CN-122023853-A - Tunnel blasting super-undermining analysis method based on radar point cloud

CN122023853ACN 122023853 ACN122023853 ACN 122023853ACN-122023853-A

Abstract

The invention relates to the field of tunnel construction, and particularly discloses a radar point cloud-based tunnel blasting super-undermining analysis method, which comprises the steps of fusing point cloud data of continuous multi-period blasting, and learning and generating a dynamic reference profile surface representing the average level of actual construction so as to replace a fixed design surface as an analysis reference; the method comprises the steps of comparing a current circulation point cloud with a dynamic reference plane, matching typical abnormal forms through pattern recognition, outputting risk types and high risk areas for positioning, carrying out deep analysis by adaptively calling a fine algorithm according to the risk types, generating a decision report containing specific process adjustment suggestions, automatically searching a knowledge base based on risks and geometric parameters, and generating a blasting parameter optimization scheme for the next circulation. The invention realizes the spanning from static measurement to dynamic learning, geometric statistics to risk early warning and from single analysis to closed-loop optimization, and remarkably improves the accuracy, predictability and process optimization capability of the super-undermining analysis.

Inventors

LIANG LONG
YIN HONGYA
SUN YUNCHEN
MA HONGWEI
SU JIHUA
ZHANG WENXI
WANG JIAN
DOU HECHAO
SUN WEILIANG
ZHOU YUELONG
Xing Gaochong
LI TAO
REN BINGXIN

Assignees

中铁十四局集团第四工程有限公司

Dates

Publication Date: 20260512
Application Date: 20260203

Claims (10)

1. A tunnel blasting super-underexcavation analysis method based on radar point cloud is characterized by comprising the following steps: acquiring point cloud data of current and historical blasting cycles, and generating a dynamic reference profile through multi-period data fusion learning; Comparing the point cloud of the current cycle with the dynamic reference profile, matching a typical abnormal mode through mode identification, and outputting a risk type and a high risk area for positioning; According to the risk type, performing depth analysis on the high-risk area by adaptively calling a corresponding analysis algorithm, and generating an adaptive construction decision report containing hidden danger geometric parameters and specific process adjustment; and automatically generating a blasting parameter optimization scheme for the next blasting cycle according to a preset optimization knowledge base based on the risk type and the hidden danger geometric parameters, and completing a closed loop from analysis to design optimization.
2. The method for analyzing the tunnel blasting over-and-under-excavation based on the radar point cloud according to claim 1, wherein the step of obtaining the point cloud data of the current and the historical blasting cycles and generating a dynamic reference profile through multi-period data fusion learning specifically comprises the following steps: collecting three-dimensional radar point cloud data of a tunnel excavation surface of a current blasting cycle and at least two blasting cycles through a fixed scanning base station or a mobile scanning platform arranged in a tunnel, and denoising point clouds of each period; Registering the point cloud data of each period to the same tunnel design coordinate system, and respectively calculating the basic super-underexcavation value of the point cloud of each period relative to the design contour surface; based on the continuous multi-period foundation super-underexcavation value, a dynamic reference profile surface representing the actual average excavation trend is learned and generated on the basis of the designed profile surface.
3. The method for analyzing the tunnel blasting super-underexcavation based on the radar point cloud according to claim 2, wherein the registering the point cloud data of each period under the same tunnel design coordinate system and calculating the basic super-underexcavation value of the point cloud of each period relative to the design contour surface respectively specifically comprises: With tunnel design axis And design the outline surface Constituted tunnel design coordinate system Based on the first Phase point cloud Rigid body transformation matrix registration by target or stable feature points Post-registration point cloud data The method comprises the following steps: Wherein: Is the registered first Phase Point cloud Three-dimensional coordinates of the sampling points; for each registered sampling point Calculate it to design contour plane Is a directional normal distance of (2): Wherein: To design the profile surface Upper distance The point of the closest approach is, To design the profile surface At the position of Unit external normal vector at position, sign function The definition is as follows: ; for the base overexcitation value, a positive value is defined to indicate overexcitation, and a negative value indicates underexcavation.
4. The method for analyzing tunnel blasting undermining based on radar point cloud according to claim 3, wherein the continuous multi-period-based basic undermining value learns and generates a dynamic reference profile surface representing actual average mining trend based on the design profile surface, specifically comprising: Carrying out statistical analysis on the basic super-underexcavation values of continuous multiple periods on each corresponding section position of the designed profile surface along the axial direction of the tunnel; calculating the average offset of the actual excavation outline at the section position relative to the design outline by adopting an exponential weighted average method; and translating each point on the design contour surface along the normal direction of the design contour surface by corresponding average offset, and obtaining the formed continuous curved surface as the dynamic reference contour surface.
5. The method for analyzing the tunnel blasting over-and-under-excavation based on the radar point cloud according to claim 1, wherein the method is characterized in that the point cloud of the current cycle is compared with the dynamic reference profile, typical abnormal patterns are matched through pattern recognition, and the output risk type and the high risk area are positioned, and specifically comprises the following steps: taking the dynamic reference profile as a comparison reference of the current cycle, and calculating a relative deviation value of the current cycle point cloud relative to the dynamic reference profile; Meanwhile, based on the relative deviation values of all sampling points in the current cycle and the three-dimensional coordinates thereof, extracting spatial clusters with significance of the relative deviation values through a spatial clustering algorithm, and calculating geometric characteristic parameters of each cluster; and if the matching degree of the feature vector of a certain cluster and a certain mode in the mode library exceeds a set threshold, judging that the potential risk type exists in the current cycle, and outputting the space range of the cluster as high risk area positioning.
6. The method for analyzing the tunnel blasting over-and-under-excavation based on the radar point cloud according to claim 5, wherein the calculating the relative deviation value of the current circulation point cloud relative to the dynamic reference contour plane by using the dynamic reference contour plane as the comparison reference of the current circulation specifically comprises: Current cycle number The registered point cloud data is Wherein: Registering the post-point cloud for the current cycle The point coordinate vector is used to determine the point coordinate, Sampling points of the current cycle; calculating the relative dynamic reference profile of each sampling point Obtain the relative deviation value: Wherein: for the current cycle The relative deviation value of the sampling points, wherein a positive value represents the overexcitation amount relative to the dynamic reference plane, and a negative value represents the underexcavation amount; Is a dynamic reference profile Upper distance The nearest point; Is a dynamic reference profile Is a unit external normal vector field function; For the symbol discriminant function, the same definition as above is given 。
7. The method for analyzing the tunnel blasting over-and-under-excavation based on the radar point cloud according to claim 6, wherein the extracting the spatial clustering with significance of the relative deviation value through the spatial clustering algorithm comprises the following steps: Construction of comprehensive feature vectors Spatial clustering algorithm based on density is adopted for point set Clustering is carried out, and similarity measurement is defined: Wherein: A comprehensive distance measurement function used in a clustering algorithm; And (3) with The weight coefficients respectively of the difference between the space distance and the deviation value satisfy the following conditions ; Through clustering to obtain Significant clustering Wherein: Is the first A sample index set of clusters, and each cluster satisfies Wherein: A minimum cluster point threshold.
8. The method for analyzing the tunnel blasting over-run and under-run based on the radar point cloud according to claim 7, wherein the calculating the geometric characteristic parameter of each cluster specifically comprises: three-dimensional space distribution form parameters, and obtaining characteristic values through principal component analysis Wherein: Is clustered The PCA feature values of (1) are arranged in descending order of size, and morphological feature vectors are defined: Wherein: is a three-dimensional space distribution morphological feature subvector, A volume estimate surrounding the ellipsoid for the cluster; the deviation value gradient change direction parameter is used for constructing a space gradient field with relative deviation value in the cluster: Wherein: In order to deviate from the value space gradient vector, Is taken as a point In the cluster A nearest neighbor index set within; Calculating a high-order statistic of relative deviation values in the clusters by calculating a deviation value statistical distribution parameter. Wherein: the feature sub-vectors are statistically distributed for the deviation values, Respectively mean value, standard deviation and skewness coefficient of the deviation value, ; Fusing the parameters to construct clusters Is a complex geometric feature parameter vector of (1) : 。
9. The method for analyzing the tunnel blasting over-and-under-excavation based on the radar point cloud according to claim 1, wherein the method is characterized in that according to the risk type, a corresponding analysis algorithm is adaptively called to carry out depth analysis on the high-risk area, and an adaptive construction decision report containing accurate geometric parameters and specific process adjustment is generated, and specifically comprises the following steps: according to the risk type, a secondary analysis algorithm uniquely corresponding to the risk type is called from a preset algorithm mapping relation library, and point clouds defined by the high-risk area positioning are re-analyzed; Executing the called secondary analysis algorithm, and calculating to obtain hidden danger geometric parameters for quantitatively describing the specific risk forms; and (3) integrating the relative deviation value statistical result, risk type and hidden danger geometric parameters of the whole current circulation, and automatically generating a self-adaptive construction decision report according to a preset report template, wherein the self-adaptive construction decision report comprises the general evaluation of the current blasting quality, a high-risk area positioning list ordered according to the risk level, hidden danger geometric parameter description aiming at each high-risk area and specific construction treatment suggestions bound with the hidden danger geometric parameter description.
10. The method for analyzing the tunnel blasting over-and-under-excavation based on the radar point cloud according to claim 1, wherein the method for automatically generating a blasting parameter optimization scheme for the next blasting cycle according to a preset optimization knowledge base based on the risk type and the hidden danger geometrical parameters to complete a closed loop from analysis to design optimization specifically comprises the following steps: Based on risk types and hidden danger geometric parameters, calling a preset blasting parameter optimization knowledge base, wherein the knowledge base defines mapping rules between different risk types and geometric parameters and adjustable blasting design parameters; According to the mapping rule, a blasting parameter optimization scheme aiming at correcting the currently identified abnormal mode for the next blasting cycle is automatically generated, the scheme explicitly lists a blast hole group or area to be adjusted, and specific adjustment quantity suggestions for at least one parameter of the blast hole distance, the blast hole depth, the charge quantity or the detonation sequence are given.

Description

Tunnel blasting super-undermining analysis method based on radar point cloud Technical Field The invention relates to the technical field of tunnel construction, in particular to a tunnel blasting super-undermining analysis method based on radar point cloud. Background In the tunnel drilling and blasting method construction, the rapid and accurate super-underexcavation analysis of the blasted excavation profile is a core link for controlling engineering quality, cost and safety. Currently, a point cloud is acquired based on three-dimensional laser radar (LiDAR) or photogrammetry and compared with a tunnel design model, and the point cloud is becoming a mainstream super-undermining detection method. Existing solutions generally follow a static analysis paradigm of "scan-contrast-statistics". Specifically, the method comprises the steps of firstly obtaining the point cloud data of the exploded excavation surface, directly calculating the normal distance from each point in the point cloud to the designed theoretical contour surface after coordinate registration, judging overexcitation or underexcavation, and finally counting indexes such as overexcitation amount, maximum overexcitation depth and the like. To improve accuracy, some improved methods are directed to point cloud registration algorithms, point cloud filtering, or finer rasterized distance calculations. However, there are two fundamental limitations common to the prior art, first, the analysis benchmark is stiff. All comparisons are based on a fixed and constant design contour, and in actual construction, the excavated contour tends to deviate from the design contour systematically and integrally (e.g., 10 cm of total excavation is normal in soft rock sections) due to factors such as the geology of the particular surrounding rock, the cluster performance of the equipment, etc. The traditional method combines the systematic deviation with the accidental abnormality caused by the improper blasting parameters, so that the analysis result is distorted, the technological level of single blasting cannot be truly reflected, the false alarm rate is high, and the guiding significance is limited. Second, the analysis logic is passive and isolated. The prior art can only finish post geometric measurement, can not identify the specific spatial morphology mode of the super-undermining, and can not be related to the potential engineering risk and the process cause. Meanwhile, analysis reports of each cycle are mutually independent, so that a plurality of data islands are formed, the closed loop capacity of continuous learning and feedback optimization of subsequent construction by utilizing historical data is lacking, precious construction data value cannot be precipitated, and autonomous optimization based on data is difficult to realize in a construction process. Therefore, a tunnel blasting super-undermining analysis method capable of dynamically learning construction normalcy, intelligently identifying risk modes and automatically forming an optimal decision closed loop is urgently needed. Disclosure of Invention The invention aims to solve the defects in the prior art, and provides a tunnel blasting super-undermining analysis method based on radar point cloud. In order to achieve the above purpose, the present invention adopts the following technical scheme: A tunnel blasting super-underexcavation analysis method based on radar point cloud comprises the following steps: acquiring point cloud data of current and historical blasting cycles, and generating a dynamic reference profile through multi-period data fusion learning; Comparing the point cloud of the current cycle with the dynamic reference profile, matching a typical abnormal mode through mode identification, and outputting a risk type and a high risk area for positioning; According to the risk type, performing depth analysis on the high-risk area by adaptively calling a corresponding analysis algorithm, and generating an adaptive construction decision report containing hidden danger geometric parameters and specific process adjustment; and automatically generating a blasting parameter optimization scheme for the next blasting cycle according to a preset optimization knowledge base based on the risk type and the hidden danger geometric parameters, and completing a closed loop from analysis to design optimization. As a further technical scheme of the invention, the method for acquiring the point cloud data of the current and historical blasting cycles and generating a dynamic reference profile through multi-period data fusion learning specifically comprises the following steps: collecting three-dimensional radar point cloud data of a tunnel excavation surface of a current blasting cycle and at least two blasting cycles through a fixed scanning base station or a mobile scanning platform arranged in a tunnel, and denoising point clouds of each period; Registering the point cloud data of each period to the