| The rich coal resources buried deep underground of kilometers have prompted China’s coal gradually shift into deep mining.The formation lithology of deep vertical shaft is various and the geological structure is complex.The larger the vertical depth of shaft is,the more likely the interaction of rock,mud and water weak interlayer will lead to the crack of shaft lining.If let cracks continue,the shaft lining will break and the structure will fail,resulting in major mine disasters.In order to realize the visual quantitative description of deep vertical shaft lining,the research on real-time monitoring and pre-warning system is carried out.In view of the difficulty of obtaining the load variation on the deep vertical shaft lining,the acoustic mode eigenfrequency of optical fiber is decoupled by different frequency shifts which are affected by both temperature and strain,so as to realize the simultaneous measurement of stress field and temperature field of shaft lining by one optical fiber.Aiming at the difficulty in locating cracks on the shaft lining and the unknown cracks’ geometry and change speed,the image processing technology is used to extract the crack features.Finally,the crack and strain characteristics are integrated to establish a prediction model to realize the visual monitoring and pre-warning of the shaft lining.The main work of this paper is as follows:(1)According to the special environment of the deep vertical shaft,the noise model of the shaft lining image is designed,and the denoising method of the deep vertical shaft lining image is put forward with the noise characteristics to realize the denoising of the shaft lining image.In the low illuminance environment of deep vertical shaft,the sampled shaft lining image including salt-and-pepper noise,gaussian noise,and multiplicity and other noises introduced by photoelectric conversion and other factors.The noise type of shaft lining image is far more complex than the general image.According to the characteristics of single background and uniform pixel value distribution of the shaft lining image,salt and pepper noise is detected first by using its extreme value characteristics,and then filter noise directly without changing the signal pixel value.After the salt and pepper noise is filtered out,the concentration of other noises in the shaft lining image is unknown.A blind denoising model is proposed by using the convolution neural network.In view of the fact that it is impossible to obtain the clean shaft lining image in the deep vertical shaft environment,the training sample is constructed with the noise image as the label.After training the denoising model,the blind denoising of the shaft lining image can be realized,and the crack characteristics of the shaft lining can be preserved.(2)Aiming at the problem of locating cracks on the shaft lining,a classification method which can recognize cracks on the shaft lining images is proposed by using the convolution neural network,and the position of cracks on the shaft lining is automatically located.In order to effectively utilize the crack information on the shaft lining image,a pooling method is designed and used to build the shaft lining image classification model.By studying the tagging method of shaft lining image,an image database for training shaft lining image classification model is established.Through training and testing the classification model on the shaft lining image database,the automatic recognition of cracks on the shaft lining image is realized.Combining with the sampling time,velocity and acceleration of the camera,the specific location of cracks on the shaft lining is located.(3)Through untying the couple between the stress and temperature,the temperature and strain of the shaft lining can be simultaneously detected by a single optical fiber,so as to solve the problem of obtaining the load variation on the deep vertical shaft lining.When the comprehensive stress caused by the interaction of rock,mud and water around the deep vertical shaft and the temperature change exceeds the ultimate strength of the shaft lining concrete,it will cause the shaft lining crack,so it is necessary to real-time monitor the strain and temperature on the shaft lining.The temperature and strain can be decoupled by using different frequency shift produced by two different acoustic modes with different sensitivity to temperature and strain.The temperature field and stress field of the shaft lining can be measured simultaneously by using a single optical fiber to obtain the temperature and strain of the shaft lining.(4)In order to quantitatively describe the situation of the deep vertical shaft lining,the evaluation model and prediction model are established by synthesizing characteristic indexes of shaft lining cracks and stress,so as to realize the evaluation and pre-warning of the shaft lining fracture index.In order to establish the mapping relationship between shaft lining fracture and shaft lining strain,cracks’ growth,8 indexes representing cracks’ geometry,change speed and 2 indexes representing strain size and change speed are extracted.Through multi-information fusion of the 10 indexes,an evaluation model and prediction model of shaft lining fracture index are established to realize the prediction of the crack development trend and the shaft lining fracture index in the future.(5)With the distributed optical fiber and the explosion-proof cameras as the sensors,the embedded system as the lower system and tower graphic workstation as upper system,the shaft lining monitoring and pre-warning system is designed.Through the network switches,the lower system obtains the image and the stress of the shaft lining collected by the sensors,so as to realize the evaluation and prediction of WFI right on-site.The upper system software is programmed with C++,which can realize the preprocessing of the shaft lining image and the online evaluation and prediction of WFI.Figure[79]table[17]reference[162]... |
