247.1 mm. A closed-loop materials testing system (MTS) is used
for the application of pressure. Ramp function load with different loading rates and
magnitudes are chosen as dynamic loads to simulate the weight loads caused by
moving vehicles on road.
Full-size image (45K)
Fig. 2. Experimental setup.
Fig. 3 shows a typical loading profile and fringe output of the interferometer
during the duration of the applied ramp function load. As can be seen, the number of
fringes at left corner of Fig. 3 is so great that it is hard to distinguish each fringe. Thus,
another figure on the right is used. The figure is the enlargement of the circled area to
show clearly the fringes. Because of the polarization effect, the amplitudes of fringes
slightly vary. But will not influence the fringe count and period, thus neglected in this
paper. In the experiment, the pressure load is applied to the chamber at 7.5 MPa
increments up to the maximum pressure level of 30 MPa, corresponding to load
increments from 44.5 kN to a maximum of 178 kN. The loading time is applied
starting from 1 to 6 s with an increment of 1 s.
Full-size image (46K)
Fig. 3. Typical loading procedure and fiber optic sensor output waveform.
3.2. Experimental data
Fig. 4 shows the experimental results of the fringe number and fringe period
readouts of the sensor output under different loads. Lines with different signs
represent relations between the sensor's outputs and the maximum amplitudes of the
load under different loading rates.
Full-size image (33K)
Fig. 4. The experimental results: the fringe numbers and fringe periods vs. loads.
The fringe number has a linear relationship with the static load, while the fringe
period has a non-linear one. Note that, the relationship differs under different loading
rates, since with increasing loading rate, the same maximum amplitude load will turn
out to a bigger dynamic load, which causes the increase in fringe number and the
decrease in fringe period. Though both the fringe number and the fringe period are
sensitive to the dynamic load, their sensitive ranges are different. The sensitivity of
the fringe number to load is a constant in the whole testing range. When load is low,
the small change of dynamic load may not be recognized by the fringe number. On
the other hand, the sensitivity of the fringe period to load is not a constant. When the
load is low, the small change of dynamic load corresponds to big change of fringe
period. These two parameters can be used together to give a more precise indication
of the load tested.
Considering Eqs.(5)and(6), the functions between loads and fringe number and
fringe period can be approached as Eqs. (7) and (8), respectively:
(7)
Ls=ksn1Nf+ksn2
(8)
whereksn1, ksn2, kst1 and kst2 are the parameters approached.
3.3. Repeatability of the sensor
Three experimental results under the same loading conditions are compared in
Fig. 5 demonstrating that the dynamic fiber optic pressure sensor has good
repeatability.
Full-size image (21K)
Fig. 5. Illustration of the repeatability of the optic fiber sensor.
3.4. Calibration of the sensor
According to the fringe number and period of the optic fiber sensor output, the
dynamic load and static load of the vehicle passed can be obtained from the
calibration process.
3.4.1. Calibration of the static weight
The function approach method was adopted to calibrate the static weight
measured. It should be mentioned that the load duration time t should also be gained
by the sensor system in order to get the static weight. Usually it is easy to get. The
following steps are taken to calibrate the static weight:
Step 1: Using the function approach method to approach the measured data when
the sensor experienced the different static applied load with the same loading rate.
Linear function and power function are adopted for the fringe number data and fringe
period data. The approached results are shown in Table 1.
Table 1.
Step 1 approached results
Approached kst1, kst2 for
Approached ksn1, ksn2 for
relation between the largest static
relation between the largest static
load (Ls) and fringe period (Tf),
load (Ls) and fringe number (Nf),
Ls = ksn1Nf + ksn2
Loading
time
ksn1
t = 6
t = 5
t = 4
t = 3
t = 2
t = 1
0.028
0.0272
0.0261
0.0235
0.0225
0.0228
ksn2
−10.81
−12.74
−12.773
−5.4984
−5.8752
−9.8251
kst1
256.58
249.09
209.11
164.96
126.49
53.903
kst2
−1.0057
−1.1759
−1.2119
−1.331
−1.8503
−1.6376
Step 2: Go on using six-order polynomial function to approach the ksn1 = f1(t),
ksn2 = f2(t), kst1 = f3(t) and kst2 = f4(t), here the fn(t) are the fitted polynomial functions.
Step 3: Using ksn1 = f1(t), ksn2 = f2(t), kst1 = f3(t) and kst2 = f4(t) obtained from step
2 together with Eqs. (7)and(8) to calibrate the fringe number and fringe period
measured to static load.
Due to the limitation of paper length, only the calibration results when t = 1 and
4 s are described. In Fig. 6, compared with the real applied static load, the precision of
the researched sensor is 5% higher using fringe number readout and 15% higher using
fringe period readout.
Full-size image (31K)
Fig. 6. Static load calibration results.
3.4.2. Calibration of the dynamic weight
As shown in Eq. (5), the fringe number has a linear relation with the dynamic
load. This relationship can be described as
(9)
Ld=kdn1Nf+kdn2
wherekdn1 and kdn2 are the parameters approached.
The dynamic weight calibration has been developed based on the following
assumption. When the loading speed is very slow, the dynamic weight the sensor
subjected to can be assumed to be the same as the static weight applied with the slow
loading speed. In the experiments, when loading time is 6 s, the largest dynamic load
the sensor subjected to is considered to be equal to the largest static load applied. So
the proportionality constants kdf1 and kdf2 between the fringe number and the dynamic
load can be calculated using data of t = 6 s, namely
(10)
kdf1=ksf1|t=6=0.028
(11)
kdf2=ksf2|t=6=−10.81
Therefore, the dynamic loads can be calculated using Eq. (9), shown in Fig. 7.
Fig. 7 shows that the same largest static loads under different loading rates produce
different dynamic loads. The higher the loading rate is, the bigger the dynamic load is.
Full-size image (17K)
Fig. 7. Dynamic load calibrated results using fringe number readout.
To the fringe period readout, since the parameters Ksp1 and Ksp2 are not constants
when sensor experiences different dynamic loads, it cannot be calibrated using the
above mentioned method.
4. Conclusion
A dynamic optic fiber pressure sensor based on the Michelson interferometer is
introduced. This sensor reported here has the advantages of simplicity and low cost.
Experiments show that both output fringe number and fringe period can be used to
indicate the load. A calibration technology is developed to calibrate the sensor. Both
the dynamic weight and static weight of the passing vehicle can be obtained. The
findings resulted from these studies have developed an understanding for the behavior
of interferometer sensor under dynamic compressive states of stress and are
fundamental to the application of this kind of sensor to monitor the truck vehicle
weights while in motion.
译文
WIM系统中以光纤为基础的动态压力传感装置
沉方元，FahardAnsarib，刘小慧和杨昭
航空重点实验室，智能材料与结构，南京航空航天大学29有道街，南京210016，中国
土木与材料工程学院，芝加哥伊利诺伊大学部，伊利诺伊州和IL 60607，美国
2004年8月16日收到;
2004年11月10日接受。
2004年12月15日在线。
摘要：
在这里描述一种光纤为基础的动态压力传感装置来测量车辆荷载的方法。在研究报告中，是专门设计的迈克尔逊干涉仪的硬件和所设计的软件遭受不同程度的动态压缩载荷和装载率的实验。实验结果表明，输出文件数据和图像均可以用来表示动态负载。于是校准传感器的校准技术被提了出来，可以得到无论是动态的荷载和通过车辆的静态荷载。该调查结果显示，在这些研究开发的干涉下的动态压缩应力状态传感器的作用是了解基本的在光纤应用的方案—光纤传感器监测车车辆荷载。
关键词：光纤传感器，动态压力，动荷载，硬件和软件
文章概要
1．介绍
2．传感器的设计
2.1．传感器安装
2.2．传感器的工作原理
3．程序实验和结果
3.1．实验装置
3.2．实验数据
3.3．重复试验
3.4．传感器的校准
3.4.1．标定的静态荷载
3.4.2．标定的动荷载
4．结论
1介绍
过去的几十年中需要权衡运动荷载，特别是在交通控制中，已经大幅增加。车辆的动荷载几个技术现在已经应用于包括压电电缆，电容垫，液压和弯曲载荷细胞[1]的使用。液压和弯板称重传感器会提高准确度（1-5％）和其动态范围，但是因其高安装成本和规模的限制而无法广泛应用。压电和电容垫技术会使成本大幅降低，但却不太准确（5-15％），在速度低于二十公里每小时[2]和[3]的无法正常工作。为了提供降低安装和维护成本的方法，以光纤为基础的WIM系统动态压力传感装置现正开发的改进，未来可能取代目前使用的装置。
安萨里等人基于光纤偏振，做出了偏振耦合的两个正交偏振效应的模态参数。做出了对压力动态测量的高双折射偏振（HiBi）与运动货车轴重动荷载的实际后果对比的报告[3]。纳瓦雷特和伯纳乌报告是对多光纤干涉仪的马赫曾德尔干涉仪结构组成用另一马赫曾德尔干涉仪取代作为研究内容[4]。科森蒂诺和格罗斯曼开发利用的是动态传感器微弯理论容重在运动中的结果[5]。
本节描述了一个基于迈克尔逊干涉仪的动态压力传感器的研制，它具有结构简单，成本低，可提供高精度等许多潜在优点。它具有基于Labview Windows /
CVI开发的特殊的硬件和系统软件，旨在实现消除环境噪声的影响，例如，自我引发的测试程序和条纹数及附带期间同时计数传感器的功能。传感器的动态响应进行研究时，受到了不同程度的负荷率和动态压缩载荷的影响。并有研究校准传感器数据标定方法。
2传感器的设计
2.1传感器安装
图 1说明了拟定的动态压力传感器系统。单模光纤是用来作为传感元件以构成一个迈克尔逊干涉仪。该干涉仪的光电组件包括一个长度在1550纳米的激光隔离器和一个光电二极管激光波长。该传感器是由光纤通信级（康宁SMF28）。从检测器，放大器的输出信号是先输入到一个特殊的硬件电路，包括一个2阶高通滤波器，一个零点检测电路和触发器电路。硬件电路的设计实施下列功能：（1）自我诊断的车辆到达时间自行触发测量过程;（2）提供消除低频干扰的功能，如温度的影响而使该元素的变化缓慢，;（3）提供减少噪音的功能，频率波段类似的迈克尔逊干涉仪的输出有用的信息。其中的一个可能的噪声源是接近路过车辆的线。自迈克尔逊干涉仪的输出是在压力下边缘可以作为高频比较受温度变化，激光二极管的性能和低频率的变化和其它环境影响，一个2阶高通滤波器所造成的噪音信号通过的消除这些低频成分。
一个零点检测电路的设计更改计算机中的数据采集系统反正弦表边缘，以脉冲信号进行计数的条纹数和测量的边缘时期。自触发功能是通过施密特电路，施密特电路的阈值电压设置根据实验来区分在测试环境造成的微小振动引起的真正的条纹信号的车辆和假边缘信号。在实际应用中，这可能是用于在相邻通道的过往车辆。系统软件基于Labview和Windows/CVI的目的是设置测量参数，控制测试程序，并以现实结果为基础。
图1光纤压力传感器的动态设置
3实验程序和结果
3.1实验装置
实验装置如图2所示。一个钢室的设计研究包含光纤。该光纤是夹在两个1厘米厚的硬橡胶垫和粘在两个硬橡胶垫件之一上。橡胶垫是为保护光纤不受损坏。该室的装载钢材表面的横截面面积为247.1毫米× 24毫米。该光纤压力传感器测量长度为247.1毫米。一个封闭的回路材料的测试系统（MTS）是用于压力的加载。在不同加载速率和加载程度下的活荷载是动态负载模拟加载的，荷载是由道路车辆造成的。
图2实验装置。
图3显示了应用在斜坡函数加载的时间，是由常用的装载配置仪器及附带的干涉仪的输出。可以看出在边缘的左上角的数字。图3是太大以至于很难区分每个边缘的数字。因此，数字有另一个用途，这个数字是该圈的地方扩大，显示清晰的边缘的数字。由于极化效应，条纹的振幅略有不同。但不会影响计数和时间，因此，在本文忽视。在实验中，负荷的压力在7.5兆帕时增量会多达30 MPa，相应的最大压力水平从44.5KN增大为最大的178KN。装载时间是从1开始从第1至6秒递增。
图3典型的装载程序和光纤传感器的输出波形。
3.2。实验数据
图4显示的条纹数的实验结果和不同负荷下的传感器输出读数的时期。不同的标志线代表不同加载速率之间的传感器的输出和负载的最大振幅的关系。
图4实验结果：边缘的数字和荷载。
附加的数字，与静载荷的线性关系，而边缘地方有一个非直线的。请注意，这种关系随着不同加载速率不同，因为相同的最高振幅负荷将变成一个更大的动态负载，这将导致在条纹数的增加和边缘减少。虽然双方的条纹数及附加时期敏感的动态负载，其敏感范围是不同的，加载时灵敏度是在整个测试范围不变。当荷载低，动态荷载小的变化可能无法识别的条纹数。另一方面，该时期的边缘敏感负荷不是一个常数。当荷载降低，动态负载小的变化对应的边缘时期大的变化。这两个参数可以一起使用提供了更精确的荷载指标进行测试。
考虑均衡器（5）及（6），与压力的条纹数及附加时期的功能
都可以接近的均衡器（7）及（8），分别为：
(7)
Ls=ksn1Nf+ksn2
(8)
其中参数ksn1，ksn2，kst1和kst2接近，参数ksn1，ksn2，kst1和kst2接近。
3.3重复性的传感器
图5是三个相同加载条件下的实验结果进行了比较，动态展示光纤压力传感器具有良好的重复性。
图5光纤传感器的重复性
3.4。该传感器的校准
根据条纹数和光纤传感器的输出期间，可从动态车辆载荷和静载荷传递的过程中校准。
3.4.1静态重量校准测量
该函数逼近法是采用静态重量校准测量。应该一提的是负载持续时间t也应获得的传感器系统，以获得静态重量。通常很容易得到的。采取以下步
第1步：使用函数方法测量数据时，经历了不同的传感器适用于具有相同的静态负荷率时。线性函数和幂函数采用的条纹数数据及附加期间的数据。在接触的结果列于表1。
表1第1步接近结果
接近ksn1，其中最大静接近kst1，其中最大静加载时间
负荷（LS）和条纹数（线性负荷（LS）和边缘期（TF）关系ksn2）,
Ls =
ksn1Nf +
ksn2
的关系kst2，
ksn1
0.028
0.0272
0.0261
0.0235
0.0225
0.0228
ksn2
−10.81
−12.74
−12.773
−5.4984
−5.8752
−9.8251
kst1
256.58
249.09
209.11
164.96
126.49
53.903
kst2
−1.0057
−1.1759
−1.2119
−1.331
−1.8503
−1.6376
t = 6
t = 5
t = 4
t = 3
t = 2
t = 1
第2步：继续使用6阶多项式函数接近ksn1 = F1类（吨），ksn2 = f2类（吨），kst1 =三级方程式（t）和kst2 = F4类（吨），这里的新的重量（吨）是拟合多项式函数。
第3步：使用ksn1 = F1类（吨），ksn2 = f2类（吨），kst1 =三级方程式（t）和kst2 = F4类（吨）均衡器，从步骤2一起获得（7）及（8）校准条纹数及附带期间测得静载荷。
由于纸张的长度的限制，只有当T =校准结果1和4秒在图6中描述。与实际应用静载荷相比，该研究传感器精度比使用条纹数读出提高5％，比使用附带期间读数高15％。
图。 6。静负荷测量结果。
3.4.2权重的动态测量
式（5）所示，条纹数已与动态负载呈线性关系。这种关系可以说是
(9)
Ld=kdn1Nf+kdn2
其中参数kdn1和kdn2接近。
动态校准权重的开发是基于以下的假设。当加载速度很慢，动态重量传感器受到可以认为是与加载速度慢的静态重量相同。在实验中，当加载时间为6秒，最大的动态负载传感器被认为是平等的最大静载荷应用。因此，相称常数kdf1和kdf2之间的条纹数和动态负载可以用吨来计算，即数据
(10)
kdf1=ksf1|t=6=0.028
(11)
kdf2=ksf2|t=6=−10.81
因此，可以用计算式（9）动态加载。，如图7所示。图7显示，根据不同的负荷率相同的最大静载荷产生不同的动态负载。较高的
负荷率，更大的动态负载。
图7动态负载校准结果使用条纹数读数。
要读出边缘时期以外的数据，参数Ksp1和Ksp2不是常数时，以不同的动态负载传感器的经验，它不能被使用上述方法校准。
4结论
这里报告在使用一个光纤压力传感器的基础上，介绍了迈克尔逊干涉仪。该传感器具有简单和成本低等优点。实验结果表明，两个输出条纹数及附加期可用于指示荷载。通过校正技术来开发校准传感器。无论是动态的车辆荷载，和静态荷载可以得到。根据上述这些研究，国家制定了动态压缩干涉下的应力传感器，这一类传感器可以应用到监测车在行驶中车辆的重量。
发布评论