Industrial internet data deep mining evaluation method

Abstract

The invention provides an industrial internet data deep mining evaluation method, which comprises the following steps of: fitting and complementing a pressure attenuation trend between adjacent sensors according to a pressure state of each monitoring point and a spatial proportional relationship between the sensors and elbow distribution distances, identifying virtual pressure monitoring points at elbow distribution positions, and determining the pressure attenuation trend of the adjacent sensors according to the virtual pressure monitoring points; determining a continuous pressure monitoring coverage range; carrying out global scanning on a continuous pressure monitoring coverage area, determining a local pressure drop sudden change section, and positioning starting and ending position coordinates of a potential sudden change area by evaluating the pressure change slope of each point location in the section; identifying a core abnormal point in the monitoring fragment set of the related pressure drop mutation of the elbow, identifying an acquisition blind area by comparing the coverage overlapping degree of the coordinates of the core abnormal point and the actual sensor mounting position, and evaluating the acquisition omission degree according to the acquisition blind area; and a monitoring point position is additionally arranged in the acquisition blind area according to the acquisition omission degree, and a pressure monitoring coverage range covering the pressure drop sudden change section is obtained.

Description

Technical Field

[0001] This invention relates to the field of information technology, and in particular to a method for in-depth data mining and evaluation in the industrial internet. Background Technology

[0002] In modern industrial production systems, fluid transport pipelines are not only the physical carriers of logistics transmission but also a crucial component of edge sensing within the industrial internet architecture. With the deepening application of industrial internet technology, the digital transformation of pipeline systems has moved from simple equipment networking to data-driven precision operation and maintenance. By deploying high-precision pressure sensors, the fluid dynamics state within the physical pipeline is converted into continuous digital signals, uploaded in real-time to the industrial internet platform, achieving real-time mapping of pipeline operating status in the digital space. However, existing industrial internet monitoring solutions mostly focus on high-frequency and low-latency data transmission, neglecting the effectiveness of the spatial distribution of data collection points. Especially for complex pipeline networks, while industrial internet platforms can aggregate massive amounts of pressure time-series data, spatial sampling distortion occurs if the physical spacing of sensor installations is disproportionate to the distribution spacing of critical components such as pipe bends. In actual operating conditions, locations with geometrical abrupt changes, such as bends, are often areas of high pressure energy consumption and high rates of internal scaling. If the spatial span of sensors covers multiple bends without collection points, the abrupt pressure drop changes (i.e., pressure drop abrupt changes) caused by the fluid flowing through the bends will be masked by the long-distance average values. This blind spot in the physical dimension means that while cloud platforms possess continuous temporal data, they lose crucial spatial characteristics. This prevents the system from using details of pressure fluctuations to infer the degree of scaling or localized blockages on the pipe's inner wall. Therefore, identifying these blind spots in digital sensing by analyzing the spatial ratio between pressure acquisition results and the distribution of physical bends, based on existing data from industrial internet platforms, and then optimizing the coverage logic of the monitoring network accordingly, has become a key technical challenge in improving the intelligent operation and maintenance level of pipeline systems. Summary of the Invention

[0003] This invention provides a method for in-depth data mining and evaluation in the industrial internet, the method comprising:

[0004] The system collects monitoring values ​​from pipeline pressure sensors, the installation spacing between adjacent sensors, and the coordinates of pipeline bend locations through an industrial internet platform. It identifies the pressure state at each monitoring point and calculates the spatial ratio between the sensor installation spacing and the bend distribution spacing. Based on the pressure state and the spatial ratio, it fits and completes the pressure attenuation trend between adjacent sensors, generating virtual pressure monitoring points at the bend locations to determine the continuous pressure monitoring coverage area. A full-area scan of the continuous pressure monitoring coverage area identifies local pressure drop abrupt change zones, calculates the pressure change slope at each point within these zones, and locates the start and end coordinates of potential abrupt change zones. Finally, it performs a spatial correlation between the start and end coordinates of these potential abrupt change zones and the pipeline bend distribution locations. Inter-regional comparison is performed to identify overlapping areas and mark them as elbow-induced pressure drop abrupt change areas. Monitoring segments containing these elbow-induced pressure drop abrupt change areas are extracted to form a monitoring segment set. Core anomaly points are identified in the monitoring segment set, and the overlap between the coordinates of the core anomaly points and the actual sensor installation locations is compared to identify blind spots and quantify the degree of data acquisition omission. Based on the degree of data acquisition omission, additional monitoring points are generated within the blind spots to form a pressure monitoring coverage area covering the pressure drop abrupt change area. Fluid resistance analysis is performed on the pressure monitoring coverage area to identify abnormal pressure drop points at non-elbow locations. By combining fluid flow velocity with reverse deduction of the pipe inner wall roughness coefficient and effective flow diameter, the coordinates of the scaling location on the pipe inner wall are determined and the scaling layer thickness is quantified.

[0005] Furthermore, the step of collecting pipeline pressure sensor monitoring values, adjacent sensor installation spacing, and pipeline bend distribution coordinates through an industrial internet platform, identifying the pressure status of each monitoring point, and calculating the spatial ratio between sensor installation spacing and bend distribution spacing includes:

[0006] The system acquires real-time monitoring values ​​and installation location coordinates of pipeline pressure sensors, calculates the pipe segment length between adjacent sensors, extracts elbow location coordinates from pipeline topology data and calculates the distribution spacing between adjacent elbows, calculates the spatial ratio value based on the pipe segment length and the distribution spacing between adjacent elbows, marks pipe segments that exceed the sparse threshold or are below the dense threshold, and constructs a spatial distribution ratio map of sensors and elbows.

[0007] Furthermore, the step of performing a full-area scan of the continuous pressure monitoring coverage area to identify local pressure drop abrupt change sections, and locating the start and end coordinates of potential abrupt change areas by evaluating the pressure change slope at each point within the section, includes:

[0008] A sliding window scan is performed on the monitoring points within the continuous pressure monitoring coverage area. The window includes adjacent monitoring points. The sum of squares of the deviations of the pressure values ​​within the window from the mean is calculated. Pipe sections where the sum of squares of deviations exceeds the fluctuation threshold are marked as local pressure drop abrupt change sections. The pressure values ​​and location coordinates of the monitoring points within the local pressure drop abrupt change sections are fitted with linear regression coefficients using the least squares method as the pressure change slope. The slope values ​​of adjacent monitoring points are compared, and the locations where the slope changes exceed the change threshold are identified as inflection points. Monitoring points with slopes less than the stability threshold are searched forward along the pipeline direction as the abrupt change starting point, and monitoring points with slopes recovering to below the stability threshold are searched backward as the abrupt change ending point. The coordinates of the abrupt change starting point and the abrupt change ending point are obtained.

[0009] Furthermore, based on the pressure status of each monitoring point and the spatial ratio between the sensor and the elbow distribution spacing, the pressure attenuation trend between adjacent sensors is fitted and completed to identify virtual pressure monitoring points at the elbow distribution location and determine the continuous pressure monitoring coverage area, including:

[0010] A pressure decay trend curve is generated by fitting the pressure values ​​between adjacent sensors using cubic spline interpolation. The pressure value and pressure change rate are obtained by numerical sampling on the pressure decay trend curve. When the pressure change rate exceeds the sudden change threshold, a virtual pressure monitoring point is created. The virtual pressure monitoring point and the actual sensor monitoring point are combined to form a monitoring point set. The area where the distance between adjacent monitoring points is less than the continuous coverage threshold is determined as the continuous pressure monitoring coverage area.

[0011] Furthermore, the starting and ending coordinates of the potential abrupt change region are spatially compared with the distribution locations of pipeline bends to identify overlapping areas and mark them as bend-induced pressure drop abrupt change regions. Monitoring segments containing these bend-induced pressure drop abrupt change regions are extracted to form a monitoring segment set, including:

[0012] Calculate the minimum distance from the boundary point of the abrupt change region to the center of the bend. If the distance is less than the spatial overlap threshold, mark it as a bend-induced pressure drop abrupt change region. Extract continuous monitoring data sequences covering the bend-induced pressure drop abrupt change region to form monitoring segments, and summarize them to form a monitoring segment set.

[0013] Furthermore, for the monitoring segment set, core anomalies are identified, and the overlap between the coordinates of the core anomalies and the actual sensor installation locations is compared to identify blind spots and quantify the degree of data loss, including:

[0014] Pressure gradients are calculated for the monitoring segment set. The isolated forest algorithm is used to identify data points whose pressure change rate exceeds the abnormal threshold. The point with the largest pressure change rate is selected as the core abnormal point. When the distance from the core abnormal point to the nearest actual sensor exceeds the coverage threshold, it is marked as a collection blind zone. The collection omission degree assessment value is calculated by weighting the proportion of core abnormal points in the blind zone with the proportion of pipeline length.

[0015] Furthermore, based on the degree of data collection omission, additional monitoring points are generated within the data collection blind zone to form a pressure monitoring coverage area covering the pressure drop abrupt change section, including:

[0016] Coordinates of additional monitoring points are generated at the geometric center of the blind zone and at the boundary between the blind zone and the non-blind zone. The coordinates of the additional monitoring points are merged with the original sensor location coordinates to generate a complete monitoring network, forming a pressure monitoring coverage area that covers the section with sudden pressure drop.

[0017] Furthermore, the fluid resistance analysis of the pressure monitoring coverage area, by identifying abnormal pressure drop points at non-elbow locations and combining them with fluid velocity, reverse-engineers the roughness coefficient and effective flow diameter of the pipe inner wall, determines the coordinates of the scaling location on the pipe inner wall, and quantifies the thickness of the scaling layer, including:

[0018] The frictional resistance coefficient of each pipe segment was calculated using the Darcy-Wiesbach equation. Pipe segments exceeding the standard value for clean pipelines and excluding elbow locations were selected as a set of abnormal pressure drop points in non-elbow locations. The actual roughness coefficient was solved in reverse using the Körbruck equation based on the pressure gradient value of the abnormal pressure drop points and the fluid velocity, and the roughness increment was calculated. For pipe segments with roughness increments exceeding the threshold, the effective flow diameter was calculated using Reynolds number correction. The scale layer thickness was obtained by subtracting the effective flow diameter from the nominal diameter and dividing by two. The start and end coordinates of continuous sections where the scale layer thickness exceeded the scale threshold were identified as the scale location coordinates on the inner wall of the pipeline.

[0019] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects:

[0020] This invention discloses a method for in-depth data mining and evaluation in the industrial internet. By real-time acquisition of pipeline pressure sensor values, sensor spacing, and elbow distribution locations, a spatial proportional relationship model between sensors and elbows is established. Addressing the issue of pressure drop abrupt changes at pipeline elbows affecting the accuracy of scale detection, this invention employs pressure attenuation trend fitting and completion technology to generate virtual pressure monitoring points at elbow locations, achieving continuous pressure monitoring coverage. Through full-domain scanning, local pressure drop abrupt change sections are identified, and the coordinates of these abrupt change areas are spatially compared with elbow locations to distinguish between elbow-induced pressure drop and abnormal pressure drop, identify blind spots in data acquisition, and optimize the layout of monitoring points. Finally, through fluid resistance analysis, combined with abnormal pressure drop points and fluid velocity data at non-elbow locations, the pipe inner wall roughness coefficient and effective flow diameter are inversely deduced to accurately locate the coordinates of scale formation on the pipe inner wall and quantify the scale layer thickness, achieving intelligent monitoring and early warning of pipeline health status. Attached Figure Description

[0021] Figure 1 This is a flowchart of an industrial internet data deep mining and evaluation method according to the present invention.

[0022] Figure 2 This is a schematic diagram of an industrial internet data deep mining and evaluation method according to the present invention.

[0023] Figure 3 This is another schematic diagram of an industrial internet data deep mining and evaluation method according to the present invention. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0025] like Figure 1-3 This embodiment of an industrial internet data deep mining and evaluation method may specifically include:

[0026] Step S101: The monitoring values ​​of the pipeline pressure sensor, the installation spacing between adjacent sensors, and the distribution location of pipeline bends are collected in real time through the industrial internet platform. The pressure status of each monitoring point is identified, and the spatial ratio between the sensor installation spacing and the bend distribution spacing is calculated.

[0027] The system acquires real-time monitoring values ​​and sensor installation coordinates from pipeline pressure sensors in an industrial internet platform. The pipe segment length between adjacent sensors is calculated using coordinate differences. Simultaneously, elbow coordinates are extracted from pre-stored pipeline topology data on the platform. After arranging the elbow coordinates sequentially along the pipeline route, the distribution spacing between adjacent elbows is calculated. A spatial ratio value is calculated for the pipe segment length and the distribution spacing. When the ratio value exceeds a preset sparse threshold, the pipe segment is marked as a sparse acquisition segment; when the ratio value is below a preset dense threshold, it is marked as a dense acquisition segment. Based on the marking results of each pipe segment and the spatial ratio value, a spatial distribution ratio map of sensors and elbows is constructed.

[0028] Specifically, in one implementation, the industrial internet platform accesses the real-time data stream of the pipeline pressure sensor via the OPCUA protocol. Each sensor node carries a unique identifier and three-dimensional coordinate information. The OPCUA protocol refers to an open platform communication unified architecture, a standard protocol used for secure data exchange between industrial devices. Based on the X, Y, and Z coordinate values ​​of the sensor, the platform calculates the straight-line distance between adjacent sensors using the Euclidean distance formula, and then combines this with a pipeline centerline path correction coefficient to obtain the actual pipe segment length.

[0029] Specifically, pipeline topology data is stored in the platform database in the form of a graph, with elbow nodes containing angle attributes and connection relationships. The system traverses the topology graph along the pipeline route, extracts the spatial coordinate sequence of elbow nodes, and obtains the distribution spacing by calculating the difference between the coordinates of adjacent elbows.

[0030] For example, the spatial ratio value is obtained by dividing the pipe segment length by the average bend spacing within the corresponding interval. When a pipe segment contains 3 bends with a total spacing of 12 meters, and the sensor spacing is 30 meters, the ratio value is 5, exceeding the preset sparsity threshold of 2.0, and the system automatically marks it as a sparse acquisition segment. The spatial distribution ratio relationship map is presented in the form of a heat map, with different colored blocks representing different acquisition density levels.

[0031] Step S102: Based on the pressure status of each monitoring point and the spatial ratio between the sensor and the elbow distribution distance, the pressure attenuation trend between adjacent sensors is fitted and completed to identify virtual pressure monitoring points at the elbow distribution location and determine the continuous pressure monitoring coverage area.

[0032] Based on the real-time pressure values ​​at each monitoring point and the spatial ratio between the sensors and the elbow distribution, cubic spline interpolation is used to fit the pressure values ​​between adjacent sensors, generating a pressure attenuation trend curve. The pressure change rate at each sampling point on the trend curve is calculated. By sampling at the elbow coordinates of the pressure attenuation trend curve, the pressure value and pressure change rate at that location are obtained. If the pressure change rate exceeds a preset abrupt change threshold, a virtual pressure monitoring point is created at that elbow location, and the pressure value obtained from the curve sampling is assigned to it. A monitoring point set is constructed based on the virtual pressure monitoring points and the actual sensor monitoring points. By determining whether the distance between adjacent monitoring points is less than a preset continuous coverage threshold, the boundary of the area with continuous monitoring capability is identified, and the continuous pressure monitoring coverage range is determined.

[0033] Specifically, in one implementation, cubic spline interpolation fits the pressure values ​​between adjacent sensors by constructing a piecewise cubic polynomial function.

[0034] Specifically, for any two adjacent sensor monitoring points on the pipeline, the system extracts their pressure values ​​as boundary conditions. Simultaneously, based on the continuity principle in fluid mechanics, the system ensures that the first and second derivatives of the interpolation function are continuous at the nodes, thereby obtaining a smooth pressure attenuation trend curve. This trend curve reflects the pressure loss pattern of the fluid within the pipe section caused by frictional and local resistance.

[0035] It should be noted that the pressure change rate is calculated by differentiating the interpolation function. At the elbow coordinate location, the system calculates the first derivative value at that point, which is the rate of change of pressure with respect to the pipe length. When fluid flows through the elbow, the local resistance caused by the change in flow direction leads to a sharp drop in pressure, manifested as a significant increase in the absolute value of the pressure change rate.

[0036] Preferably, the preset mutation threshold is obtained based on historical data statistics under normal pipeline operating conditions.

[0037] For example, the pressure change rate is typically 0.5 kPa per meter in straight pipe sections, but can reach over 3 kPa per meter at elbows. When the pressure change rate exceeds the threshold, the system creates a virtual pressure monitoring point at the elbow location and samples the estimated pressure value at that point from the interpolation curve, thus achieving a digital reconstruction of the pressure state at the elbow.

[0038] For example, the monitoring point set includes the spatial coordinates and pressure values ​​of actual sensors and virtual monitoring points. The system iterates through all monitoring points in the set and calculates the Euclidean distance between each pair of adjacent points. A preset continuous coverage threshold is typically set to 10 times the pipe diameter; when the distance between adjacent monitoring points is less than this threshold, the pipe section between the two points is considered to have continuous monitoring capability.

[0039] In one possible implementation, a depth-first search algorithm is used to identify connected regions consisting of all monitoring points that meet distance criteria. Starting from any monitoring point, neighboring points with a distance less than a threshold are recursively searched until no further expansion is possible, forming a continuous monitoring region. The continuous pressure monitoring coverage area is the union of all connected regions, achieving a complete digital representation of the pipeline pressure field.

[0040] Step S103: Perform a full-area scan of the continuous pressure monitoring coverage area to identify local pressure drop abrupt change sections, and locate the start and end coordinates of potential abrupt change areas by evaluating the pressure change slope at each point within the section.

[0041] A sliding window scan is performed on all monitoring points within the continuous pressure monitoring coverage area. The window contains a preset number of adjacent monitoring points. The sum of squares of the deviations of the pressure values ​​of each monitoring point within the window from the window mean is calculated. If the sum of squares exceeds a preset fluctuation threshold, the pipe segment covered by the window is marked as a local pressure drop mutation zone. For each monitoring point pressure value and corresponding location coordinate within the local pressure drop mutation zone, the least squares method is used to fit the linear regression coefficient of pressure change with location as the pressure change slope of that point. By comparing the slope values ​​of adjacent monitoring points, when the slope change exceeds a preset change threshold, the location is identified as a turning point. Based on the location of the turning point, monitoring points with slope values ​​less than a preset stability threshold are searched forward along the pipeline direction as the mutation start point, and monitoring points with slope values ​​recovering to below the stability threshold are searched backward as the mutation end point. The coordinates of the mutation start point and mutation end point are obtained to determine the start and end coordinates of the potential mutation area.

[0042] Specifically, in one implementation, the sliding window scanning mechanism is achieved by moving a set of monitoring points of a fixed size along the pipe axis.

[0043] Specifically, the window contains seven consecutive monitoring points, moving one point at a time to form an overlapping scan. Within the window, the arithmetic mean μ of the pressure values ​​at all monitoring points is first calculated. Then, the deviation of each point from the mean is calculated. All deviations are squared and summed to obtain the sum of squared deviations SSE = ∑(xi-μ)^2, where xi is the pressure value at each monitoring point. When the sum of squared deviations exceeds a fluctuation threshold calculated based on historical operational data, it indicates that the pressure distribution within that section is uneven, and there are local abrupt changes.

[0044] It should be noted that the principle of applying the least squares method in pipeline pressure scenarios is to obtain the linear fitting parameters of the pressure-position relationship by minimizing the sum of squared residuals. For each monitoring point within the local pressure drop abrupt change zone, the system extracts the pressure values ​​and position coordinates of three monitoring points before and after that point, constructing seven data pairs. By solving the system of equations, the local linear regression coefficient at that point is obtained, and this coefficient is the pressure change slope. The slope reflects the rate of pressure attenuation along the pipeline axis. Under normal circumstances, the slope value is relatively stable, but abrupt changes occur at locations of scaling or blockage.

[0045] Preferably, the inflection point is identified using a differential comparison method. The system calculates the slope difference between two adjacent monitoring points. When the absolute value of the difference exceeds a preset change threshold, it is considered that there is a change in the pressure change pattern between these two points. The change threshold is typically set to 50% of the normal slope value, thereby avoiding misjudging normal fluctuations as inflection points.

[0046] For example, the search process for the start and end points of a mutation employs a bidirectional expansion algorithm. Starting from the identified inflection point, the system checks the slope value point by point along the upstream direction of the pipeline. When the slope values ​​of three consecutive monitoring points are all less than the stability threshold, the first point that meets the condition is taken as the mutation start point. Similarly, the search continues along the downstream direction, and when the slope value recovers to below the stability threshold and remains stable, the mutation end point is determined.

[0047] In one possible implementation, the stabilization threshold is dynamically adjusted based on the average pressure drop gradient of the straight pipe section, typically set at a pressure drop rate of 0.8 kPa per meter. By recording the three-dimensional coordinates of the start and end points of abrupt changes, the system accurately locates potential areas of abnormal pressure drop, providing spatial location data for pipeline maintenance decisions.

[0048] Step S104: Spatial comparison of the start and end coordinates of the potential mutation area with the distribution of pipe bends is performed to identify overlapping areas and mark them as bend-induced pressure drop mutation areas. Based on the bend-induced pressure drop mutation areas and local pressure drop mutation segments, the set of monitoring segments for bend-related pressure drop mutations is determined.

[0049] Obtain the start and end coordinates of the potential abrupt change region and the center coordinates of the pipe bend. Calculate the minimum distance from the boundary point of the abrupt change region to the center of the bend. If this distance is less than a preset spatial overlap threshold, spatial overlap is determined, and the overlapping region is marked as a bend-induced pressure drop abrupt change region. Based on the location distribution of the bend-induced pressure drop abrupt change region within the local pressure drop abrupt change segment, extract continuous monitoring data sequences covering the bend-induced pressure drop abrupt change region as monitoring segments. Summarize all monitoring segments to form a monitoring segment set for bend-related pressure drop abrupt changes.

[0050] Specifically, in one implementation, the determination of spatial overlap is based on the pipeline's three-dimensional coordinate system.

[0051] Specifically, the system calculates the Euclidean distances from the start and end points of the potential abrupt change region to the center point of the elbow, and takes the smaller of the two distance values ​​as the minimum distance. The spatial overlap threshold is usually set to twice the pipe diameter. When the minimum distance is less than this threshold, it is considered that there is a spatial correlation between the pressure drop abrupt change and the elbow location.

[0052] It should be noted that the marking of the elbow-induced pressure drop abrupt change area is done by attribute assignment. The system adds an "elbow-induced" identifier to the area that meets the overlap condition in the pipeline digital model and records the associated elbow number and overlap value.

[0053] For example, the extraction of monitoring data sequences is performed in timestamp order. The system filters all monitoring point data located within the elbow-induced pressure drop abrupt change area from the monitoring data of the local pressure drop abrupt change section, including pressure values, acquisition time, and spatial coordinates, forming a continuous data sequence. Each data sequence serves as an independent monitoring segment, and multiple segments are aggregated to form a set of monitoring segments related to elbow-related pressure drop abrupt changes, achieving accurate identification of the elbow-affected area.

[0054] Step S105: Identify the core anomaly points in the monitoring segment set of elbow-related pressure drop mutations. By comparing the coverage overlap between the coordinates of the core anomaly points and the actual sensor installation location, identify the acquisition blind spots and assess the degree of acquisition omission based on the acquisition blind spots.

[0055] For each segment in the monitoring segment set of elbow-related pressure drop abrupt changes, pressure gradient calculation is performed. The isolated forest algorithm is used to identify data points whose pressure change rate exceeds a preset anomaly threshold. The spatial coordinates of these anomaly points are obtained, and the point with the largest pressure change rate is selected as the core anomaly point. Based on the coordinates of the core anomaly point, the distance from the core anomaly point to the nearest actual sensor is calculated. If the distance exceeds a preset threshold for the sensor's coverage area, the area where the core anomaly point is located is marked as a blind zone. The number of core anomaly points within the blind zone is counted. The omission rate is calculated as a percentage, and a weighted assessment value for the degree of omission is obtained based on the proportion of the pipe length in the blind zone to the total pipe length corresponding to the monitoring segment.

[0056] Specifically, in one implementation, the isolated forest algorithm identifies stress anomalies by constructing multiple isolated trees.

[0057] Specifically, the algorithm randomly samples the pressure gradient data in the monitored segments and constructs a binary tree structure through recursive segmentation. Abnormal data points, deviating from the normal distribution, have shorter path lengths within the tree. When the average path length of a data point is less than a preset anomaly detection threshold, that point is marked as an anomaly. The system further calculates the absolute value of the pressure change rate for each anomaly point and selects the top 10% of anomalies by change rate as core anomalies.

[0058] It should be noted that the preset threshold for sensor coverage is determined based on the sensor's physical characteristics and the pipeline environment. In typical industrial pipelines, the effective monitoring distance of a pressure sensor is usually 15 to 20 times the pipe diameter. When the Euclidean distance from the core anomaly point to the nearest sensor exceeds this threshold, it indicates that the pressure anomaly at that location cannot be accurately captured by existing sensors, creating a blind spot in the data acquisition.

[0059] Preferably, the assessment of the degree of data acquisition omission employs a dual-indicator weighted method. The omission rate reflects the proportion of core anomalies that were not detected, while the pipe length ratio reflects the spatial coverage of the blind zone. The system assigns a weight coefficient of 0.6 to the omission rate and a weight coefficient of 0.4 to the pipe length ratio, and the weighted sum of the two yields a comprehensive assessment value for the degree of data acquisition omission.

[0060] For example, if a pipe segment has 8 core anomalies, 3 of which are located in the blind zone, the omission rate is 37.5%. Meanwhile, the pipe length corresponding to the blind zone is 12 meters, and the total length of the monitored segment is 40 meters, representing 30% of the total length. The comprehensive evaluation value is calculated as 0.6 × 37.5% + 0.4 × 30% = 34.5%, indicating a moderate degree of monitoring omission in this pipe segment.

[0061] In one possible implementation, the system automatically classifies the omission rate based on an assessment value. An assessment value below 20% indicates mild omission, 20%-50% indicates moderate omission, and above 50% indicates severe omission. Different omission levels correspond to different sensor replenishment deployment schemes, enabling dynamic optimization of the monitoring network.

[0062] Step S106: Based on the degree of data loss, add monitoring points in the data loss blind zone to obtain the pressure monitoring coverage range covering the pressure drop abrupt change section.

[0063] Based on the assessment value of the degree of data loss, if the assessment value exceeds a preset threshold for adding monitoring points, the coordinates of additional monitoring points are calculated at the geometric center of the blind zone, and boundary monitoring points are set at the boundary between the blind zone and the non-blind zone to obtain the coordinate set of the additional points. The coordinate set of the additional points is then spatially merged with the original sensor location coordinates to generate a complete monitoring network including both the original and additional points, thus determining the pressure monitoring coverage area for the pressure drop abrupt change section.

[0064] Specifically, and exemplaryly, in one implementation, the threshold for adding monitoring points is dynamically set based on the safety level of the pipeline operation. For high-pressure pipelines, the threshold is typically set at 15% to ensure rapid response in high-risk environments; for medium- and low-pressure pipelines, the threshold can be relaxed to 25% to balance monitoring costs and safety requirements. When the assessment value of the degree of data omission exceeds the corresponding threshold, the system triggers the monitoring point addition process. The geometric center of the blind zone is determined by calculating the arithmetic mean of all coordinate points within the blind zone, and is identified as the first monitoring point to be added.

[0065] It should be noted that the boundary monitoring points are set according to the principle of equidistant distribution. The system identifies the boundary line of the blind zone and evenly sets monitoring points on the boundary according to the preset boundary point spacing. The boundary point spacing is usually 8 times the pipe diameter, thus forming a ring-shaped monitoring zone around the blind zone.

[0066] For example, during the spatial merging process, the system checks the distance relationship between the newly added points and the existing sensors. If the distance between the newly added point and an existing sensor is less than a minimum spacing threshold (set to 10 meters), the newly added point is cancelled to avoid monitoring redundancy. The merged complete monitoring network includes three types of nodes: existing sensors, blind zone center monitoring points, and boundary monitoring points. Together, they constitute a pressure monitoring coverage area covering sections with sudden pressure drop, achieving comprehensive perception of the pipeline pressure field.

[0067] Step S107: Perform fluid resistance analysis on the pressure monitoring coverage area. By identifying abnormal pressure drop points at non-elbow locations and combining them with fluid velocity, reverse-engineer the roughness coefficient and effective flow diameter of the pipe inner wall to determine the coordinates of the scaling location on the pipe inner wall and quantify the thickness of the scaling layer.

[0068] Fluid resistance calculations are performed on the monitoring data within the pressure monitoring coverage area. The frictional resistance coefficient of each pipe section is obtained using the Darcy-Wiesbach equation. Pipe sections with resistance coefficients exceeding the standard value for clean pipelines are screened out, and data points at bends are excluded, resulting in a set of abnormal pressure drop points outside bends. Based on the pressure gradient value and corresponding fluid velocity at each point in the abnormal pressure drop point set, the actual roughness coefficient of the pipe inner wall is solved inversely using the Colbrook equation. The roughness increment is determined by the ratio of the actual roughness to the new standard roughness of the pipeline. For pipe sections where the roughness increment exceeds a preset threshold, the effective flow diameter is calculated using Reynolds number correction based on the relationship between flow velocity change and pressure drop. The scale layer thickness is obtained by subtracting the effective flow diameter from the nominal pipe diameter and then dividing by two. Using the distribution data of the scale layer thickness along the pipe axis, continuous sections with thicknesses exceeding the preset scale threshold are identified. The start and end coordinates of these sections are extracted as the scale location coordinates on the pipe inner wall, and the quantitative value of the scale layer thickness at the corresponding location is output.

[0069] Specifically, and exemplaryly, in one embodiment, the frictional resistance coefficient calculation method employs the Darcy-Wiesbach equation to perform fluid resistance analysis. This method calculates the pressure loss value ΔP of each pipe segment within the pressure monitoring coverage area using the pressure difference between adjacent monitoring points, obtains pipe geometric parameters including pipe segment length L and pipe inner diameter D, and fluid physical property parameters including fluid density ρ and average flow velocity v. According to the Darcy-Wiesbach equation, ΔP = f × (L / D) × (ρv) 2 / 2), where f is the frictional resistance coefficient (parameter to be determined), substitute the known parameters into the equation; then, solve the frictional resistance coefficient f=△P / (L / D×ρv) by transforming the equation. 2 / 2); Finally, the calculated friction resistance coefficient is compared with the standard friction coefficient of clean pipelines, which is typically in the range of 0.015 to 0.025. When the calculated friction coefficient exceeds 0.03, it is determined that the pipe section has abnormal resistance caused by scaling. The output of this method is the friction resistance coefficient value of each pipe section, as well as a list of identified abnormal resistance pipe sections. For the identified abnormal resistance pipe sections, the system further uses the Colbrook equation and the effective flow diameter estimation method for detailed analysis to quantify the degree of scaling.

[0070] It should be noted that the screening process for abnormal pressure drop points at non-elbow locations is achieved through coordinate comparison. The system marks monitoring points within a 2-meter radius before and after the elbow as the elbow influence zone, and removes data points within these areas from the abnormal pressure drop point set. The remaining abnormal pressure drop points are the pressure anomalies caused by changes in the inner wall condition on the straight pipe section, and these points become the focus of subsequent scaling analysis. Preferably, the roughness coefficient is accurately estimated using the Körbruck equation and the Newton-Raphson iteration method. This method first calculates the friction coefficient f using the Darcy-Wiesbach equation; obtains the fluid velocity v and pipe inner diameter D for calculating the Reynolds number; and calculates the fluid kinematic viscosity ν. In the processing, the Reynolds number Re = vD / ν is first calculated to reflect the fluid flow state; secondly, according to the Körbruck equation 1 / √f = -2.0 × log 10 (ε / (3.7D)+2.51 / (Re√f)), where ε is the absolute roughness (the parameter to be determined), this equation establishes an implicit relationship between the friction coefficient, Reynolds number, and roughness; next, the equation is transformed into a form for solving ε. Since the equation is a transcendental equation and cannot be solved directly, the Newton-Raphson iteration method is used to solve it. The iteration formula is ε_{n+1}=ε_n-F(ε_n) / F'(ε_n), where F(ε) is the error function of the Colbrook equation, and F'(ε) is its derivative; then, the initial roughness estimate ε0 is set (usually the standard roughness of a clean pipe is 0.045 mm), and it is substituted into the iteration formula for repeated calculation until the roughness difference between two adjacent iterations is less than the convergence threshold (usually 0.001 mm) or the maximum number of iterations is reached (usually 20 times); finally, the converged absolute roughness value ε is obtained, and the relative roughness ε / D and the roughness increment Δε=ε-ε0 are calculated. The output yields the actual roughness coefficient ε and roughness increment Δε of the pipe's inner wall, along with their growth rate relative to the standard roughness. The standard roughness of a new pipe is typically 0.045 mm. When the calculated actual roughness exceeds 0.2 mm, the roughness increment exceeds 300%, indicating that a significant scale layer has formed on the pipe's inner wall, and the surface roughness has increased significantly.

[0071] For example, the calculation of the effective flow diameter takes into account both fluid continuity and momentum conservation. In fouled pipe sections, although the outer diameter of the pipe remains constant, the effective flow area inside decreases due to the presence of the fouling layer. The system inversely calculates the reduction in cross-sectional area by monitoring local increases in flow velocity. When fluid enters a fouled pipe section from a clean section, according to mass conservation, the flow velocity is inversely proportional to the cross-sectional area. If the flow velocity increases from 2 m / s to 2.5 m / s, it indicates a 20% reduction in the effective flow area, from which the reduction in the effective flow diameter can be calculated.

[0072] In one possible implementation, the Reynolds number correction calculation considers the impact of scale surface irregularities on the flow regime. Scale typically has a rough, porous surface, altering the flow characteristics in the near-wall region. The system considers the influence of scale surface roughness on the flow regime in the effective flow diameter estimation. When the roughness increment is large, the irregularity of the pipe inner wall surface increases, leading to increased frictional resistance. By comprehensively considering the roughness coefficient and flow velocity changes, the system calculates a corrected Reynolds number to determine whether a flow regime transition has occurred. When the Reynolds number is below 2300, the fluid may transition from turbulent to laminar or transitional flow regimes. This flow regime transition will cause changes in pressure drop characteristics, further affecting the accuracy of the effective flow diameter estimation.

[0073] Specifically, the calculation of scale thickness is derived using geometric relationships. The difference between the nominal pipe diameter D0 and the effective flow diameter De represents twice the scale thickness; therefore, the scale thickness t on one side is equal to the difference between D0 and De divided by 2.

[0074] For example, for a pipe with a nominal diameter of 100 mm, if the effective flow diameter is reduced to 94 mm, the average scale thickness will be 3 mm. The system performs this calculation for each monitoring point, generating spatial distribution data of the scale thickness. Furthermore, the method for determining the scale location coordinates is based on the continuity of the thickness distribution. The system scans the axial distribution data of the scale thickness; when the scale thickness at three or more consecutive monitoring points exceeds a preset scale threshold (set based on pipe material tolerance and empirical data, typically 1.5 mm), the pipe section covered by these points is identified as a scaled area. The system extracts the three-dimensional coordinates of the starting and ending points of this area to precisely locate the specific location where the scale occurs.

[0075] Understandably, the quantitative output of scale thickness employs a comprehensive description using multiple indicators. The system calculates and outputs the average thickness value of the scaled area, reflecting the overall degree of scale buildup in that area; it also calculates the maximum thickness value to assess the location of the most severe scale buildup; furthermore, the system calculates the standard deviation of the thickness to assess the uniformity of the scale buildup—a large standard deviation indicates uneven scale buildup, while a small standard deviation indicates relatively uniform scale buildup. These quantitative indicators provide comprehensive data support for pipeline maintenance decisions, achieving a complete technical closed loop from pressure monitoring to scale diagnosis.

[0076] The above description is merely a preferred embodiment of this application and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in this application is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the concept of this application. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions disclosed in this application.

Claims

1. A method for in-depth data mining and evaluation in the industrial internet, characterized in that, The method includes: The system collects monitoring values ​​from pipeline pressure sensors, the installation spacing between adjacent sensors, and the coordinates of pipeline bend locations through an industrial internet platform. It identifies the pressure state at each monitoring point and calculates the spatial ratio between the sensor installation spacing and the bend distribution spacing. Based on the pressure state and the spatial ratio, it fits and completes the pressure attenuation trend between adjacent sensors, generating virtual pressure monitoring points at the bend locations to determine the continuous pressure monitoring coverage area. A full-area scan of the continuous pressure monitoring coverage area identifies local pressure drop abrupt change zones, calculates the pressure change slope at each point within these zones, and locates the start and end coordinates of potential abrupt change zones. Finally, it performs a spatial correlation between the start and end coordinates of these potential abrupt change zones and the pipeline bend distribution locations. Inter-regional comparison is performed to identify overlapping areas and mark them as elbow-induced pressure drop abrupt change areas. Monitoring segments containing these elbow-induced pressure drop abrupt change areas are extracted to form a monitoring segment set. Core anomaly points are identified in the monitoring segment set, and the overlap between the coordinates of the core anomaly points and the actual sensor installation locations is compared to identify blind spots and quantify the degree of data acquisition omission. Based on the degree of data acquisition omission, additional monitoring points are generated within the blind spots to form a pressure monitoring coverage area covering the pressure drop abrupt change area. Fluid resistance analysis is performed on the pressure monitoring coverage area to identify abnormal pressure drop points at non-elbow locations. By combining fluid flow velocity with reverse deduction of the pipe inner wall roughness coefficient and effective flow diameter, the coordinates of the scaling location on the pipe inner wall are determined and the scaling layer thickness is quantified.

2. The industrial internet data deep mining and evaluation method according to claim 1, characterized in that, The process of collecting pipeline pressure sensor monitoring values, adjacent sensor installation spacing, and pipeline bend location coordinates through an industrial internet platform, identifying the pressure status of each monitoring point, and calculating the spatial ratio between sensor installation spacing and bend distribution spacing includes: The system acquires real-time monitoring values ​​and installation location coordinates of pipeline pressure sensors, calculates the pipe segment length between adjacent sensors, extracts elbow location coordinates from pipeline topology data and calculates the distribution spacing between adjacent elbows, calculates the spatial ratio value based on the pipe segment length and the distribution spacing between adjacent elbows, marks pipe segments that exceed the sparse threshold or are below the dense threshold, and constructs a spatial distribution ratio map of sensors and elbows.

3. The industrial internet data deep mining and evaluation method according to claim 1, characterized in that, The process of performing a full-area scan of the continuous pressure monitoring coverage area to identify local pressure drop abrupt change zones, and locating the start and end coordinates of potential abrupt change zones by evaluating the pressure change slope at each point within the zone, includes: A sliding window scan is performed on the monitoring points within the continuous pressure monitoring coverage area. The window includes adjacent monitoring points. The sum of squares of the deviations of the pressure values ​​within the window from the mean is calculated. Pipe sections where the sum of squares of deviations exceeds the fluctuation threshold are marked as local pressure drop abrupt change sections. The pressure values ​​and location coordinates of the monitoring points within the local pressure drop abrupt change sections are fitted with linear regression coefficients using the least squares method as the pressure change slope. The slope values ​​of adjacent monitoring points are compared, and the locations where the slope changes exceed the change threshold are identified as inflection points. Monitoring points with slopes less than the stability threshold are searched forward along the pipeline direction as the abrupt change starting point, and monitoring points with slopes recovering to below the stability threshold are searched backward as the abrupt change ending point. The coordinates of the abrupt change starting point and the abrupt change ending point are obtained.

4. The industrial internet data deep mining and evaluation method according to claim 1, characterized in that, Based on the pressure status of each monitoring point and the spatial ratio between the sensor and the elbow distribution spacing, the pressure attenuation trend between adjacent sensors is fitted and completed to identify virtual pressure monitoring points at the elbow distribution location and determine the continuous pressure monitoring coverage area, including: A pressure decay trend curve is generated by fitting the pressure values ​​between adjacent sensors using cubic spline interpolation. The pressure value and pressure change rate are obtained by numerical sampling on the pressure decay trend curve. When the pressure change rate exceeds the sudden change threshold, a virtual pressure monitoring point is created. The virtual pressure monitoring point and the actual sensor monitoring point are combined to form a monitoring point set. The area where the distance between adjacent monitoring points is less than the continuous coverage threshold is determined as the continuous pressure monitoring coverage area.

5. The industrial internet data deep mining and evaluation method according to claim 1, characterized in that, The start and end coordinates of the potential abrupt change region are spatially compared with the distribution locations of pipe bends. Overlapping areas are identified and marked as bend-induced pressure drop abrupt change regions. Monitoring segments containing these regions are extracted to form a monitoring segment set, including: Calculate the minimum distance from the boundary point of the abrupt change region to the center of the bend. If the distance is less than the spatial overlap threshold, mark it as a bend-induced pressure drop abrupt change region. Extract continuous monitoring data sequences covering the bend-induced pressure drop abrupt change region to form monitoring segments, and summarize them to form a monitoring segment set.

6. The industrial internet data deep mining and evaluation method according to claim 1, characterized in that, For the monitoring segment set, identify core anomalies, compare the overlap between the coordinates of the core anomalies and the actual sensor installation locations, identify blind spots in data acquisition, and quantify the degree of data acquisition omissions, including: Pressure gradients are calculated for the monitoring segment set. The isolated forest algorithm is used to identify data points whose pressure change rate exceeds the abnormal threshold. The point with the largest pressure change rate is selected as the core abnormal point. When the distance from the core abnormal point to the nearest actual sensor exceeds the coverage threshold, it is marked as a collection blind zone. The collection omission degree assessment value is calculated by weighting the proportion of core abnormal points in the blind zone with the proportion of pipeline length.

7. The industrial internet data deep mining and evaluation method according to claim 1, characterized in that, Based on the degree of data loss, additional monitoring points are generated within the data loss blind zone to form a pressure monitoring coverage area covering the section of sudden pressure drop, including: Coordinates of additional monitoring points are generated at the geometric center of the blind zone and at the boundary between the blind zone and the non-blind zone. The coordinates of the additional monitoring points are merged with the original sensor location coordinates to generate a complete monitoring network, forming a pressure monitoring coverage area that covers the section with sudden pressure drop.

8. The industrial internet data deep mining and evaluation method according to claim 1, characterized in that, The fluid resistance analysis of the pressure monitoring coverage area involves identifying abnormal pressure drop points at non-elbow locations and combining this with fluid velocity to inversely deduce the roughness coefficient and effective flow diameter of the pipe inner wall, determine the coordinates of the scale location on the pipe inner wall, and quantify the thickness of the scale layer. This includes: The frictional resistance coefficient of each pipe segment was calculated using the Darcy-Wiesbach equation. Pipe segments exceeding the standard value for clean pipelines and excluding elbow locations were selected as a set of abnormal pressure drop points in non-elbow locations. The actual roughness coefficient was solved in reverse using the Körbruck equation based on the pressure gradient value of the abnormal pressure drop points and the fluid velocity, and the roughness increment was calculated. For pipe segments with roughness increments exceeding the threshold, the effective flow diameter was calculated using Reynolds number correction. The scale layer thickness was obtained by subtracting the effective flow diameter from the nominal diameter and dividing by two. The start and end coordinates of continuous sections where the scale layer thickness exceeded the scale threshold were identified as the scale location coordinates on the inner wall of the pipeline.

Images