Adaptive access and intelligent cleaning processing method for multi-source heterogeneous data

Abstract

This invention provides an adaptive access and intelligent cleaning method for multi-source heterogeneous data, belonging to the field of intelligent cleaning technology. The method includes: constructing triangular mesh surfaces connecting adjacent measuring points based on a sequence of fiber optic strain measurement points arranged along the blade spanwise and their three-dimensional spatial coordinates; calculating the spatial normal vector of each triangular mesh surface to obtain a set of mesh surface normal vectors; calculating the dihedral angle between the spatial normal vectors of adjacent triangular mesh surfaces based on the set of mesh surface normal vectors; and marking the mesh surface regions where the dihedral angle represents a local morphological abrupt change as ice ridge growth areas to obtain a subset of stagnation point classifications. This invention improves the spatial distribution and detachment early warning capabilities of ice cladding thickness.

Description

Technical Field

[0001] This invention relates to the field of intelligent cleaning and processing technology, and in particular to an adaptive access and intelligent cleaning and processing method for multi-source heterogeneous data. Background Technology

[0002] When wind turbines operate in low-temperature and high-humidity areas, icing often occurs on the blade surface. Icing not only alters the aerodynamic shape of the blades and reduces power generation efficiency, but it can also detach during operation, posing safety hazards to the turbine and surrounding facilities. Accurate detection of blade icing is a key technical requirement for ensuring the safe operation of wind turbines, identifying ice ridge growth areas, obtaining the spanwise icing thickness distribution, and providing early warning of icing detachment risks. In existing technologies, a wind farm uses accelerometers installed at the blade root to monitor icing, indirectly determining the icing status by utilizing the vibration frequency shift caused by changes in blade mass. This scheme installs the accelerometers approximately 1.5 meters from the blade root, collecting vibration signals during blade rotation, extracting the first three natural frequencies, and comparing them with the reference frequency in the un-iced state. When the frequency shift exceeds a preset threshold, the system determines the blade is iced and issues an alarm. This scheme uses a concentrated sensor installation location and a small number of sensors, requiring only a few sensors to cover the vibration monitoring of the entire blade, resulting in lower equipment costs and simpler construction and maintenance.

[0003] The drawback of this scheme is that the vibration frequency monitoring of the accelerometer can only provide a binary judgment of the overall icing status of the blade (iced / uniced), and it is difficult to obtain the spatial distribution information of icing on the blade surface along the span. The icing thickness is not uniformly distributed along the blade span. The stagnation area at the leading edge is more likely to form local accumulations of ice ridges, while the trailing edge area is affected by eddies and presents different icing morphologies. The scheme cannot identify the location of ice ridge growth, the trend of spanwise thickness change, and the risk area of ​​icing detachment. When the icing thickness in a certain local area of ​​the blade has reached the critical condition for detachment but the overall vibration frequency of the blade has not shifted significantly, the scheme is unable to provide an early warning and there is a safety blind spot for sudden icing detachment. Summary of the Invention

[0004] This invention provides an adaptive access and intelligent cleaning method for multi-source heterogeneous data, which improves the ability to predict the spatial distribution of ice thickness and the early warning of ice shedding.

[0005] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows: Firstly, a method for adaptive access and intelligent cleaning of multi-source heterogeneous data, the method comprising: Based on the sequence of fiber strain measurement points arranged along the blade span and their three-dimensional spatial coordinates, a triangular mesh surface connecting adjacent measurement points is constructed, and the spatial normal vector of each triangular mesh surface is calculated to obtain the set of mesh surface normal vectors. Based on the set of grid surface normal vectors, the dihedral angle between the spatial normal vectors of adjacent triangular grid surfaces is calculated. The grid surface regions where the dihedral angle represents a local morphological abrupt change are marked as ice ridge growth zones, thus obtaining a classification subset of stationary areas. Obtain the trailing edge point cloud sequence and its spatial coordinates. Determine the spatial projection reference of the trailing edge point cloud sequence based on the classification subset of the stagnation area. Calculate the spatial straight-line distance difference between adjacent point clouds under the spatial projection reference to obtain the spatial discrete jump field of the trailing edge. Based on the discrete jump field of the trailing edge space, the spatial straight-line distance difference is accumulated segment by segment along the vortex direction in the trailing edge vortex region to obtain the cumulative thickness field of the trailing edge ice ridge; the scattering noise points in the cumulative thickness field of the trailing edge ice ridge are removed according to the classification subset of the stagnation area to obtain the pure thickness field of the trailing edge ice ridge. Based on the classification subset of the stationary area and the thickness field of the pure ice ridge at the trailing edge, spatial coordinate alignment and data interpolation are performed along the blade span to obtain the full span ice thickness distribution matrix. Based on the spanwise icing thickness distribution matrix, the thickness difference between adjacent spanwise locations is calculated to obtain the spanwise thickness difference field; based on the spanwise thickness difference field, the physical region that characterizes the icing detachment trend by the thickness difference is extracted to obtain the icing detachment early warning zone.

[0006] Secondly, the adaptive access and intelligent cleaning system for multi-source heterogeneous data includes: The module is used to construct triangular mesh surfaces connecting adjacent measurement points based on the sequence of fiber strain measurement points arranged along the blade span and their three-dimensional spatial coordinates, calculate the spatial normal vector of each triangular mesh surface, and obtain the set of mesh surface normal vectors. The labeling module is used to calculate the dihedral angle between the spatial normal vectors of adjacent triangular mesh surfaces based on the set of mesh surface normal vectors, and to label the mesh surface regions where the dihedral angle represents a local morphological change as ice ridge growth areas, thus obtaining a classification subset of stationary areas; The calculation module is used to obtain the trailing edge point cloud sequence and its spatial coordinates, determine the spatial projection reference of the trailing edge point cloud sequence based on the classification subset of the stagnation area, calculate the difference in spatial straight-line distance between adjacent point clouds under the spatial projection reference, and obtain the spatial discrete jump field of the trailing edge. The elimination module is used to accumulate the spatial straight-line distance difference segment by segment along the vortex direction in the trailing vortex region based on the discrete jump field of the trailing space to obtain the cumulative thickness field of the trailing ice ridge; and to eliminate the scattering noise points in the cumulative thickness field of the trailing ice ridge based on the classification subset of the stagnation area to obtain the pure thickness field of the trailing ice ridge. The interpolation module is used to perform spatial coordinate alignment and data interpolation along the blade span based on the classification subset of the stationary area and the thickness field of the pure ice ridge at the trailing edge, so as to obtain the full spanwise ice thickness distribution matrix. The extraction module is used to calculate the thickness difference between adjacent spanwise positions based on the spanwise ice thickness distribution matrix, and obtain the spanwise thickness difference field; based on the spanwise thickness difference field, it extracts the physical region that represents the ice detachment trend by the thickness difference, and obtains the ice detachment early warning zone.

[0007] The above-described solution of the present invention has at least the following beneficial effects: By integrating two types of heterogeneous data—fiber optic strain measurement points and trailing edge point clouds—automatic identification of ice ridge growth zones is achieved using triangular mesh surface normal vectors and dihedral angle calculations. This solves the problem that a single sensor cannot distinguish ice cladding morphology types. The cumulative thickness field of the ice ridge is obtained by progressively accumulating the trailing edge spatial discrete jump field and eddy current direction. Scattering noise is eliminated by constraining the stagnation point area classification subset, improving the accuracy and spatial continuity of ice cladding thickness inversion. Multi-source data spatial coordinate alignment and interpolation are performed along the blade span to output a spanwise ice cladding thickness distribution matrix, filling the technical gap of existing schemes that can only provide overall binary judgment and cannot obtain spanwise thickness distribution. Based on the spanwise thickness drop field, the physical region of ice cladding detachment trend is extracted to generate an ice cladding detachment early warning zone, providing spatial positioning basis for wind turbine active de-icing decisions and operational safety. Attached Figure Description

[0008] Figure 1 This is a flowchart illustrating the adaptive access and intelligent cleaning process for multi-source heterogeneous data provided in an embodiment of the present invention.

[0009] Figure 2 This is a schematic diagram of the process for obtaining the discrete jump field of the trailing edge space provided by an embodiment of the present invention.

[0010] Figure 3 This is a schematic diagram of an adaptive access and intelligent cleaning system for multi-source heterogeneous data provided in an embodiment of the present invention. Detailed Implementation

[0011] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0012] like Figure 1 As shown, embodiments of the present invention propose an adaptive access and intelligent cleaning method for multi-source heterogeneous data, the method comprising the following steps: Step 1: Based on the sequence of fiber strain measurement points arranged along the blade span and their three-dimensional spatial coordinates, construct a triangular mesh surface connecting adjacent measurement points, calculate the spatial normal vector of each triangular mesh surface, and obtain the set of mesh surface normal vectors. Step 2: Based on the set of normal vectors of the grid surface, calculate the dihedral angle between the spatial normal vectors of adjacent triangular grid surfaces, and mark the grid surface regions where the dihedral angle represents local morphological abrupt changes as ice ridge growth areas to obtain the classification subset of stationary areas; Step 3: Obtain the trailing edge point cloud sequence and its spatial coordinates. Determine the spatial projection reference of the trailing edge point cloud sequence based on the classification subset of the stagnation area. Calculate the spatial straight-line distance difference between adjacent point clouds under the spatial projection reference to obtain the spatial discrete jump field of the trailing edge. Step 4: Based on the discrete jump field of the trailing edge space, accumulate the spatial straight-line distance difference segment by segment along the vortex direction in the trailing edge vortex region to obtain the cumulative thickness field of the trailing edge ice ridge; remove the scattering noise points in the cumulative thickness field of the trailing edge ice ridge according to the classification subset of the stagnation area to obtain the pure thickness field of the trailing edge ice ridge. Step 5: Based on the classification subset of the stationary area and the thickness field of the pure ice ridge at the trailing edge, perform spatial coordinate alignment and data interpolation along the blade span to obtain the full spanwise ice thickness distribution matrix; Step 6: Calculate the thickness difference between adjacent spanwise positions based on the spanwise icing thickness distribution matrix to obtain the spanwise thickness difference field; extract the physical region that represents the icing detachment trend based on the spanwise thickness difference field to obtain the icing detachment early warning zone.

[0013] In this embodiment of the invention, two types of heterogeneous data, namely fiber optic strain measurement points and trailing edge point clouds, are integrated. The automatic identification of ice ridge growth zones is achieved by using triangular mesh surface normal vectors and dihedral angle calculations, which solves the problem that a single sensor cannot distinguish the type of icing morphology. The cumulative thickness field of the ice ridge is obtained by accumulating the trailing edge spatial discrete jump field and eddy current direction segment by segment. Scattering noise points are eliminated by constraining the stagnation point area classification subset, which improves the accuracy and spatial continuity of icing thickness inversion. Multi-source data spatial coordinate alignment and interpolation are performed along the blade span to output a spanwise icing thickness distribution matrix, filling the technical gap of existing solutions that can only provide overall binary judgment and cannot obtain spanwise thickness distribution. Based on the spanwise thickness drop field, the physical region of icing detachment trend is extracted to generate an icing detachment early warning zone, providing spatial positioning basis for wind turbine active de-icing decision-making and operational safety.

[0014] In a preferred embodiment of the present invention, step 1 above may include: Step 11: Based on the sequence of fiber optic strain measurement points arranged along the blade spanwise and their three-dimensional spatial coordinates, extract the three-dimensional spatial coordinates of three adjacent fiber optic strain measurement points. Connect the adjacent measurement points with spatial straight lines to construct a physical triangular mesh surface, thus obtaining a set of triangular mesh surfaces on the blade surface. Specifically, this includes: based on the sequence of fiber optic strain measurement points arranged along the blade spanwise and their three-dimensional spatial coordinates, extract the three-dimensional spatial coordinates of three adjacent fiber optic strain measurement points. Connect the adjacent measurement points with spatial straight lines to construct a physical triangular mesh surface, thus obtaining a set of triangular mesh surfaces on the blade surface. The sequence of fiber optic strain measurement points arranged along the blade spanwise is such that the fiber optic strain measurement points are arranged sequentially on the blade surface along the physical extension direction of the blade spanwise. Each measurement point has a unique three-dimensional spatial coordinate, denoted as the first point. The spatial coordinates of each measuring point are as follows: the measuring points are arranged at equal or varying intervals along the spanwise direction from the blade root to the blade tip. The spanwise spacing between adjacent measuring points is determined by the spatial resolution of the fiber optic sensor and the blade span. The construction process proceeds group by group according to the physical extension order along the blade spanwise. Three adjacent measuring points along the spanwise direction are taken as a group, and their three-dimensional spatial coordinates are set as the coordinates of the first measuring point, the second measuring point, and the third measuring point, respectively. The first measuring point and the second measuring point are connected by a straight line in space, forming the first physical edge of the triangular mesh surface. The second measuring point and the third measuring point are connected by a straight line in space, forming the second physical edge of the triangular mesh surface. The third measuring point and the first measuring point are connected by a straight line in space, forming the third physical edge of the triangular mesh surface. The three physical edges are connected end to end in three-dimensional space, forming a closed physical triangular mesh surface. This physical triangular mesh surface is directly attached to the physical surface of the blade, and its three vertices are the physical positions of the three fiber optic strain measuring points on the blade surface. Adjacent triangular mesh surfaces are spliced ​​together by sharing physical edges. Specifically, the first... The triangular mesh surface constructed by the three measuring points in the group and the first The triangular mesh surface constructed by the three measuring points in the group is shared by the first... measuring points and the first The physical boundary between measurement points.

[0015] The process iterates through all fiber optic strain gauges, advancing along the spanwise sliding window in groups of three, with a window step size of one gauge point forward for each additional group. This continues until the last three gauges participate in constructing the final triangular mesh surface, at which point the sliding window terminates. All constructed physical triangular mesh surfaces are numbered according to their spanwise physical extension order, resulting in a set of triangular mesh surfaces on the blade surface. The data structure of this set is as follows: each entry contains the triangular mesh surface number, the numbers of the three vertex gauges, and the three-dimensional spatial coordinates of the three vertices. The entries are arranged in ascending order of their triangular mesh surface numbers, consistent with the spanwise physical extension direction. This set uses the physical triangular mesh surfaces as the basic units, discretizing the blade surface along the spanwise direction into a structured surface mesh.

[0016] Step 12: Based on the set of triangular mesh faces on the blade surface, extract the spatial direction vectors of the two intersecting physical edges within each physical triangular mesh face. Calculate the cross product of the two intersecting physical edge spatial direction vectors to obtain the perpendicular normal direction, thus obtaining the spatial normal vector for each physical triangular mesh face. Specifically, this includes: based on the set of triangular mesh faces on the blade surface, extract the spatial direction vectors of the two intersecting physical edges within each physical triangular mesh face. Calculate the cross product of the two intersecting physical edge spatial direction vectors to obtain the perpendicular normal direction, thus obtaining the spatial normal vector for each physical triangular mesh face. Receive the set of triangular mesh faces on the blade surface, and for the first... A physical triangular mesh surface has three physical edges. Two physical edges intersecting at the same vertex are selected from these three edges. Let these two selected physical edges be the first edge pointing from the first vertex to the second vertex, and the second edge pointing from the first vertex to the third vertex. The spatial direction vector of the first edge is the first direction vector, which is calculated by subtracting the coordinates of the first vertex from the coordinates of the second vertex. The spatial direction vector of the second edge is the second direction vector, which is calculated by subtracting the coordinates of the first vertex from the coordinates of the third vertex.

[0017] Both the first and second direction vectors lie within the spatial plane containing the triangular mesh face, and they are not collinear. The cross product of the first and second direction vectors yields the spatial normal vector of the triangular mesh face. This spatial normal vector is perpendicular to any two non-collinear direction vectors within the triangular mesh face, and its direction points to one side of the plane containing the triangular mesh face. The formula for calculating the cross product is: ; In the formula, Indicates the first The spatial normal vector of each physical triangular mesh surface is a three-dimensional vector, whose three scalar components correspond to the three coordinate axes of the spatial rectangular coordinate system, respectively. Indicates the first The first spatial direction vector of a physical triangular mesh surface points from the first vertex to the second vertex, and its three components are: , respectively, represent the projection lengths of the direction vector on the first, second, and third coordinate axes; Indicates the first The second spatial direction vector of each physical triangular mesh surface points from the first vertex to the third vertex, and its three components are: ; The scalar component of the spatial normal vector in the direction of the first coordinate axis is obtained by subtracting the product of the second component of the first direction vector and the third component of the second direction vector from the product of the second component of the first direction vector and the third component of the second direction vector. This represents the scalar component of the spatial normal vector along the second coordinate axis. The scalar component of the spatial normal vector in the direction of the third coordinate axis is represented. The magnitude of the spatial normal vector is equal to the area of ​​the parallelogram formed by the first direction vector and the second direction vector as adjacent sides, which is twice the area of ​​the triangular mesh surface. The direction of the spatial normal vector is determined by the right-hand rule: the four fingers of the right hand turn from the direction of the first direction vector to the direction of the second direction vector, and the thumb points to the positive direction of the normal vector. For the outer surface of the blade, it is uniformly stipulated that the normal vector points to the outer side of the blade.

[0018] Iterate through all triangular mesh faces in the triangular mesh facet set on the blade surface, extract the direction vectors of the two intersecting edges of each triangular mesh facet, perform a cross product operation, and obtain the spatial normal vector. Store the index of each triangular mesh facet as associated with its spatial normal vector to obtain the mapping relationship between the spatial normal vector and the facet index.

[0019] Step 13: Based on the spatial normal vector of each physical triangular mesh surface, summarize the spatial normal vectors of all physical triangular mesh surfaces in the physical extension order along the blade span, and perform spatial coordinate index mapping to obtain the set of mesh surface normal vectors. Specifically, this includes: based on the spatial normal vector of each physical triangular mesh surface, summarizing the spatial normal vectors of all physical triangular mesh surfaces in the physical extension order along the blade span, and performing spatial coordinate index mapping to obtain the set of mesh surface normal vectors. The spatial normal vector of each physical triangular mesh surface is received. The physical triangular mesh surfaces on the blade surface have a natural order relationship along the physical extension direction along the span. The triangular mesh face and the first The three triangular mesh faces are adjacent to each other in the span direction and share a physical edge that spans the span direction. The spatial normal vectors of all the triangular mesh faces are arranged in the order of physical extension of the span direction to form a span normal vector sequence. The specific operation of the spatial coordinate index mapping is as follows: the span direction number of each triangular mesh face is used as the primary index key, and the three-dimensional spatial coordinates of the centroid of the triangular mesh face are used as the secondary index key. The three scalar components of the corresponding spatial normal vector are associated and stored in the index structure.

[0020] The three-dimensional spatial coordinates of the centroid of the triangular mesh face are obtained by taking the arithmetic mean of the coordinates of the three vertices of the face. The index structure supports fast retrieval by spanwise position range or spatial coordinate range. It can directly extract the normal vectors of all triangular mesh faces within a specified spanwise segment by spanwise index range, or retrieve the normal vectors of triangular mesh faces within a specific physical region by spatial bounding box. The spanwise normal vector sequence and its spatial coordinate index are merged and organized to form a set of mesh face normal vectors. The data structure of the set of mesh face normal vectors is as follows: each entry contains the spanwise index of the triangular mesh face, the three-dimensional spatial coordinates of the centroid, and the three scalar component values ​​of the spatial normal vector. All entries are arranged in ascending order by spanwise index. Based on the normal vectors of adjacent triangular mesh faces in this set, the dihedral angle of each face is calculated, thereby identifying the ice ridge growth zone at the leading edge.

[0021] In this embodiment of the invention, a physical triangular mesh surface is constructed by extracting three adjacent fiber strain measurement points along the blade spanwise, and the spatial normal vector of each surface element is obtained by cross product operation and spanwise index mapping is performed to transform the discrete fiber measurement point data into a mesh surface normal vector field with a continuous topological structure.

[0022] In a preferred embodiment of the present invention, step 2 above may include: Step 21: Based on the set of normal vectors for the mesh surfaces, extract the spatial normal vectors of two adjacent physical triangular mesh surfaces, calculate the spatial angle between the two spatial normal vectors, and obtain the set of dihedral angles of adjacent mesh surfaces. Specifically, this includes: extracting the spatial normal vectors of two adjacent physical triangular mesh surfaces based on the set of normal vectors for the mesh surfaces, calculating the spatial angle between the two spatial normal vectors, obtaining the set of dihedral angles of adjacent mesh surfaces, and receiving the set of normal vectors for the mesh surfaces. The condition for determining whether two physical triangular mesh surfaces are adjacent is that the two triangular mesh surfaces share the same physical edge. In the set of normal vectors for the mesh surfaces, each entry contains the vertex index information of the corresponding triangular mesh surface. The adjacent relationship can be determined by checking whether the sets of the three vertex indexes of the two triangular mesh surfaces have an intersection. If the intersection contains two vertex indexes, they are adjacent; if the intersection contains three vertex indexes, they are the same triangular mesh surface; if the intersection contains zero or one vertex index, they are not adjacent. For the first vertex index that satisfies the adjacent condition... The triangular mesh face and the first For each triangular mesh facet, extract the spatial normal vectors of the two adjacent triangular mesh faces from the set of mesh facet normal vectors, and calculate the spatial angle between the two spatial normal vectors. This angle is the dihedral angle between the adjacent mesh faces. The dihedral angle characterizes the degree of spatial bending of the two adjacent triangular mesh faces at the shared edge. Its value reflects the smoothness of the geometric transition of the blade surface at that adjacent position. The formula for calculating the dihedral angle is: ; In the formula, Indicates the first The triangular mesh face and the first The dihedral angle between three triangular mesh faces, in radians, has a range of values ​​of [value missing]. ; Indicates the first The spatial normal vector of each triangular mesh face is a three-dimensional vector with three scalar components; Indicates the first Spatial normal vectors of triangular mesh faces; The inner product of two normal vectors is equal to the sum of the products of their corresponding components. ; normal vector The length of the Euclidean module is equal to ; Normal vector The Euclidean modulus is calculated in the same way as above. The physical meaning of the dihedral angle is: when... When approaching zero radian, the two triangular mesh surfaces tend to be coplanar, and the blade surface transitions smoothly at these adjacent positions without geometric abrupt changes; when When the angle is large, the spatial orientation difference between the two faces is significant, and there are obvious bends or edges on the blade surface at these adjacent positions. The normal aerodynamic design curvature of the blade leading edge region produces a certain dihedral angle baseline value; the sharp geometric transition at the boundary of the ice ridge growth zone, caused by local ice accumulation, results in a dihedral angle value significantly higher than the baseline value. The dihedral angle calculation process involves three steps: calculating the inner product of the two normal vectors; dividing the product of the magnitudes of the two normal vectors by the inner product to obtain the cosine value of the dihedral angle; and finally taking the inverse cosine function of the cosine value to obtain the dihedral angle radian value. This process is repeated for each pair of adjacent triangular mesh faces in the set of normal vectors. The dihedral angle values ​​of each pair of adjacent triangular mesh faces are associated with the corresponding face element index pairs and stored. These pairs are then organized and sorted by face element index to obtain the dihedral angle set of adjacent mesh faces. This set of dihedral angles of adjacent mesh faces encodes the geometric bending distribution of the blade surface into a discrete numerical field in the form of dihedral angles of each adjacent face.

[0023] Step 22: Based on the dihedral angle set of adjacent grid surfaces, extract the physical triangular grid surfaces whose dihedral angle values ​​satisfy the critical dihedral angle condition characterizing ice ridge growth, thus obtaining the set of morphological abrupt change grid surfaces, including: Step 22.1: Based on the set of dihedral angles of adjacent mesh faces, extract the dihedral angle values ​​between each physical triangular mesh face and all its adjacent physical triangular mesh faces, calculate the absolute value of the difference between adjacent dihedral angle values, and obtain the set of local dihedral angle change rates. Specifically, this includes: based on the set of dihedral angles of adjacent mesh faces, extract the dihedral angle values ​​between each physical triangular mesh face and all its adjacent physical triangular mesh faces, calculate the absolute value of the difference between adjacent dihedral angle values, and obtain the set of local dihedral angle change rates. The set of dihedral angles of adjacent mesh faces is used for the... For each triangular mesh face, first determine the set of indices of its adjacent triangular mesh faces. The set of adjacent triangular mesh face indices contains all faces that are adjacent to the first triangular mesh face. The index set is determined by retrieving the index of triangular mesh faces that share at least one physical edge from the dihedral angle set of adjacent mesh faces, and all triangular mesh faces that share at least one physical edge. The face elements of the triangular mesh are obtained by the first pair of face elements. The local rate of change of the dihedral angle of the triangular mesh face is defined as follows: [The text abruptly ends here, so the translation stops as well.] The definition describes the maximum absolute value of the difference between any two dihedral angle values ​​calculated using a triangular mesh face as the center and facing all its adjacent triangular mesh faces. The maximum discrete jump amplitude of the dihedral angle values ​​around each triangular mesh surface, and the formula for calculating the local rate of change of the dihedral angle are: ; In the formula, Indicates the first The local rate of change of the dihedral angle of a triangular mesh facet, in radians. The larger the value, the higher the degree of dispersion of the dihedral angle around the mesh facet. Indicates the first The set of indices of adjacent triangular mesh faces of the i-th triangular mesh face, containing all those adjacent to the i-th triangular mesh face. The triangular mesh face shares the index of at least one physical edge; Indicates the first The triangular mesh face and the first Dihedral angles between adjacent triangular mesh faces, in radians; Indicates the first The triangular mesh face and the first Dihedral angles between adjacent triangular mesh faces, in radians; This represents the absolute value of the difference between two dihedral angle values, expressed in radians. This value represents the first... The physical meaning of the dihedral angle local change rate is as follows: the dihedral angle between adjacent surfaces in the smooth area of ​​the blade surface tends to be uniform, the orientation of the surface elements gradually changes in space, the dihedral angle values ​​in different adjacent directions are close, and the local change rate is low.

[0024] The situation is reversed at the boundary of the ice ridge growth zone. The dihedral angle between facets on one side of the ridge and facets on the other side is much larger than the dihedral angle between adjacent facets on the same side of the ridge. This spatial directional difference leads to an increased local rate of change. The local rate of change of the dihedral angle uses the second-order change characteristic (the facet-by-face change rate of the dihedral angle) instead of the absolute value of the first-order dihedral angle as the criterion, making it more sensitive to small local morphological abrupt changes. When the number of adjacent faces of a triangular mesh face is less than 2 (i.e., the mesh face is located at the spatial boundary of the triangular mesh face set on the blade surface, with only one or zero adjacent faces), there is insufficient information about adjacent faces to assess the local dispersion of the dihedral angle. In this case, the local rate of change of the dihedral angle is directly taken as zero radians and, according to the conservative principle, is not included in the subsequent abrupt change labeling. Each triangular mesh face involved in the dihedral angle set of adjacent mesh faces is traversed, and its local rate of change of dihedral angle is calculated face by face. The index of each triangular mesh face is associated with its local rate of change and stored to obtain the set of local rates of change of dihedral angle.

[0025] Step 22.2: Based on the set of local dihedral change rates, extract physical triangular mesh surfaces whose local dihedral change rate values ​​satisfy the preset critical condition for morphological abrupt change, and obtain a set of candidate abrupt change mesh surfaces. Specifically, this includes: based on the set of local dihedral change rates, extract physical triangular mesh surfaces whose local dihedral change rate values ​​satisfy the preset critical condition for morphological abrupt change, and obtain a set of candidate abrupt change mesh surfaces; receive the set of local dihedral change rates; set a critical threshold for morphological abrupt change, denoted as the first threshold. The first threshold is used to distinguish between normal aerodynamic curvature changes on the blade surface and geometric morphological abrupt changes caused by ice ridge growth. The specific meaning of the first threshold is: when the local dihedral change rate of a certain triangular mesh surface exceeds the first threshold, the dihedral angle difference in different adjacent directions around the mesh surface has exceeded the normal fluctuation range that the blade's own designed curvature can produce, and it is determined that there is a local geometric discontinuity caused by ice ridge growth at this location. The value of the first threshold is 0.25 radians, and the range of the first threshold is from 0.15 radians to 0.40 radians. The values ​​are based on the following: Geometric analysis of megawatt-class wind turbine blades shows that, under normal, ic-free conditions, the design aerodynamic curvature of the blade leading edge changes gently along the spanwise and chordwise directions, and the local difference in dihedral angles between adjacent grid surfaces is generally less than 0.15 radians. When ice ridges accumulate at the blade leading edge, a geometric transition with sharp edges is formed at the ice ridge boundary. There is a significant difference between the dihedral angles of adjacent surfaces on both sides of this transition edge and the dihedral angles of adjacent surfaces on the same side of the transition edge, and the local rate of change of the dihedral angle generally rises to more than 0.25 radians.

[0026] Using 0.25 radians as a typical first threshold value, an engineering-acceptable balance is achieved between the detection rate of abrupt changes in ice ridge morphology and the false alarm rate of normal blade curvature. Under icing conditions dominated by rime (relatively regular and smooth ice shapes), a first threshold of 0.20 radians can be selected to improve detection sensitivity. Under icing conditions dominated by glaze (irregular ice shapes and high surface roughness), a first threshold of 0.30 radians can be selected to reduce false alarms caused by local noise from rough surfaces. For each triangular mesh face in the set of local dihedral angle change rates, its dihedral angle change rate value is extracted. This value is then compared with the first threshold. When the... When the local rate of change of the dihedral angle of a triangular mesh surface is greater than a first threshold, the mesh surface is marked as a candidate abrupt change mesh surface, and its index is added to the candidate abrupt change mesh surface set. When the local rate of change of the dihedral angle is not greater than the first threshold, after traversing all triangular mesh surfaces and completing the above comparison-marking operation, a candidate abrupt change mesh surface set is obtained. The candidate abrupt change mesh surface set contains all triangular mesh surfaces that satisfy the critical condition of morphological abrupt change in the sense of the local rate of change of the dihedral angle. Due to the possibility of local noise interference in fiber optic strain measurement, isolated mislabeled points may be mixed in with the candidate abrupt change mesh surfaces.

[0027] Step 22.3: Based on the candidate mutation mesh surface set, extract the three-dimensional spatial connectivity distribution features of the candidate mutation mesh surfaces in the physical region of the blade leading edge and merge the spatial connectivity domains to obtain the morphological mutation mesh surface set. Specifically, this includes: based on the candidate mutation mesh surface set, extract the three-dimensional spatial connectivity distribution features of the candidate mutation mesh surfaces in the physical region of the blade leading edge and merge the spatial connectivity domains to obtain the morphological mutation mesh surface set. The candidate mutation mesh surface set is received. The ice ridge growth zone extends continuously along the spanwise direction at the blade leading edge, forming a spatially connected physical region. The actual ice ridge morphological mutation mesh surfaces are distributed in a continuous sheet-like pattern in space. Several adjacent candidate mutation mesh surfaces are connected into one piece, which is measured locally by fiber optic strain measurement. Noise-induced mislabeled points are spatially isolated, appearing as a single or a few candidate mutation mesh surfaces, not forming a spatially connected patch with the surrounding mesh surfaces. The spatial connectivity merging operation utilizes this difference in distribution characteristics to separate the real ice ridge morphological mutation regions from the noise-induced mislabeled points. The specific process of the connectivity merging operation is as follows: Constructing a spatial adjacency graph: Using the index of all mesh surfaces in the candidate mutation mesh surface set as the node set of the graph, for any two nodes in the graph, if the corresponding two candidate mutation mesh surfaces share a physical edge in the set of triangular mesh surfaces on the blade surface, then an undirected edge is established between them. The establishment of the undirected edge is based on the inherent mesh surface adjacency relationship in the set of triangular mesh surfaces on the blade surface.

[0028] Perform connected component analysis, traversing the spatial adjacency graph using either breadth-first search or depth-first search strategies. Starting from any unvisited node, search along the edges of the graph for all reachable nodes, grouping the starting node and all its reachable nodes into the same connected component, and assigning a unique number to this component. Repeat this process until all nodes in the graph are grouped into a connected component. Physical scale filtering: For each connected component, count the number of triangular mesh faces it contains. Set a minimum area threshold for the connected component, denoted as the second threshold. The second threshold is used to determine whether a connected component possesses the minimum physical scale for ice ridge growth. Specifically, if the number of triangular mesh faces in a connected component is less than this threshold, the actual area covered by the connected component on the blade surface is insufficient to accommodate the spanwise extension length of a typical ice ridge, and it is classified as noise or surface defects and discarded. The value of the second threshold is 10 triangular mesh faces. The range of the second threshold is from 5 to 30 triangular mesh faces. Its value is based on the spanwise coverage length of each triangular mesh face, determined by the spacing of the fiber optic strain gauges along the blade spanwise. Based on the typical megawatt-class blade span and fiber optic strain gauge density, the spanwise physical coverage length corresponding to 10 triangular mesh faces is approximately 1 to 3 meters, comparable to the minimum continuous extension scale of a typical ice ridge along its span. Small connected regions containing fewer than 5 triangular mesh faces have insufficient physical coverage to form identifiable ice ridge growth zones. This is due to factors such as local measurement noise at the fiber optic strain gauges, isolated manufacturing defects or localized contamination on the blade surface. The number of triangular mesh faces in each connected region is compared to a second threshold. When the number of triangular mesh faces in a connected region is less than the second threshold, the entire region is removed from the candidate aberration mesh face set, and all mesh faces within it are restored to their normal state. When the number of triangular mesh faces in a connected region is not less than the second threshold, the region is retained, and all mesh faces within it continue to be marked as a morphological aberration mesh face. All mesh faces in all retained connected regions are combined to form a morphological aberration mesh face set. The mesh faces in this set spatially form several continuous regions, each corresponding to an ice ridge growth location on the blade's leading edge.

[0029] Step 23: Based on the set of morphologically abruptly changed mesh surfaces, the physical region at the leading edge of the blade where the morphologically abruptly changed mesh surfaces are located is marked as the ice ridge growth region. The spatial boundary coordinates and normal vector distribution features of the ice ridge growth region are extracted to obtain a subset of stationary point areas containing spatial boundary information of the ice ridge growth region. Specifically, this includes: based on the set of morphologically abruptly changed mesh surfaces, marking the physical region at the leading edge of the blade where the morphologically abruptly changed mesh surfaces are located as the ice ridge growth region, and extracting the spatial boundary coordinates and normal vector distribution features of the ice ridge growth region to obtain a subset of stationary point areas containing spatial boundary information of the ice ridge growth region; receiving the set of morphologically abruptly changed mesh surfaces, where each connected domain occupies a spatially continuous region within the physical region at the leading edge of the blade, which is the ice ridge growth region. This step extracts the spatial boundary coordinates and normal vector distribution features for each ice ridge growth region to form standardized stationary point area information. Spatial boundary coordinate extraction: for each connected domain, the three-dimensional spatial coordinates of all triangular mesh surface nodes within that domain are extracted. The minimum and maximum values ​​of all node coordinates in the spanwise physical extension direction are taken and recorded as the minimum spanwise coordinate value and the maximum spanwise coordinate value, respectively. Take the minimum and maximum values ​​of all nodal coordinates in the chordal direction of the blade, and denote them as the minimum and maximum chordal coordinate values, respectively. Take the minimum and maximum values ​​of all nodal coordinates in the thickness direction of the blade, and denote them as the minimum and maximum thickness coordinate values, respectively.

[0030] Six scalar values ​​collectively define the three-dimensional bounding box of the ice ridge growth region on the blade surface, i.e., its spatial boundary coordinates. The spatial boundary coordinates physically define the spatial extent of the ice ridge growth region along the spanwise, chordwise, and thickness directions. The spanwise boundary coordinates (spanwise minimum and maximum values) are used to constrain the continuity of thickness variation in the cumulative thickness field of the trailing edge ice ridge and to align the spatial coordinates. The chordwise boundary coordinates are used to determine the position of the ice ridge growth region in the leading-to-trailing-edge direction, assisting in the construction parameters of the projection reference plane.

[0031] Normal vector distribution feature extraction: For each connected domain, the spatial normal vectors of all triangular mesh surfaces within it are extracted. The arithmetic mean of all normal vectors is calculated for each of the three components in the spatial coordinate system, yielding the average normal vector. The average normal vector reflects the dominant spatial orientation of the surface in the ice ridge growth zone. Local accumulation of ice ridges alters the curvature of the blade surface at that location, causing the dominant orientation of the surface normal vectors in the ice ridge region to deviate from the design normal vector directions of a smooth blade surface at those spanwise and chordwise positions. This deviation direction and magnitude provide key input for constructing the trailing edge point cloud projection reference. Simultaneously, the standard deviation of all normal vectors is calculated for each of the three components, yielding the standard deviation of the normal vector components. The standard deviation of the normal vector components reflects the spatial dispersion of the surface orientation in the ice ridge growth zone. Due to their irregular shape and the presence of sharp ridges, the surface of ice ridges exhibits a higher degree of dispersion of their normal vectors around the dominant direction compared to the smooth surface of normal blades. To summarize the stationary region information: the spatial boundary coordinates (six scalar values), the average value of the normal vector components (three scalar values), and the standard deviation of the normal vector components (three scalar values) of each connected domain are associated and stored, each accompanied by a unique number representing the ice ridge growth region, thus forming a stationary region entry for that connected domain. By traversing all connected domains and summarizing all stationary region entries, a stationary region classification subset is obtained.

[0032] In this embodiment of the invention, by calculating the dihedral angle between the normal vectors of adjacent grid surfaces and extracting the morphologically abrupt grid surfaces that meet the critical conditions, the boundary and normal vector distribution characteristics of the ice ridge growth zone are determined by merging spatial connected domains. This achieves automatic marking of the ice ridge growth zone based on geometric discontinuity, avoiding the error of manual interpretation that relies on empirical thresholds.

[0033] like Figure 2 As shown, in a preferred embodiment of the present invention, step 3 above may include: Step 31: Based on the classification subset of the stationary area, extract the spatial normal vector distribution characteristics of the ice ridge growth area, construct a spatial projection reference surface for the trailing edge point cloud sequence orthogonal to the average normal vector of the ice ridge growth area, and obtain the spatial projection reference. Specifically, this includes: based on the classification subset of the stationary area, extract the spatial normal vector distribution characteristics of the ice ridge growth area, construct a spatial projection reference surface for the trailing edge point cloud sequence orthogonal to the average normal vector of the ice ridge growth area, and obtain the spatial projection reference. Receive the classification subset of the stationary area, which contains the average normal vector of each ice ridge growth area. Extract the average normal vector of all ice ridge growth areas from the classification subset of the stationary area. Take the arithmetic mean of all extracted average normal vectors again. Sum the average normal vector of each ice ridge growth area with respect to each coordinate component and divide by the total number of ice ridge growth areas to obtain the overall average normal vector of the ice ridge growth area. The overall average normal vector represents the comprehensive dominant direction of the surface normal vectors of all ice ridge growth areas at the leading edge of the blade.

[0034] A spatial projection reference plane is constructed using the overall average normal vector of the ice ridge growth region as the reference direction. The mathematical definition of the spatial projection reference plane is: a spatial plane passing through an arbitrarily selected spatial reference point, with the normalized direction vector of the overall average normal vector of the ice ridge growth region as its normal vector. The spatial reference point can be any known spatial coordinate point in the trailing edge region of the blade. The physical consideration is that the growth morphology of the ice ridge is dominated by the impact and freezing process of supercooled water droplets in the leading edge stagnation zone, and the spatial geometric orientation of the ice ridge determines the offset direction of the leading edge surface normal vector. The most significant direction of icing undulations in the trailing edge point cloud is physically related to the direction of the normal vector of the leading edge ice ridge. After being diverted at the leading edge, supercooled water droplets flow along the blade surface to the trailing edge, and the distribution pattern of icing at the trailing edge is influenced by the aerodynamic flow effect around the leading edge ice ridge. Projecting the trailing edge point cloud along a direction orthogonal to the average normal vector of the leading edge ridge allows the undulation characteristics of the projected trailing edge point cloud along the direction of the ridge's influence to be preserved to the maximum extent. The resulting spatial projection reference plane is a plane with a definite normal vector direction and passing through a definite spatial reference point, denoted as the spatial projection reference.

[0035] Step 32: Based on the spatial projection datum, project the acquired blade trailing edge point cloud sequence and its three-dimensional spatial coordinates onto the spatial projection datum plane along the normal direction of the spatial projection datum plane to obtain the trailing edge projected point cloud sequence. Specifically, this includes receiving the spatial projection datum and the externally acquired blade trailing edge point cloud sequence. Each point in the blade trailing edge point cloud sequence has known three-dimensional spatial coordinates. The projection operation is performed along the normal direction of the spatial projection datum plane. The normal direction is obtained by normalizing the overall average normal vector of the determined ice ridge growth region. For the first point in the trailing edge point cloud sequence... Given a point, calculate the directed vertical distance from that point to the spatial projection reference plane. The directed vertical distance is equal to the scalar projection of the spatial displacement vector of that point relative to the reference point on the projection reference plane along the normal direction of the reference plane. Translate the original point's spatial coordinates along the normal direction of the reference plane by this directed vertical distance. After the translation, the point will fall exactly on the projection reference plane, thus obtaining the projected coordinates of that point. The formula for calculating the spatial coordinates of the projected point is: ; In the formula, Indicates the first The three-dimensional spatial coordinates of a point after the projection of a trailing edge point cloud point, which lies on the spatial projection reference plane; Indicates the first The original three-dimensional spatial coordinates of the trailing edge point cloud points before projection; Represents the three-dimensional spatial coordinates of an arbitrary reference point on the spatial projection reference plane; The unit normal vector representing the spatial projection reference plane is obtained by normalizing the overall average normal vector of the ice ridge growth region, and its magnitude is 1. For the first The spatial displacement vector of a point relative to the reference point of the projection datum plane; It represents the scalar projection of the spatial displacement vector onto the normal direction of the reference plane, that is, the directed vertical distance from the original point to the projection reference plane. This represents the translation vector from the original point along the normal direction towards the projection reference plane.

[0036] The geometric effect of this projection operation is as follows: the trailing edge point cloud is flattened from three-dimensional space along a fixed direction onto the projection reference plane. The relative positional relationship of each point in two directions within the reference plane is preserved, while the coordinate degrees of freedom along the normal direction of the reference plane are eliminated. All points in the projected trailing edge point cloud sequence are coplanar and fall within the spatial projection reference plane. The projection operation is performed point by point, traversing all points in the blade's trailing edge point cloud sequence. All projected points form the trailing edge projected point cloud sequence.

[0037] Step 33: Based on the trailing edge projection point cloud sequence, calculate the spatial straight-line distance difference between adjacent trailing edge projection point clouds to characterize the local undulation and jump features of the trailing edge surface, and obtain the trailing edge spatial discrete jump field. Specifically, this includes: receiving the trailing edge projection point cloud sequence, in which each projection point is arranged according to its spatial order within the projection reference plane, for the... The projection point and the first Given a number of projection points, calculate the spatial straight-line distance between them. The formula for calculating the spatial straight-line distance between adjacent projection points is: ; In the formula, Indicates the first The projection point and the first The spatial straight-line distance between projection points is expressed in units of length. Indicates the first The three-dimensional spatial coordinates of each projection point have the following components: ; Indicates the first The three-dimensional spatial coordinates of each projection point have the following components: ; Let Euclidean norm be the difference vector between the two projection points, i.e. The physical meaning of the spatial straight-line distance between adjacent projection points is: the three-dimensional spatial distance between adjacent sampling points taken point by point along the trailing edge surface. This distance comprehensively reflects the local undulation characteristics of the trailing edge iced surface within the projection reference plane. When the ice thickness on the trailing edge surface is uniform, the change trend of the distance between adjacent projection points is gradual and the value is stable. When the trailing edge surface experiences local undulations due to ice ridge accumulation, abrupt changes in ice thickness, or local variations in surface roughness, the distance between adjacent projection points jumps accordingly. The jump amplitude is positively correlated with the amount of local ice thickness change. The spatial straight-line distance is calculated for each pair of adjacent points in the trailing edge projection point cloud sequence. All distance values ​​are arranged according to the spatial order of the projection points within the reference plane, forming a discrete jump field in the trailing edge space.

[0038] In this embodiment of the invention, the projection reference of the trailing edge point cloud is determined by the average normal vector of the ice ridge growth area. After the trailing edge point cloud is projected onto the reference plane, the spatial straight-line distance difference between adjacent point clouds is calculated. This makes the extraction direction of the trailing edge ice undulation features adaptively associated with the morphological type of the leading edge ice ridge, thereby improving the physical consistency of the leading edge-trailing edge data fusion.

[0039] In a preferred embodiment of the present invention, step 4 above may include: Step 41: Based on the discrete jump field of the trailing edge space, extract continuous trailing edge projected point clouds along the vortex direction within the trailing edge vortex region. Accumulate the spatial straight-line distance difference between adjacent trailing edge projected point clouds segment by segment to characterize the ice accumulation effect, obtaining the trailing edge ice ridge cumulative thickness field reflecting the trailing edge icing accumulation state. Specifically, this includes: receiving the discrete jump field of the trailing edge space; the trailing edge vortex region is the physical area near the trailing edge of the blade affected by airflow separation and vortex effects. Within this region, airflow containing supercooled water droplets flows sequentially along the vortex direction across the trailing edge surface. Supercooled water droplets adhere and freeze segment by segment on the trailing edge surface during the flow, forming an icing layer accumulating along the vortex direction. The spatial straight-line distance difference between adjacent projected points in the discrete jump field of the trailing edge space reflects the segment-by-segment increment of the trailing edge surface undulations caused by icing adhesion along the vortex path, determining the spatial order relationship of each point in the trailing edge projected point cloud within the vortex region. The vortex direction is determined by the physical characteristics of airflow separation at the trailing edge. The airflow separates at the trailing edge of the blade, forming vortices. Fluid particles move in a direction roughly parallel to the trailing edge surface and with a certain chordal component. The vortex direction determines the sequential relationship of points in the trailing edge projection point cloud, forming a vortex direction sequence index. Following the vortex direction sequence, segments are accumulated sequentially starting from the initial point. For the first... The cumulative segment is the number of segments, and the cumulative thickness value corresponding to this segment is equal to the value from the starting segment to the current segment. The formula for calculating the sum of the differences in the straight-line distances between all segments, accumulated segment by segment, is as follows: ; In the formula, Indicates the direction along the vortex. The cumulative thickness of the trailing edge ridge corresponding to the segment, in units of length; The segment number is along the direction of the vortex, and its value is a positive integer from 1 to the total number of segments; For the first The spatial straight-line distance difference between the projection segments, in units of length; From paragraph 1 to paragraph 2 The summation of the spatial straight-line distance differences for each segment represents the cumulative thickness of the ice layer along the eddy current direction. The physical meaning of this summation is that the ice layer adheres to the trailing edge surface segment by segment along the eddy current direction. The amount of adhesion in each segment is characterized by the local undulation amplitude of the trailing edge surface (i.e., the spatial straight-line distance difference between adjacent projection points). The summation of all segments from the starting position of the eddy current direction to the current node represents the cumulative thickness of the ice layer along the eddy current streamline direction. This operation is based on the following physical assumptions: within the eddy current region, supercooled water droplets sequentially contact each segment of the trailing edge surface along the streamline. The amount of adhesion contributed to each segment is positively correlated with the local undulation amplitude of that segment's surface. After the summation of all segments is completed sequentially along the eddy current direction, each spatial node obtains a cumulative thickness value. The cumulative thickness values ​​of all nodes are stored in association with their spatial positions and arranged in the order of the eddy current direction to obtain the cumulative thickness field of the trailing edge ice ridge.

[0040] Step 42: Based on the stagnation zone classification subset, extract the continuous physical morphological boundary of the ice ridge growth zone. Based on the thickness change continuity of the cumulative thickness field of the trailing edge ice ridge constrained by the continuous physical morphological boundary, obtain the thickness continuity constraint condition for filtering dielectric fluctuation noise. Specifically, this includes: receiving the stagnation zone classification subset, which contains the spatial boundary coordinates (spanwise minimum and maximum values ​​and chordwise minimum and maximum values) of each ice ridge growth zone; extracting the spanwise and chordwise boundaries of all ice ridge growth zones from the stagnation zone classification subset; defining the continuous physical morphological boundary of the ice ridge growth zone along the blade surface as: the spatial range of the blade surface enclosed by the continuous interval from the minimum to the maximum spanwise coordinate of each ice ridge growth zone in the spanwise direction and the continuous interval from the minimum to the maximum chordwise coordinate of each ice ridge growth zone in the chordwise direction; within the coverage of this continuous physical morphological boundary, the physical process of the ice ridge attaching to the blade surface is dominated by aerodynamic continuity, and the ice layer attached to the blade surface extends continuously along both the spanwise and chordwise directions, with its thickness gradually changing smoothly along the spatial direction. This constitutes the physical premise of the thickness continuity constraint: in the cumulative thickness field of the trailing edge ice ridge, the absolute difference between the cumulative thickness value of any node located within the physical boundary of the ice ridge growth zone and the cumulative thickness value of its spatially adjacent node should not exceed the reasonable upper limit of change determined by the boundary scale of the continuous physical form. Jumps exceeding this upper limit are caused by abnormal dielectric fluctuations in the trailing edge point cloud measurement, rather than physical changes in the actual ice thickness.

[0041] Quantitative thickness constraints are derived from the continuous physical morphological boundaries of the ice ridge growth zone. An upper limit for thickness abrupt changes is set, denoted as the third threshold. The formula for calculating the third threshold is: the spatial distance between adjacent nodes multiplied by a preset maximum thickness gradient coefficient, which is 0.05. The preset maximum thickness gradient coefficient ranges from 0.02 to 0.10. Its value is based on the fact that the rate of change of ice ridge thickness along the spanwise and tangential directions on the blade surface is physically constrained by the aerodynamic adhesion process. Analysis of measured data from typical icing conditions shows that the spatial gradient of ice ridge thickness along the spanwise direction generally does not exceed 5% (i.e., the thickness change per unit length along the spanwise direction does not exceed 5% of the average thickness at that location). A value of 0.05 is taken as the typical preset maximum thickness gradient coefficient to achieve a balance between the ability to constrain noise removal and the preservation of true thickness gradient characteristics. The continuous physical morphological boundary (stretching and chordal range) of the ice ridge growth zone is combined with the upper limit of thickness jump tolerance (third threshold) to form a thickness continuity constraint condition. This constraint condition stipulates that for any node in the cumulative thickness field of the trailing edge ice ridge, if the spatial coordinates of the node fall within the physical boundary of the ice ridge growth zone, and the absolute value of the difference between the cumulative thickness value of the node and any spatially adjacent node exceeds the third threshold, then the node is judged to violate the continuity of thickness change.

[0042] Step 43: Based on the thickness continuity constraint, isolated jump points that violate the continuity of thickness variation in the cumulative thickness field of the trailing ridge are removed to obtain a clean trailing ridge thickness field free from scattering noise interference. Specifically, this includes: receiving the cumulative thickness field of the trailing ridge and the thickness continuity constraint. The generation mechanism of dielectric fluctuation scattering noise in the trailing point cloud measurement is as follows: During use, the trailing point cloud sensor is affected by factors such as local water accumulation on the blade surface and abrupt changes in dielectric constant at the ice-air interface. Individual measurement points exhibit measurement deviations from the true surface. These deviations are amplified into abnormal jump values ​​through spatial straight-line distance difference calculation. After segment-by-segment accumulation, in the cumulative thickness field of the trailing ridge, this manifests as a significant deviation between the thickness value of individual nodes and the normal continuous thickness variation trend within their spatial neighborhood. These abnormal points that deviate from the continuity of thickness variation are scattering noise. Each node in the cumulative thickness field of the trailing ridge is traversed. For each node, its set of spatially adjacent nodes is extracted, taking the nearest neighbor nodes in both the spanwise and chordwise spatial directions. Calculate the absolute difference of the cumulative thickness value between the node and each of its spatial neighbors. Compare each absolute difference with the third threshold specified by the thickness continuity constraint. When the absolute difference of the cumulative thickness between the node and a spatial neighbor exceeds the third threshold, the node is marked as an isolated transition point.

[0043] All nodes marked as isolated transition points are directly removed from the cumulative thickness field of the trailing ridge. The removal operation marks the spatial location of the corresponding node as null, and the original cumulative thickness value of the removed node is no longer included in subsequent calculations. The remaining nodes that are not removed retain their original cumulative thickness values. The cumulative thickness field after removal is the pure trailing ridge thickness field. This thickness field eliminates isolated transitions caused by dielectric wave scattering noise, while maintaining the physical continuity of the icing thickness along the spatial direction. The absolute difference in cumulative thickness between any two spatially adjacent nodes in the field does not exceed the third threshold.

[0044] In this embodiment of the invention, the spatial linear distance difference of the trailing edge projection point cloud is accumulated segment by segment along the eddy current direction to obtain the cumulative thickness field of the ice ridge. Isolated jump points are eliminated by the continuous physical morphological boundary constraint of the stationary point region classification subset. While preserving the real ice ridge thickness signal, dielectric wave scattering noise is effectively filtered out, thereby improving the spatial continuity of thickness inversion.

[0045] In a preferred embodiment of the present invention, step 5 above may include: Step 51: Based on the stagnation area classification subset, extract the physical boundary coordinates of the ice ridge growth area. Perform spatial registration on the pure ice ridge thickness field at the trailing edge based on the physical boundary coordinates to obtain the aligned trailing edge icing thickness field after eliminating spatial misalignment errors from multi-source data. Specifically, this includes receiving the stagnation area classification subset and the pure ice ridge thickness field at the trailing edge. The data for the stagnation area classification subset originates from the spatial coordinate system of the fiber optic strain measurement points, while the data for the pure ice ridge thickness field at the trailing edge originates from the spatial coordinate system of the trailing edge point cloud. There is a systematic spatial misalignment error between the two reference systems. The sources of this error include translational and rotational deviations between the installation coordinate systems of the fiber optic strain sensor and the trailing edge point cloud sensor, as well as the inconsistency of the spatial reference origins of the two types of sensors. The goal of spatial registration is to eliminate this systematic misalignment error and align the two types of data in a unified spatial coordinate system. Extract the spatial range of the physical boundary coordinates of all ice ridge growth areas from the stagnation area classification subset, and calculate the geometric center point of this spatial range. The spanwise coordinate center value is the average of the spanwise minimum and maximum values ​​of all ice ridge growth areas, and the chordwise coordinate center value is calculated similarly. The geometric center point is the registration reference point. The three-dimensional spatial coordinates of all nodes in the trailing edge pure ice ridge thickness field are translated globally. The translation vector equals the three-dimensional spatial coordinates of the registration reference point minus the spatial reference origin coordinates of the trailing edge pure ice ridge thickness field. The translation operation aligns the trailing edge thickness field data spatially with the data of the stationary area classification subset, using the registration reference point as the common origin. The translation operation does not change the thickness values ​​of each node in the trailing edge pure ice ridge thickness field, but adjusts their spatial coordinates. The translated trailing edge thickness field is denoted as the aligned trailing edge icing thickness field.

[0046] Step 52: Based on the aligned trailing edge icing thickness field and the thickness data of the stagnation zone classification subset, extract the grid surface regions with missing thickness data along the blade spanwise to obtain the set of interpolation regions reflecting the multi-source data fusion faults. Specifically, this includes receiving the aligned trailing edge icing thickness field and the stagnation zone classification subset. The spatial coverage relationship between the two data sources is as follows: fiber optic strain gauges mainly cover the leading edge region of the blade, the stagnation zone classification subset provides thickness-related data within the ice ridge growth zone, and the trailing edge point cloud mainly covers the trailing edge region of the blade. The aligned trailing edge icing thickness field provides icing thickness data for the trailing edge region. Along the blade spanwise, the leading edge and trailing edge data form spatial complementarity at the chordal position, jointly covering most of the circumferential range of the blade. However, at certain locations along the spanwise direction, there may be data gaps where neither the leading edge fiber optic gauges nor the trailing edge point cloud cover the data.

[0047] A full-range scan is performed along the spanwise physical extension of the blade, from the spanwise coordinates at the leaf root to the spanwise coordinates at the leaf tip. At each spanwise scan location, the existence of valid icing thickness data is checked. Valid thickness data is determined by the following criteria: the grid node corresponding to the spanwise location possesses ridge growth zone thickness information from the stationary zone classification subset, or possesses thickness values ​​from the aligned trailing edge icing thickness field. When a grid node at a spanwise location has neither stationary zone thickness data nor trailing edge thickness data, that location is marked as a data-missing grid surface. All grid surfaces marked as data-missing are arranged sequentially along the spanwise direction. Continuously distributed data-missing grid surfaces along the spanwise direction are merged into a single data-missing segment. Each data-missing segment records its spanwise start and end coordinates. The sum of all data-missing segments constitutes the set of regions to be interpolated.

[0048] Step 53: Based on the set of regions to be interpolated, linear interpolation is performed using the thickness data of adjacent grid surfaces to fill data gaps, resulting in a full spanwise icing thickness distribution matrix containing complete spanwise spatial information. Specifically, this includes: receiving the set of regions to be interpolated; and performing spanwise linear interpolation for each missing data segment in the set. The spanwise linear interpolation process is as follows: Find the effective thickness data immediately adjacent to the missing data segment on both sides in the spanwise direction: the nearest grid surface node with effective thickness data before the spanwise start position of the missing segment is taken as the left neighbor node, and the nearest grid surface node with effective thickness data after the spanwise end position of the missing segment is taken as the right neighbor node. The spanwise coordinates and icing thickness values ​​of both the left and right neighbor nodes are known. Along the spanwise direction between the left and right neighbor nodes, the interpolated thickness value is calculated for each missing grid surface node within the missing data segment. The interpolated thickness value is obtained by linearly weighting the known thickness values ​​of the left and right neighbor nodes according to the spanwise distance. The formula for calculating the spanwise linear interpolation is: ; In the formula, spanwise coordinate position is The interpolated ice thickness value for the missing mesh surface nodes, in units of length; The coordinates of the currently missing mesh node along the blade span are given in length units, satisfying the following conditions: ; This represents the ice thickness value of the left neighbor node, expressed in units of length. This represents the ice thickness value of the right neighbor node, expressed in units of length. The coordinates of the left adjacent node along the blade span are expressed in units of length. The coordinates of the right adjacent node along the blade span are expressed in units of length. This represents the thickness difference between the right and left neighboring nodes, reflecting the total change in thickness between the two neighboring nodes. This represents the outward distance from the current node to its left neighbor. This represents the average linear rate of change of thickness along the span between left and right adjacent nodes; The interpolated thickness increment from the left neighbor node to the current node is based on the physical principle that the spatial variation of ice ridge thickness along the blade spanwise between adjacent valid data nodes approximates a linear trend under the condition that the spacing between adjacent nodes is reasonable. This linearization assumption has an acceptable level of accuracy in engineering icing thickness detection applications. After sequentially filling all missing nodes in all data missing segments of the set to be interpolated, the interpolated thickness values ​​are integrated with the original valid thickness data into the full spanwise grid. The icing thickness values ​​of each grid surface node are arranged in ascending order of spanwise coordinates, and each node is associated with three attributes: its spanwise coordinates, chordal coordinates, and thickness value, forming a full spanwise icing thickness distribution matrix.

[0049] In this embodiment of the invention, spatial registration is performed using the physical boundary coordinates of the ice ridge growth zone as a reference to eliminate misalignment errors of multi-source data, and linear interpolation is performed to fill the fault area of ​​data fusion, outputting a full spanwise ice thickness distribution matrix, thus realizing the complete fusion of leading-edge fiber data and trailing-edge point cloud data under a unified spanwise coordinate system.

[0050] In a preferred embodiment of the present invention, step 6 above may include: Step 61: Based on the spanwise icing thickness distribution matrix, extract the icing thickness values ​​of adjacent physical triangular mesh nodes arranged along the spanwise physical extension direction of the blade, and arrange them continuously according to the spanwise physical extension order to obtain a spanwise thickness value sequence. Specifically, this includes: receiving the spanwise icing thickness distribution matrix, where each node is arranged in ascending order of spanwise coordinates; extracting the icing thickness values ​​corresponding to adjacent physical triangular mesh nodes sequentially along the spanwise physical extension direction of the blade, i.e., from the blade root coordinates to the blade tip coordinates; determining the adjacency relationship between nodes based on spanwise coordinate adjacency: the pair of nodes with the smallest spanwise coordinate difference constitutes a spanwise adjacent node pair; and arranging the extracted icing thickness values ​​sequentially from the first mesh node at the blade root to the last mesh node at the blade tip, with each thickness value associated with its corresponding spanwise coordinate and mesh node number. The arrangement result is a one-dimensional thickness value sequence along the spanwise direction, denoted as the spanwise thickness value sequence.

[0051] Step 62: Based on the spanwise thickness numerical sequence, calculate the absolute difference in icing thickness values ​​between adjacent physical triangular mesh nodes to obtain the spanwise thickness difference sequence. Specifically, this includes: receiving the spanwise thickness numerical sequence, and the... The thickness value and the first Each thickness value corresponds to the ice thickness of two adjacent grid surface nodes along the spanwise direction. The absolute value of the difference between the two values ​​is calculated to obtain the variation in ice thickness along the spanwise direction between adjacent nodes. The formula for calculating the absolute difference in ice thickness between adjacent nodes is: ; In the formula, For the first The node and the first The absolute difference in ice thickness along the span between nodes, in units of length; For the first The ice thickness value for each grid node, in units of length; For the first The ice thickness value for each grid node, in units of length; The absolute difference between two thicknesses is calculated by eliminating the influence of the sign of the difference and retaining only the magnitude of the change. The physical meaning of the absolute thickness difference is the magnitude of the change in ice thickness along the spanwise direction between adjacent spanwise positions. The larger this magnitude value, the more significant the difference in ice thickness within that spanwise section. In spanwise sections with a large difference in ice thickness, the difference in mechanical stress along the spanwise direction within the ice layer is also larger, and the risk of ice layer detaching under its own gravity, centrifugal force, and aerodynamic force is higher. The absolute difference is calculated for each adjacent pair in the spanwise thickness value sequence, and all absolute differences are arranged in spanwise order to obtain the spanwise thickness difference value sequence.

[0052] Step 63: Based on the spanwise thickness difference sequence, map the absolute differences back to the physical locations of the corresponding physical triangular mesh nodes, and retain the three-dimensional spatial coordinate indices to obtain the full spanwise thickness difference field. Specifically, this includes: receiving the spanwise thickness difference sequence, where each absolute difference in the sequence corresponds in physical space to a segment between a pair of adjacent mesh nodes in the spanwise direction. Mapping each absolute difference from the one-dimensional spanwise sequence back to the three-dimensional physical space. The mapping method is as follows: for the... The absolute difference is taken as the physical location of the nth absolute difference. The node and the first The midpoint of the space between nodes is the three-dimensional spatial coordinate of which is obtained by taking the arithmetic mean of the three-dimensional spatial coordinates of the two adjacent nodes in three components. The coordinates of each three-dimensional spatial midpoint are associated with and stored with its corresponding absolute thickness difference. At the same time, the spanwise index is retained to enable tracing from the spanwise direction to the three-dimensional space. All entries are organized in ascending order of the spanwise direction to form a full spanwise thickness difference field. This difference field encodes the segmented gradient information of the icing thickness along the spanwise direction in the form of a three-dimensional spatial scalar field.

[0053] In this embodiment of the invention, the absolute difference of the ice thickness values ​​of adjacent grid surface nodes is calculated along the spanwise physical extension sequence and mapped back to the physical location to form a spanwise thickness difference field. The absolute value information of ice thickness is transformed into spanwise variation gradient information, providing a quantitative basis for the spatial positioning of ice shedding trend.

[0054] In a preferred embodiment of the present invention, step 6 above may include: Step 64: Based on the spanwise thickness difference field, extract the physical triangular mesh nodes whose absolute difference satisfies the preset detachment critical condition to obtain the initial detachment risk node set. Specifically, this includes: receiving the spanwise thickness difference field; the physical critical condition for ice detachment from the blade surface is: when the difference in ice thickness between two adjacent nodes along the spanwise direction exceeds a certain critical threshold, the shear stress generated along the spanwise direction within that ice layer reaches or exceeds the bonding strength between the ice and the blade surface, indicating a risk of ice detachment. A detachment critical threshold is set, denoted as the fourth threshold. The fourth threshold is used to determine whether the absolute difference in ice thickness between adjacent nodes along the spanwise direction reaches the level indicating a risk of ice detachment. The specific meaning of the fourth threshold is: the upper limit of the maximum allowable variation in ice thickness between adjacent nodes along the spanwise direction. The fourth threshold depends on the spanwise distance between adjacent nodes and the mechanical properties of the ice. The fourth threshold is calculated by multiplying the distance between adjacent nodes by a preset detachment gradient coefficient. The preset detachment gradient coefficient is 0.08. The value range of the preset detachment gradient coefficient is 0.05 to 0.15. The values ​​are based on the following: the shear bond strength between ice and the blade surface coating is generally in the range of 0.1 to 1.0 MPa. The interlaminar shear stress within the ice layer caused by the change in ice thickness along the span is proportional to the thickness gradient. When the thickness gradient exceeds 8%, the corresponding interlaminar shear stress reaches the lower limit of the typical bond strength range, and the risk of detachment increases significantly. Under low-temperature, high-density icing conditions, a preset detachment gradient coefficient of 0.05 can be used to provide a conservative early warning threshold; under high-temperature, low-density icing conditions, a coefficient of 0.15 can be used to reduce the false alarm rate.

[0055] For each entry in the spanwise thickness difference field, its absolute thickness difference is extracted and compared with a fourth threshold. When the absolute thickness difference is greater than the fourth threshold, both adjacent grid surface nodes corresponding to the entry are marked as initial detachment risk nodes. When the absolute thickness difference is not greater than the fourth threshold, no marking operation is performed. All entries in the spanwise thickness difference field are traversed, and all marked grid surface nodes are deduplicated to form the initial detachment risk node set.

[0056] Step 65: Based on the initial set of detachment risk nodes, extract the physically connected domains composed of initially detachment risk nodes that are adjacent in three-dimensional spatial coordinates and physically continuous, to obtain the initial detachment risk physical region. Specifically, this includes: receiving the initial set of detachment risk nodes; the icing detachment risk has continuity in spatial distribution on the blade surface; and the actual risk region consists of multiple risk nodes that are continuously distributed in space. The initial set of detachment risk nodes may contain isolated single-node markers, or mislabeling caused by local thickness measurement anomalies or data fluctuations. For all nodes in the initial set of detachment risk nodes, construct a spatial adjacency graph. In the spatial adjacency graph, nodes are the sequence numbers of risk nodes; there is an undirected edge between two nodes if and only if these two nodes share a physical edge in the triangular mesh face set on the blade surface. Perform connected component analysis on the spatial adjacency graph, using a breadth-first search strategy, starting from any unclassified node, and classify all nodes reachable along the edge into the same connected domain. Perform physical scale filtering on each connected domain. A minimum node count threshold is set for the connected regions with detachment risk, denoted as the fifth threshold. The fifth threshold is set to 5 grid surface nodes, with a range of 3 to 8 nodes. The basis for this value is that the physical process of ice detachment requires the risk area to have sufficient spatial extent to generate macroscopic detachment mechanical conditions. Isolated connected regions containing fewer than 3 grid surface nodes have insufficient physical coverage to form a true ice detachment risk area and are classified as anomalies and removed. Connected regions with a node count not less than the fifth threshold are marked to constitute the initial physical detachment risk area.

[0057] Step 66: Based on the initial detachment risk physical region, and combined with the physical boundary coordinates along the blade spanwise, a region expansion process with a preset physical step size is performed to obtain the icing detachment warning zone. Specifically, this includes: receiving the initial detachment risk physical region. The boundary of the initial detachment risk physical region only includes the spatial range of the risk nodes directly marked in step 64. At the physical level, once icing detachment occurs near the boundary of this region, stress release may affect the adjacent area outside the boundary, causing the detachment range to expand outward along the spanwise and chordwise directions. To form a conservative and safe warning range, the boundary of the initial detachment risk physical region is expanded, and an expansion physical step size is set, denoted as the sixth threshold. The sixth threshold is used to determine the distance the boundary of the initial detachment risk physical region expands outward. The value of the sixth threshold is the spacing between three fiber strain measurement points. The sixth threshold ranges from 2 to 5 measurement point intervals. The basis for this value is that the cascading expansion range of ice shedding along the span generally does not exceed the characteristic scale determined by the local curvature change in the span of the blade. Using a 3 measurement point interval as the expansion step size provides a moderate safety margin between the warning zone and the actual shedding range.

[0058] The operation of the region expansion process is as follows: For each connected component of the initial icing risk physical region, the boundary of the connected component is expanded outward by a physical step specified by the sixth threshold in both the spanwise and chordwise directions. The expanded region boundary is obtained by adding the expansion step to the original connected component boundary. All nodes inside the expanded region are included in the warning range. The expansion process for each connected component requires checking the spanwise physical boundary coordinates (root and tip coordinates) and the chordwise physical boundary coordinates (leading and trailing edge coordinates). When the expanded region boundary exceeds the blade physical boundary, the blade physical boundary is used as the cutoff point. This operation ensures that the warning area does not exceed the physical range of the blade itself. All expanded connected components are merged, and after removing the overlapping parts, the icing detachment warning area is formed.

[0059] In this embodiment of the invention, nodes whose absolute differences satisfy the critical condition for detachment are extracted to form an initial risk node set. The initial detachment risk area is obtained by merging the physical connected domains and the area is expanded to generate an ice detachment warning area. The ice detachment risk range is marked by spatial boundary demarcation, providing clear physical area guidance for de-icing operations rather than just giving a single-point alarm.

[0060] like Figure 3 As shown, embodiments of the present invention also provide an adaptive access and intelligent cleaning system for multi-source heterogeneous data, including: The module is used to construct triangular mesh surfaces connecting adjacent measurement points based on the sequence of fiber strain measurement points arranged along the blade span and their three-dimensional spatial coordinates, calculate the spatial normal vector of each triangular mesh surface, and obtain the set of mesh surface normal vectors. The labeling module is used to calculate the dihedral angle between the spatial normal vectors of adjacent triangular mesh surfaces based on the set of mesh surface normal vectors, and to label the mesh surface regions where the dihedral angle represents a local morphological change as ice ridge growth areas, thus obtaining a classification subset of stationary areas; The calculation module is used to obtain the trailing edge point cloud sequence and its spatial coordinates, determine the spatial projection reference of the trailing edge point cloud sequence based on the classification subset of the stagnation area, calculate the difference in spatial straight-line distance between adjacent point clouds under the spatial projection reference, and obtain the spatial discrete jump field of the trailing edge. The elimination module is used to accumulate the spatial straight-line distance difference segment by segment along the vortex direction in the trailing vortex region based on the discrete jump field of the trailing space to obtain the cumulative thickness field of the trailing ice ridge; and to eliminate the scattering noise points in the cumulative thickness field of the trailing ice ridge based on the classification subset of the stagnation area to obtain the pure thickness field of the trailing ice ridge. The interpolation module is used to perform spatial coordinate alignment and data interpolation along the blade span based on the classification subset of the stationary area and the thickness field of the pure ice ridge at the trailing edge, so as to obtain the full spanwise ice thickness distribution matrix. The extraction module is used to calculate the thickness difference between adjacent spanwise positions based on the spanwise ice thickness distribution matrix, and obtain the spanwise thickness difference field; based on the spanwise thickness difference field, it extracts the physical region that represents the ice detachment trend by the thickness difference, and obtains the ice detachment early warning zone.

[0061] It should be noted that this system is a system corresponding to the above method. All implementation methods in the above method embodiments are applicable to this embodiment and can achieve the same technical effect.

[0062] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for adaptive access and intelligent cleaning of multi-source heterogeneous data, characterized in that, The method includes: Based on the sequence of fiber strain measurement points arranged along the blade span and their three-dimensional spatial coordinates, a triangular mesh surface connecting adjacent measurement points is constructed, and the spatial normal vector of each triangular mesh surface is calculated to obtain the set of mesh surface normal vectors. Based on the set of grid surface normal vectors, the dihedral angle between the spatial normal vectors of adjacent triangular grid surfaces is calculated. The grid surface regions where the dihedral angle represents a local morphological abrupt change are marked as ice ridge growth zones, thus obtaining a classification subset of stationary areas. Obtain the trailing edge point cloud sequence and its spatial coordinates. Determine the spatial projection reference of the trailing edge point cloud sequence based on the classification subset of the stagnation area. Calculate the spatial straight-line distance difference between adjacent point clouds under the spatial projection reference to obtain the spatial discrete jump field of the trailing edge. Based on the discrete jump field of the trailing edge space, the spatial straight-line distance difference is accumulated segment by segment along the vortex direction in the trailing edge vortex region to obtain the cumulative thickness field of the trailing edge ice ridge; the scattering noise points in the cumulative thickness field of the trailing edge ice ridge are removed according to the classification subset of the stagnation area to obtain the pure thickness field of the trailing edge ice ridge. Based on the classification subset of the stationary area and the thickness field of the pure ice ridge at the trailing edge, spatial coordinate alignment and data interpolation are performed along the blade span to obtain the full span ice thickness distribution matrix. Based on the spanwise icing thickness distribution matrix, the thickness difference between adjacent spanwise locations is calculated to obtain the spanwise thickness difference field; based on the spanwise thickness difference field, the physical region that characterizes the icing detachment trend by the thickness difference is extracted to obtain the icing detachment early warning zone.

2. The adaptive access and intelligent cleaning method for multi-source heterogeneous data according to claim 1, characterized in that, Based on the sequence of fiber strain gauge points arranged along the blade span and their three-dimensional spatial coordinates, a triangular mesh surface connecting adjacent gauge points is constructed. The spatial normal vector of each triangular mesh surface is calculated, resulting in a set of mesh surface normal vectors, including: Based on the sequence of fiber strain measurement points arranged along the blade span and their three-dimensional spatial coordinates, the three-dimensional spatial coordinates of three adjacent fiber strain measurement points are extracted. Physical triangular mesh surfaces are constructed by connecting adjacent measurement points with spatial straight lines to obtain a set of triangular mesh surfaces on the blade surface. Based on the set of triangular mesh surfaces on the blade surface, extract the spatial direction vectors of the two intersecting physical edges in each physical triangular mesh surface, calculate the cross product of the spatial direction vectors of the two intersecting physical edges to obtain the vertical normal direction, and obtain the spatial normal vector of each physical triangular mesh surface. Based on the spatial normal vector of each physical triangular mesh surface, the spatial normal vectors of all physical triangular mesh surfaces are summarized in the order of physical extension along the blade span, and spatial coordinate index mapping is performed to obtain the set of mesh surface normal vectors.

3. The adaptive access and intelligent cleaning method for multi-source heterogeneous data according to claim 2, characterized in that, Based on the set of grid surface normal vectors, the dihedral angle between the spatial normal vectors of adjacent triangular grid surfaces is calculated. Grid surface regions where the dihedral angle represents a local morphological abrupt change are marked as ridge growth zones, resulting in a subset of stationary point classifications, including: Based on the set of normal vectors of the mesh surface, extract the spatial normal vectors of two adjacent physical triangular mesh surfaces, calculate the spatial angle between the two spatial normal vectors, and obtain the set of dihedral angles of adjacent mesh surfaces. Based on the set of dihedral angles of adjacent grid surfaces, physical triangular grid surfaces whose dihedral angle values ​​satisfy the critical conditions for dihedral angles characterizing ice ridge growth are extracted to obtain the set of morphologically abrupt grid surfaces. Based on the set of morphologically aberrant grid surfaces, the physical region at the leading edge of the blade where the morphologically aberrant grid surface is located is marked as the ice ridge growth zone. The spatial boundary coordinates and normal vector distribution features of the ice ridge growth zone are extracted to obtain a subset of stationary areas containing spatial boundary information of the ice ridge growth zone.

4. The adaptive access and intelligent cleaning method for multi-source heterogeneous data according to claim 3, characterized in that, Based on the set of dihedral angles of adjacent grid surfaces, physically triangular grid surfaces whose dihedral angle values ​​satisfy the critical conditions for characterizing ice ridge growth are extracted, resulting in a set of morphologically abrupt grid surfaces, including: Based on the set of dihedral angles of adjacent grid faces, extract the dihedral angle values ​​between each physical triangular grid face and all its adjacent physical triangular grid faces, calculate the absolute value of the difference between adjacent dihedral angle values, and obtain the set of local dihedral angle change rates. Based on the set of local dihedral angle change rates, physical triangular mesh surfaces whose local dihedral angle change rate values ​​satisfy the preset morphological abrupt change critical conditions are extracted to obtain a set of candidate abrupt change mesh surfaces. Based on the candidate mutation mesh surface set, the three-dimensional spatial connectivity distribution features of the candidate mutation mesh surfaces in the physical region of the blade leading edge are extracted and the spatial connectivity domains are merged to obtain the morphological mutation mesh surface set.

5. The adaptive access and intelligent cleaning method for multi-source heterogeneous data according to claim 3, characterized in that, Obtain the trailing edge point cloud sequence and its spatial coordinates. Determine the spatial projection reference of the trailing edge point cloud sequence based on the stagnation zone classification subset. Calculate the spatial straight-line distance difference between adjacent point clouds under the spatial projection reference to obtain the trailing edge spatial discrete jump field, including: Based on the classification subset of the stationing area, the spatial normal vector distribution characteristics of the ice ridge growth area are extracted, and a spatial projection reference surface of the trailing edge point cloud sequence orthogonal to the average normal vector of the ice ridge growth area is constructed to obtain the spatial projection reference. Based on the spatial projection reference, the acquired trailing edge point cloud sequence and its three-dimensional spatial coordinates are projected onto the spatial projection reference plane along the normal direction of the spatial projection reference plane to obtain the trailing edge projection point cloud sequence. Based on the trailing edge projection point cloud sequence, the spatial straight-line distance difference between adjacent trailing edge projection point clouds is calculated to characterize the local undulation and jump features of the trailing edge surface, thus obtaining the trailing edge spatial discrete jump field.

6. The adaptive access and intelligent cleaning method for multi-source heterogeneous data according to claim 4, characterized in that, Step 4: Based on the discrete jump field of the trailing edge space, accumulate the spatial straight-line distance difference segment by segment along the vortex direction within the trailing edge vortex region to obtain the cumulative thickness field of the trailing edge ice ridge; remove scattering noise points in the cumulative thickness field of the trailing edge ice ridge based on the stagnation region classification subset to obtain the pure thickness field of the trailing edge ice ridge, including: Based on the discrete jump field of the trailing edge space, continuous trailing edge projection point clouds are extracted along the vortex direction in the trailing edge vortex region. The spatial straight-line distance difference between adjacent trailing edge projection point clouds is accumulated segment by segment to characterize the ice layer accumulation effect, and the cumulative thickness field of the trailing edge ice ridge reflecting the ice accumulation state of the trailing edge is obtained. Based on the classification subset of the stationary area, the continuous physical morphological boundary of the ice ridge growth area is extracted. Based on the continuous physical morphological boundary constraining the thickness change continuity of the cumulative thickness field of the trailing ice ridge, the thickness continuity constraint condition for filtering out dielectric fluctuation noise is obtained. Based on the thickness continuity constraint, isolated jump points that violate the thickness change continuity in the cumulative thickness field of the trailing ridge are removed, resulting in a pure trailing ridge thickness field free from scattering noise interference.

7. The adaptive access and intelligent cleaning method for multi-source heterogeneous data according to claim 5, characterized in that, Based on the classification subset of the stationary area and the thickness field of the pure ice ridge at the trailing edge, spatial coordinate alignment and data interpolation are performed along the blade span to obtain the spanwise icing thickness distribution matrix, including: Based on the classification subset of the stationing area, the physical boundary coordinates of the ice ridge growth area are extracted. Based on the physical boundary coordinates, the spatial position registration of the pure ice ridge thickness field at the trailing edge is performed to obtain the aligned trailing edge ice thickness field after eliminating the spatial misalignment error of multi-source data. Based on the thickness data of the aligned trailing edge icing thickness field and the classification subset of the stagnation area, the grid surface regions with missing thickness data are extracted along the blade span to obtain a set of interpolation regions reflecting the multi-source data fusion fault. Based on the set of regions to be interpolated, linear interpolation calculations are performed using the thickness data of adjacent grid surfaces to fill data gaps, resulting in a full spanwise ice thickness distribution matrix containing complete spanwise spatial information.

8. The adaptive access and intelligent cleaning method for multi-source heterogeneous data according to claim 6, characterized in that, Based on the spanwise ice thickness distribution matrix, the thickness difference between adjacent spanwise locations is calculated to obtain the spanwise thickness difference field. Based on the overall spanwise thickness difference field, the physical regions characterizing the icing detachment trend are extracted, resulting in icing detachment early warning zones, including: Based on the spanwise icing thickness distribution matrix, the icing thickness values ​​of adjacent physical triangular mesh nodes arranged along the spanwise physical extension direction of the blade are extracted and continuously arranged in the spanwise physical extension order to obtain the spanwise thickness value sequence. Based on the spanwise thickness numerical sequence, the absolute difference in ice thickness values ​​between adjacent physical triangular mesh nodes is calculated to obtain the spanwise thickness difference sequence. Based on the spanwise thickness difference sequence, the absolute difference is mapped back to the physical location of the corresponding physical triangular mesh surface node, and the three-dimensional spatial coordinate index is retained to obtain the full spanwise thickness difference field.

9. The adaptive access and intelligent cleaning method for multi-source heterogeneous data according to claim 7, characterized in that, Based on the overall spanwise thickness difference field, the physical regions characterizing the icing detachment trend are extracted, resulting in icing detachment early warning zones, including: Based on the full span thickness difference field, physical triangular mesh nodes whose absolute difference satisfies the preset critical detachment condition are extracted to obtain the initial set of detachment risk nodes. Based on the initial set of detachment risk nodes, the physical connected domains formed by the initial detachment risk nodes that are adjacent in three-dimensional spatial coordinates and physically continuous are extracted to obtain the initial detachment risk physical region. Based on the initial physical area of ​​icing risk, and combined with the physical boundary coordinates of the blade spanwise, a region expansion process with a preset physical step size is performed to obtain the icing detachment warning zone.

10. An adaptive access and intelligent cleaning system for multi-source heterogeneous data, wherein the system implements the method as described in any one of claims 1 to 9, characterized in that, include: The module is used to construct triangular mesh surfaces connecting adjacent measurement points based on the sequence of fiber strain measurement points arranged along the blade span and their three-dimensional spatial coordinates, calculate the spatial normal vector of each triangular mesh surface, and obtain the set of mesh surface normal vectors. The labeling module is used to calculate the dihedral angle between the spatial normal vectors of adjacent triangular mesh surfaces based on the set of mesh surface normal vectors, and to label the mesh surface regions where the dihedral angle represents a local morphological change as ice ridge growth areas, thus obtaining a classification subset of stationary areas; The calculation module is used to obtain the trailing edge point cloud sequence and its spatial coordinates, determine the spatial projection reference of the trailing edge point cloud sequence based on the classification subset of the stagnation area, calculate the difference in spatial straight-line distance between adjacent point clouds under the spatial projection reference, and obtain the spatial discrete jump field of the trailing edge. The elimination module is used to accumulate the spatial straight-line distance difference segment by segment along the vortex direction in the trailing vortex region based on the discrete jump field of the trailing space to obtain the cumulative thickness field of the trailing ice ridge; and to eliminate the scattering noise points in the cumulative thickness field of the trailing ice ridge based on the classification subset of the stagnation area to obtain the pure thickness field of the trailing ice ridge. The interpolation module is used to perform spatial coordinate alignment and data interpolation along the blade span based on the classification subset of the stationary area and the thickness field of the pure ice ridge at the trailing edge, so as to obtain the full spanwise ice thickness distribution matrix. The extraction module is used to calculate the thickness difference between adjacent spanwise positions based on the spanwise ice thickness distribution matrix, and obtain the spanwise thickness difference field. Based on the overall spanwise thickness difference field, the physical regions that characterize the icing detachment trend by thickness difference are extracted to obtain the icing detachment early warning zone.

Images