Diagnosis and analysis method for cross-fault tunnel diseases
阅读说明：本技术 跨断层隧道病害的诊断分析方法 (Diagnosis and analysis method for cross-fault tunnel diseases ) 是由 王峥峥 张玉堂 贾佳欣 赵敏 覃晖 唐佳 于 2019-10-30 设计创作，主要内容包括：本发明公开了一种跨断层隧道病害的诊断分析方法,其包括获取当前时间,在满足时间要求时,第二摄像机采集第一图像,并判断第一图像和参考图像中的第一摄像机位置是否处于同一位置,若是启动第二摄像机采集第二图像进行错台判断,当存在错台且错台的宽度大于一定值时,将将采集的第一图像、第二图像和隧道在断层处出现严重错台的信息压缩后发送给道路养护部；当存在错台且错台的宽度小于一定值时,判断第二摄像机两次采集第一图像之间是否存在设定比例的湿度传感器持续上传湿度信息,若是则对第二图像进行裂纹检测,并采集的第一图像、第二图像和裂纹检测结果及隧道在断层处出现错台和渗水的危害信息压缩后发送给道路养护部。(The invention discloses a cross-fault tunnel disease diagnosis and analysis method, which comprises the steps of obtaining current time, acquiring a first image by a second camera when the time requirement is met, judging whether the positions of the first camera in the first image and a reference image are in the same position, if so, starting the second camera to acquire a second image to carry out slab staggering judgment, and compressing the acquired first image, the acquired second image and information of serious slab staggering of a tunnel at a fault and then sending the compressed information to a road maintenance part when the slab staggering exists and the width of the slab staggering is greater than a certain value; when the slab staggering exists and the width of the slab staggering is smaller than a certain value, whether humidity information is continuously uploaded by a humidity sensor with a set proportion exists between the two collected first images of the second camera is judged, if yes, crack detection is carried out on the second image, and the collected first image, the collected second image, the collected crack detection result and the damage information of slab staggering and water seepage at the fault of the tunnel are compressed and then sent to a road maintenance part.)
1. The cross-fault tunnel disease diagnosis and analysis method is characterized by comprising a diagnosis and analysis system for implementing the diagnosis and analysis method, wherein the diagnosis and analysis system comprises a single chip microcomputer, a first camera and a second camera which are respectively installed on two sides of a tunnel, which are opposite to a fault plane, and a plurality of humidity sensors which are installed on two sides of the tunnel, which are opposite to the fault plane;
the first camera, the second camera, the humidity sensor and the single chip microcomputer are all connected with a power module connected with commercial power;
the first camera is arranged on one side of the tunnel in the width direction through a support and used for collecting image information of the tunnel at the fault plane; the second camera is arranged on the opposite side of the first camera through the bracket, and the first camera is positioned in the view field of the second camera when not started;
the method for diagnosing and analyzing the cross-fault tunnel diseases by the diagnostic and analysis system comprises the following steps:
s1, acquiring the current time;
s2, judging whether the current time and the time of the last image acquisition of the second camera are greater than or equal to the preset time, if so, entering the step S3, otherwise, returning to the step S1;
s3, starting a second camera to acquire image information of the first camera, and receiving a first image;
s4, judging whether the position of the first camera in the stored reference image is at the same position as the position of the first camera in the first image, if so, returning to the step S1, otherwise, entering the step S5;
s5, starting a first camera to collect a second image of the tunnel at the fault, receiving the second image, inputting the second image into the first convolution neural network model to perform slab staggering detection, and outputting a slab staggering detection result;
s6, judging whether the wrong station detection result indicates that the wrong station exists in the second image information, if so, entering a step S7, otherwise, updating the reference image by adopting the first image, and returning to the step S1;
s7, respectively calculating the distance between the camera of the first camera in the reference image and the first image and the upper edge of the image;
s8, judging whether the difference between the two distances is smaller than a preset distance, if so, entering a step S9, otherwise, entering a step S12;
s9, judging whether a humidity sensor with a set proportion continuously uploads humidity information exists between the two first images acquired by the second camera, if so, entering a step S10, otherwise, updating the reference image by using the first image, and returning to the step S1;
s10, inputting the second image information into a second convolutional neural network model for crack detection, and outputting a crack detection result;
s11, compressing the acquired first image, second image and crack detection result and the damage information of dislocation and water seepage of the tunnel at the fault, sending the compressed information to a road maintenance part, and returning to the step S1;
s12, compressing the acquired first image, second image and information of severe dislocation of the tunnel at the fault, sending the compressed information to a road maintenance department, and returning to the step S1.
2. The method for diagnosing and analyzing the cross-fault tunnel disease according to claim 1, wherein a bracket for mounting the first camera comprises a vertical rod and a support plate for mounting the first camera, the support plate is hinged to the vertical rod, an electric push rod which is inclined downwards and is connected with the single chip microcomputer is mounted on the vertical rod above the hinged position of the support plate, the tail end of the push rod of the electric push rod is fixedly connected with the support plate, and when the electric push rod moves to the minimum stroke, the first camera is located in the view field of the second camera;
when the electric push rod is in the minimum stroke stage, the camera of the first camera inclines upwards, and when the electric push rod is in the maximum stroke stage, the camera of the first camera inclines downwards; after the bracket for mounting the first camera comprises the electric push rod, step S5 further comprises starting the electric push rod while starting the first camera, and after the electric push rod moves to the maximum stroke, closing the first camera and controlling the stroke of the electric push rod to return to the initial position; the first image comprises the graphic information of the top wall, the side wall and the tunnel pavement of the tunnel, which is acquired by the first camera.
3. The method for diagnosing and analyzing the cross-fault tunnel disease according to claim 1, wherein the training method of the first convolutional neural network model comprises the following steps:
a1, collecting a plurality of cement pavement slab staggering images, a plurality of vertical wall slab staggering images and a plurality of shield segment slab staggering images, and marking slab staggering areas in all the slab staggering images;
a2, preprocessing all the marked slab staggering images, and then initializing the weight of the convolutional neural network;
a3, inputting all preprocessed staggered platform images, transmitting the images forward through a convolutional layer, a downsampling layer and a full-link layer to obtain output values, and calculating errors between the output values and target values of a convolutional neural network;
a4, when the error is larger than the expected value, the error is transmitted back to the convolutional neural network, and the errors of the full connection layer, the down sampling layer and the convolutional layer are calculated in sequence;
and A5, when the error is equal to or less than the expected value, adopting the error to update the weight of the neural network, and ending the training to obtain a first convolution neural network model.
4. The method for diagnosing and analyzing a cross-fault tunnel disease according to claim 1, wherein the training method of the second convolutional neural network model comprises the following steps:
b1, manufacturing a plurality of arc templates with different radians and similar colors to the inner surface of the tunnel, and drawing cracks with different types and sizes on each arc template;
b2, mounting a plurality of arc templates at different positions of a circle in the width direction of the tunnel, and collecting crack images of the inner surface of the tunnel, to which the arc templates are adhered, by using a rotating camera;
b3, changing the position of the arc-shaped plate in one circle in the width direction of the tunnel, and then continuously adopting a rotating camera to collect crack images of the inner surface of the tunnel adhered with the arc-shaped template;
b4, judging whether the number of the acquired crack images is larger than the total number of the preset images, if so, executing the step B5, otherwise, returning to the step B3;
b5, marking cracks in all the crack images, preprocessing all the marked crack images, and then initializing the weight of the convolutional neural network;
b6, inputting all preprocessed crack images, transmitting the preprocessed crack images forwards through a convolution layer, a down-sampling layer and a full-connection layer to obtain an output value, and calculating the error between the output value and a target value of the convolution neural network;
b7, when the error is larger than the expected value, the error is transmitted back to the convolutional neural network, and the errors of the full connection layer, the down sampling layer and the convolutional layer are calculated in sequence;
and B8, when the error is equal to or less than the expected value, updating the weight of the neural network by using the error, and finishing training to obtain a second convolutional neural network model.
5. The method for diagnosing and analyzing the cross-fault tunnel disease according to claim 3 or 4, wherein the preprocessing of the image comprises sequentially performing size adjustment, contrast transformation and image distortion processing on the image.
Technical Field
The invention relates to the field of geological disaster monitoring, in particular to a diagnosis and analysis method for a cross-fault tunnel disease.
Background
Faults are architectural forms that develop extensively in architectural movement. It is of different sizes and scales, with the smaller size less than one meter, and the larger size hundreds of meters, up to kilometers. But all disrupt the continuity and integrity of the formation. On the fault zone, rocks are broken and are easy to be weathered and eroded. Often developing into valleys along fault lines, sometimes in springs or lakes.
With the increasing force of constructing the great southwest in China, in the construction of heavy projects such as high-speed railways, expressways, diversion tunnels and the like, particularly long tunnels, a plurality of fracture zones are inevitably required to pass through. For example, the Dang lawn tunnel is 10km long and passes through 9 fracture zones; the mud hill tunnel with elegant filtration and high speed is 10km long and passes through 15 fracture zones; the tunnel of the Dian Zhong Water-diversion incense burner mountain is about 63km long and passes through 16 main fracture zones, wherein 40 percent of the tunnels are gliding fault zones. At present, most faults generate a certain amount of peristalsis every year, the peristalsis is a few centimeters every year, more than 10cm every year, and the peristalsis is further aggravated when an earthquake occurs.
However, China builds a large number of cross-fault tunnels in the southwest, and the southwest is an earthquake-prone area, so that the creeping of faults is aggravated, at present, corresponding monitoring measures are not adopted for cross-fault tunnels by cross-fault tunnel road maintenance departments, and a road is closed to take remedial measures when tunnel collapse or serious safety problems occur frequently, and the mode brings great potential safety hazards to road safety operation.
Disclosure of Invention
Aiming at the defects in the prior art, the diagnosis and analysis method for the cross-fault tunnel diseases solves the problem that the damage generated by the cross-fault tunnel cannot be monitored in the prior art.
In order to achieve the purpose of the invention, the invention adopts the technical scheme that:
the diagnosis and analysis system comprises a single chip microcomputer, a first camera and a second camera which are respectively arranged on two sides of the tunnel opposite to the fault plane and a plurality of humidity sensors which are arranged on two sides of the tunnel opposite to the fault plane;
the first camera, the second camera, the humidity sensor and the single chip microcomputer are all connected with a power module connected with commercial power;
the first camera is arranged on one side of the tunnel in the width direction through a support and used for collecting image information of the tunnel at the fault plane; the second camera is arranged on the opposite side of the first camera through the bracket, and the first camera is positioned in the view field of the second camera when not started;
the method for diagnosing and analyzing the cross-fault tunnel diseases by the diagnostic and analysis system comprises the following steps:
s1, acquiring the current time;
s2, judging whether the current time and the time of the last image acquisition of the second camera are greater than or equal to the preset time, if so, entering the step S3, otherwise, returning to the step S1;
s3, starting a second camera to acquire image information of the first camera, and receiving a first image;
s4, judging whether the position of the first camera in the stored reference image is at the same position as the position of the first camera in the first image, if so, returning to the step S1, otherwise, entering the step S5;
s5, starting a first camera to collect a second image of the tunnel at the fault, receiving the second image, inputting the second image into the first convolution neural network model to perform slab staggering detection, and outputting a slab staggering detection result;
s6, judging whether the wrong station detection result indicates that the wrong station exists in the second image information, if so, entering a step S7, otherwise, updating the reference image by adopting the first image, and returning to the step S1;
s7, respectively calculating the distance between the camera of the first camera in the reference image and the first image and the upper edge of the image;
s8, judging whether the difference between the two distances is smaller than a preset distance, if so, entering a step S9, otherwise, entering a step S12;
s9, judging whether a humidity sensor with a set proportion continuously uploads humidity information exists between the two first images acquired by the second camera, if so, entering a step S10, otherwise, updating the reference image by using the first image, and returning to the step S1;
s10, inputting the second image information into a second convolutional neural network model for crack detection, and outputting a crack detection result;
s11, compressing the acquired first image, second image and crack detection result and the damage information of dislocation and water seepage of the tunnel at the fault, sending the compressed information to a road maintenance part, and returning to the step S1;
s12, compressing the acquired first image, second image and information of severe dislocation of the tunnel at the fault, sending the compressed information to a road maintenance department, and returning to the step S1.
The invention has the beneficial effects that: the method is applied to a tunnel crossing a fault, particularly a fault plane is located in the width direction of the tunnel, the tunnel at the fault is subjected to image acquisition at each period of time, whether the fault creeps is initially judged through the relative positions of a first camera and a second camera, images on the surface of the tunnel are acquired after the initial judgment, whether the tunnel has a dislocation phenomenon is checked, so that the fault creep is confirmed, if the dislocation does not occur, the fact that the relative position change of the first camera and the second camera does not occur and the fault creep is not caused is shown, and at the moment, a reference image stored in the singlechip is updated, so that the subsequent accurate detection is guaranteed.
If the wrong platform appears and shows that the wrong platform has appeared in fault department, when the wriggling takes place in the fault, not as long as the wrong platform appears will cause the influence to the safe handling in tunnel, this scheme can be close step judge the displacement size that the wrong platform took place through two intervals of calculation, if the wrong platform displacement that appears is than hour, can detect through looking over whether there is infiltration and crackle in fault department, tunnel way pipe piece erosion can aggravate because the infiltration in addition to the crackle appears in the tunnel, and seriously influence tunnel section of jurisdiction quality, road maintenance portion is required to be informed this moment and is maintained.
If the dislocation displacement is large, the road maintenance part is informed to maintain, and the tunnel segment structure is likely to be loosened due to large dislocation when the dislocation displacement is large, so that the tunnel segment is easy to fall off to cause safety accidents.
Drawings
FIG. 1 is a schematic diagram of a structure in which parts of a diagnostic analysis system are installed on a tunnel at a fault.
Fig. 2 is a schematic structural view of a bracket for mounting a first camera.
FIG. 3 is a flow chart of a cross-fault tunnel disease diagnosis and analysis method.
Wherein, 1, a tunnel; 2. fault plane; 3. a first camera; 4. a second camera; 5. a support; 51. a vertical rod; 52. a support plate; 53. an electric push rod.
Detailed Description
The following description of the embodiments of the present invention is provided to facilitate the understanding of the present invention by those skilled in the art, but it should be understood that the present invention is not limited to the scope of the embodiments, and it will be apparent to those skilled in the art that various changes may be made without departing from the spirit and scope of the invention as defined and defined in the appended claims, and all matters produced by the invention using the inventive concept are protected.
As shown in fig. 1, the diagnostic analysis system implementing the diagnostic analysis method includes a single chip microcomputer, a
the
the
According to the scheme, the installation positions of the
As shown in fig. 2, in an embodiment of the present invention, the
the end of the
After the
Referring to fig. 3, fig. 3 is a flowchart illustrating a diagnosis and analysis method of a cross-fault tunnel 1 defect, which includes steps S1 to S12, as shown in fig. 3.
In step S1, the current time is acquired; because the creeping of the fault is a slow geological change process, the creeping is generally several centimeters in one year, so frequent monitoring is not needed during monitoring, and the scheme preferably monitors once every month in summer, once every half and a month in spring and once every 2 months in autumn and winter.
In step S2, it is determined whether the current time and the time of the last image capture by the second camera 4 are greater than or equal to the preset time, if yes, the process proceeds to step S3, otherwise, the process returns to step S1;
in step S3, the second camera is started to capture the image information of the
in step S4, it is determined whether the position of the
in step S5, the
In one embodiment of the invention, the training method of the first convolution neural network model comprises the following steps:
a1, collecting a plurality of cement pavement slab staggering images, a plurality of vertical wall slab staggering images and a plurality of shield segment slab staggering images, and marking slab staggering areas in all the slab staggering images;
a2, preprocessing all the marked slab staggering images, and then initializing the weight of the convolutional neural network;
a3, inputting all preprocessed staggered platform images, transmitting the images forward through a convolutional layer, a downsampling layer and a full-link layer to obtain output values, and calculating errors between the output values and target values of a convolutional neural network;
a4, when the error is larger than the expected value, the error is transmitted back to the convolutional neural network, and the errors of the full connection layer, the down sampling layer and the convolutional layer are calculated in sequence;
and A5, when the error is equal to or less than the expected value, adopting the error to update the weight of the neural network, and ending the training to obtain a first convolution neural network model.
In step S6, it is determined whether the detection result of the wrong station indicates that the wrong station exists in the second image information, if yes, step S7 is performed, otherwise, the reference image is updated by using the first image, and the process returns to step S1;
in step S7, the distances from the camera of the
In step S8, determining whether the difference between the two distances is smaller than a preset distance, if so, going to step S9, otherwise, going to step S12;
in step S9, it is determined whether there is a humidity sensor with a set ratio continuously uploading humidity information between the two times of acquiring the first image by the second camera 4, if so, step S10 is performed, otherwise, the first image is used to update the reference image, and step S1 is returned; the present embodiment preferably sets the ratio to 40%.
In step S10, the second image information is input to the second convolutional neural network model for crack detection, and a crack detection result is output.
In implementation, the method for training the second convolutional neural network model preferably includes:
b1, manufacturing a plurality of arc templates with different radians and similar colors to the inner surface of the tunnel 1, and drawing cracks with different types and sizes on each arc template;
b2, installing a plurality of arc-shaped templates at different positions of a circle in the width direction of the tunnel 1, and acquiring crack images of the inner surface of the circle of the tunnel 1 adhered with the arc-shaped templates by a rotating camera.
The arc-shaped template is made of flexible materials, and is further preferably made of plastic with the thickness of 0.1-0.2 cm, so that the arc-shaped template can be better attached to the pipe edges of the tunnel 1 with different radians, so that the cracks deform to form diversified cracks; the thinner the arc-shaped template is, the closer the crack shooting effect is to the real crack.
B3, changing the position of the arc-shaped plate in one circle in the width direction of the tunnel 1, and then continuously adopting a rotating camera to collect crack images of the inner surface of the tunnel 1 adhered with the arc-shaped template;
b4, judging whether the number of the acquired crack images is larger than the total number of the preset images, if so, executing the step B5, otherwise, returning to the step B3;
according to the scheme, the training pictures are obtained in the steps B1-B4, so that cracks in the training pictures can be infinitely close to real cracks, and cracks in different forms (due to the fact that radians of all positions of the tunnel 1 duct pieces are different, arcs are pasted on the tunnel 1 duct pieces) can be captured as many as possible by changing the installation positions of the arc-shaped plates, so that samples which are recorded as full as possible are guaranteed, and the crack recognition rate in the subsequent detection process is improved.
B5, marking cracks in all the crack images, preprocessing all the marked crack images, and then initializing the weight of the convolutional neural network;
b6, inputting all preprocessed crack images, transmitting the preprocessed crack images forwards through a convolution layer, a down-sampling layer and a full-connection layer to obtain an output value, and calculating the error between the output value and a target value of the convolution neural network;
b7, when the error is larger than the expected value, the error is transmitted back to the convolutional neural network, and the errors of the full connection layer, the down sampling layer and the convolutional layer are calculated in sequence;
and B8, when the error is equal to or less than the expected value, updating the weight of the neural network by using the error, and finishing training to obtain a second convolutional neural network model.
In step S11, compressing the collected first image, second image, crack detection result and damage information of slab staggering and water seepage of the tunnel 1 at the fault, and sending the compressed information to the road maintenance department, and returning to step S1;
in step S12, the acquired first image, second image and information that a severe slab staggering occurs at the fault of the tunnel 1 are compressed and sent to the road maintenance department, and the process returns to step S1.
In this embodiment, after the
The first convolution neural network model and the second convolution neural network model relate to image preprocessing in the training process, and the image preprocessing comprises the steps of sequentially carrying out size adjustment, contrast transformation and image distortion processing on an image.
In conclusion, the diagnosis and analysis method provided by the scheme can monitor slab staggering, cracks and water seepage which occur near the
- 上一篇：一种医用注射器针头装配设备
- 下一篇：混凝土护栏的裂缝自动探伤设备及其使用方法