A sensor data acquisition and processing system for digital engineering

Abstract

The application relates to the technical field of engineering monitoring data processing, and discloses a sensor data acquisition and processing system for digital engineering, wherein the system comprises an excess seepage event identification module, a conventional baseline acquisition module, an expected response envelope generation module, a two-dimension abnormal index calculation module, a fault discrimination module, a data fusion correction module and a closed-loop updating module. The system identifies the excess seepage event by comparing the instantaneous external water load intensity with the matrix saturation permeability coefficient, generates the expected response envelope of each depth layer by using an exponential decay depth distribution function, performs two-dimension abnormal discrimination on the measured response, distinguishes the real preferential flow breakthrough event from the sensor fault by combining the cross-correlation shape similarity coefficient, and performs weighted fusion correction on the fault node to output the multi-depth monitoring data correction sequence.

Description

Technical Field

[0001] The embodiments disclosed herein relate to the field of engineering monitoring data processing technology, and in particular to a sensor data acquisition and processing method, system, electronic device, and storage medium for digital engineering. Background Technology

[0002] With the continuous development of digital engineering technology, the large-scale deployment of multi-depth sensor arrays in complex engineering scenarios such as deep foundation pits, tunnels, and underground engineering has become an important means of engineering safety monitoring. Real-time perception of the seepage status of the retaining structure is of fundamental significance for engineering safety early warning. Various types of sensors, such as pore water pressure sensors, are widely deployed at different depths to obtain spatiotemporal distribution information of the seepage field.

[0003] However, with the continuous increase in the number of monitoring sections and sensor density, data acquisition and processing at engineering sites face significant technical bottlenecks. Under heavy rainfall conditions, the preferential flow phenomenon in the retaining structure causes the deep sensor response to exhibit time-delay and amplitude characteristics that are drastically different from those of conventional matrix seepage. Baseline models established under conventional seepage conditions are insufficient to accurately describe the sensor response patterns under preferential flow conditions. Simultaneously, during periods of excessive infiltration, sensors are affected by complex hydrological environments, significantly increasing the probability of physical failures such as probe contact degradation and range saturation drift. The abnormal response characteristics overlap with the rapid response characteristics of genuine preferential flow intrusion events in both time-delay and amplitude dimensions, posing considerable challenges to anomaly identification and data quality control.

[0004] Existing engineering sensor data processing schemes typically detect sensor anomalies based on single threshold discrimination or statistical control chart methods. While these schemes are applicable to certain conditions under normal operating conditions, they have significant limitations under conditions of preferential flow with excessive permeability. Firstly, the baseline models used in existing schemes do not incorporate the fissure ratio parameter of the retaining structure and the excess permeability ratio into the modeling process of the expected response, leading to a systematic deviation between the expected response envelope and the actual response under preferential flow conditions. Secondly, existing schemes lack quantitative characterization methods for the waveform morphology of sensor responses, failing to effectively distinguish between the rapid response caused by preferential flow intrusion and the drift or abrupt changes caused by sensor physical faults. This results in fault node data being misused for downstream engineering analysis or normal rapid response data being incorrectly discarded. The inventors of this disclosure note that current sensor data processing schemes do not consider the differentiated impact of fissure transmission characteristics on the response at different depths under conditions of preferential flow with excessive permeability. This leads to the inability to effectively resolve the aliasing problem between sensor faults and the actual preferential flow response during heavy rainfall, thereby affecting the integrity and reliability of multi-depth monitoring data. Summary of the Invention

[0005] The purpose of this invention is to provide a sensor data acquisition and processing system for digital engineering, which solves the problems in the prior art that cannot effectively distinguish between real preferential flow penetration events and sensor failures under superpermeable preferential flow conditions, and that fault nodes lack physical consistency data repair methods.

[0006] This invention discloses a sensor data acquisition and processing method for digital engineering, including the following steps: determining the over-permeability event identifier and calculating the priority flow activation intensity index based on the comparison results between the instantaneous external water load intensity and the saturated permeability coefficient of the retaining structure matrix; The seepage time delay characteristic parameters between layers are generated by sliding window cross-correlation calculation based on the historical data sequence of multiple depth sensors at the same monitoring section; Based on the priority flow activation intensity index and the fracture rate parameter, the expected shortening time delay value and the expected response amplitude envelope of each depth layer are generated by the exponential decay depth allocation function. During an over-permeability event, a two-dimensional anomaly index is generated based on the deviation between the measured response of the deep sensor and the expected shortened time delay value and the expected response amplitude envelope. For deep sensors whose dual-dimensional abnormal indicators exceed the allowable range, the cross-correlation shape similarity coefficient between the deep sensor and the shallow sensor response waveform is used to determine whether it is a real priority flow penetration event or a sensor failure. For nodes identified as sensor faults, the expected deep response value obtained by the shallow sensor measured value through the exponential decay depth allocation function is weighted and fused with the spatial interpolation estimate of the same-layer horizontal neighboring sensor to generate corrected deep monitoring data. The corrected deep data is combined with the measured data that are determined to be normal preferential flow response and output. The measured time delay data during the over-permeability period is added to the preferential flow condition-specific sample set to update the seasonal baseline range of the seepage time delay characteristic parameters.

[0007] In a preferred embodiment, the step of determining the over-permeability event identifier and calculating the preferential flow activation intensity index based on the comparison result between the instantaneous external water load intensity and the saturated permeability coefficient of the building envelope matrix includes: step-by-step calculating the instantaneous external water load intensity... With matrix saturation permeability coefficient Compare and maintain the continuous over-permeability duration counter ;when Reaching the preset time window The system outputs an over-permeation event marker indicating it is active and records the start time of the over-permeation. According to the onset time of hyperosmosis Up to the current moment Time-average value of external water load intensity and The ratio is used to calculate the superosmosis ratio. As the priority flow activation intensity index, where , It is a dimensionless number.

[0008] In a preferred embodiment, the arrival time allocation of the exponential decay depth allocation function is calculated as follows: ,in, For shallow reference layer depth in priority flow mode To the depth The expected shortening of the time lag between them, This represents the baseline time lag value between layers under conventional matrix seepage conditions. The parameter is the crack rate of the building envelope. The permeability ratio, The maximum monitoring depth span of the monitoring section is represented by the depth difference in the independent variable of the exponent, which cancels out the dimension of the maximum depth span, making the whole a dimensionless number.

[0009] In a preferred embodiment, the response amplitude allocation of the exponentially decaying depth allocation function is calculated as follows: ,in, For depth The median of the expected response magnitude Depth under conventional matrix seepage conditions The baseline response amplitude, This is the amplitude depth attenuation coefficient. The depth is the shallow reference layer; the upper and lower bounds of the expected response amplitude envelope are respectively... and ,in This is the envelope tolerance coefficient.

[0010] In a preferred embodiment, the two-dimensional anomaly index includes a time lag deviation standardized score. and amplitude deviation standardized score : , ,in, For the actual time delay measured by the deep sensor, This represents the standard deviation of the interlayer historical time lag. To measure the actual response amplitude, The envelope half-width is used as the normalized reference for amplitude deviation.

[0011] In a preferred embodiment, the step of determining a true priority flow intrusion event or sensor failure based on the cross-correlation shape similarity coefficient includes: when the cross-correlation shape similarity coefficient... Not lower than the preset similarity threshold Furthermore, when the response waveform of the deep sensor maintains the same exponential decay characteristic after the rising edge as that of the shallow sensor, it is determined to be a true priority flow penetration event; when Below the preset similarity threshold Furthermore, when the response waveform of the deep sensor exhibits monotonic drift characteristics or step change characteristics, it is determined to be a sensor fault. Among them, monotonic drift characteristics correspond to probe contact degradation faults, and step change characteristics correspond to range saturation drift faults.

[0012] In a preferred embodiment, the vertical dimension substitution estimate in the weighted fusion is ,in This is the current measured value from the shallow reference sensor. and The median expected response amplitudes for the corresponding strata are generated by the exponential decay depth allocation function; the fused and corrected monitoring data are... ,in For spatial interpolation estimates of sensors in the same horizontal neighborhood at the same depth, , , For vertical cross-correlation strength, It is a reliability index for the horizontal spatial variability function.

[0013] In a preferred embodiment, the method further includes: merging the corrected deep monitoring data with the sensor measured data determined to be a true preferential flow penetration event and outputting it as a multi-depth monitoring data correction sequence; appending the interlayer time delay data measured by sensors at each layer during the over-permeability event to a preferential flow condition-specific sample set, wherein the preferential flow condition-specific sample set and the regular condition sample set are statistically analyzed separately, the regular condition sample set is used to update the baseline time delay values ​​of the seepage time delay characteristic parameters, and the preferential flow condition-specific sample set is used to provide time delay reference information under over-permeability conditions for the exponential decay depth allocation function.

[0014] An electronic device includes: at least one memory storing computer-executable instructions non-transitory; and at least one processor configured to run the computer-executable instructions, wherein the computer-executable instructions, when run by the processor, implement the sensor data acquisition and processing method for digital engineering as described in any embodiment of this disclosure.

[0015] A sensor data acquisition and processing system for digital engineering includes: an over-permeability event identification module, a conventional baseline acquisition module, an expected response envelope generation module, a two-dimensional anomaly index calculation module, a fault discrimination module, and a data fusion correction module. The over-permeability event identification module is configured to determine the over-permeability event identifier and calculate the preferred flow activation intensity index based on a comparison between the instantaneous external water load intensity and the saturated permeability coefficient of the retaining structure matrix. The conventional baseline acquisition module is configured to generate interlayer seepage time delay characteristic parameters based on the sliding window cross-correlation of historical data sequences from multiple depth sensors at the same monitoring section. The expected response envelope generation module is configured to perform exponential decay based on the preferred flow activation intensity index and the fracture rate parameter. The depth allocation function generates the expected shortened time delay value and the expected response amplitude envelope for each depth layer; the two-dimensional anomaly index calculation module is configured to generate a two-dimensional anomaly index based on the deviation between the measured response of the deep sensor and the expected shortened time delay value and the expected response amplitude envelope; the fault discrimination module is configured to determine whether the fault is a true priority flow intrusion event or a sensor fault based on the cross-correlation shape similarity coefficient of the response waveforms of the abnormal deep sensor and the shallow sensor; the data fusion correction module is configured to perform weighted fusion of the expected deep response value obtained by the shallow sensor measured value through the exponential decay depth allocation function and the spatial interpolation estimate of the sensor in the same horizontal neighborhood for nodes determined to be sensor faults, thereby generating corrected deep monitoring data.

[0016] The beneficial effects of this invention are: 1. This invention compares the instantaneous external water load intensity with the matrix saturated permeability coefficient step by step and introduces a continuous over-permeability duration counter. This enables timely output of over-permeability event identification and over-permeability ratio when an over-permeability event occurs, allowing subsequent processing procedures to adaptively switch between over-permeability conditions and normal conditions. This avoids misjudgment caused by inappropriate comparison of sensor responses under priority flow conditions with the normal baseline.

[0017] 2. This invention utilizes the crack rate parameter of the building envelope. With superosmosis ratio By jointly introducing an exponentially decaying depth allocation function, the expected shortening time delay value and expected response amplitude envelope of each depth layer can reflect the transmission characteristics of the preferred flow in the fracture channel, which decays exponentially with depth. The systematic deviation between the generated expected response envelope and the actual preferred flow response is objectively suppressed.

[0018] 3. This invention calculates standardized deviation scores in the time delay dimension and amplitude dimension respectively to form a dual-dimensional anomaly index, enabling anomaly patterns that cannot be triggered in a single dimension to be identified in another dimension, thereby improving the detection coverage of anomaly sensor nodes during over-permeability.

[0019] 4. This invention further extracts the cross-correlation shape similarity coefficient between the response waveform and the shallow sensor response waveform of the dual-dimensional abnormal indicator trigger node, and combines the deviation direction characteristics to distinguish between the real priority flow penetration event and the sensor physical fault, so that the fault type determination has a traceable waveform shape basis, reducing the probability of normal fast response data being erroneously rejected.

[0020] 5. For nodes identified as faulty, this invention generates correction data by weighted fusion of vertical dimension substitution estimation and horizontal dimension spatial interpolation estimation. The fusion weight is jointly determined by the vertical cross-correlation strength and the reliability of the horizontal spatial variogram, enabling the correction result to achieve an objective adaptive allocation between the reliability of the two estimation sources.

[0021] 6. This invention adds the measured time delay data of each layer during the superpermeability period to the special sample set for the priority flow condition and statistically analyzes it separately from the sample set for the normal condition condition. This ensures that the baseline of the normal condition condition is not affected by the time delay shortening effect during the superpermeability period. At the same time, it provides continuously updated superpermeability condition time delay reference information for the exponential decay depth allocation function, thus realizing closed-loop output of sensor data acquisition and processing. Attached Figure Description

[0022] Figure 1 This is a flowchart of a sensor data acquisition and processing method for digital engineering provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the pore water pressure distribution (corrected) at various depths of section M1 provided in an embodiment of the present invention; Figure 3 This is a schematic diagram comparing the interlayer seepage time delay characteristic parameters with the expected time delay of the preferred flow provided in the embodiments of the present invention; Figure 4 This is a schematic diagram of the expected response amplitude envelope of the preferred flow at each depth layer provided in the embodiments of the present invention; Figure 5 This is a schematic diagram comparing the standardized scores of the two-dimensional anomaly indicators of various sensors provided in the embodiments of the present invention; Figure 6 This is a schematic diagram of the deep foundation pit monitoring data processing flowchart provided in an embodiment of the present invention; Figure 7 This is a schematic diagram of the sensor S5 fault detection and fusion correction data flow diagram provided in an embodiment of the present invention; Figure 8 This is a schematic diagram comparing the weights and contributions of sensor S5 fusion correction data sources provided in an embodiment of the present invention; Figure 9 This is a schematic diagram comparing the measured time delays of each sensor at section M1 with the expected time delays of the preferred flow, provided in an embodiment of the present invention. Detailed Implementation

[0023] The present application will now be described in further detail with reference to the accompanying drawings. It should be noted that the following specific embodiments are only used to further illustrate the present application and should not be construed as limiting the scope of protection of the present application. Those skilled in the art can make some non-essential improvements and adjustments to the present application based on the above application content. Example

[0024] refer to Figures 1 to 9 A sensor data acquisition and processing method for digital engineering includes the following steps S100~S700.

[0025] Step S100: Obtain rainfall data and saturated permeability coefficient of the building envelope matrix at the engineering site. Based on the comparison between the instantaneous external water load intensity and the saturated permeability coefficient, determine the over-permeability event identifier and calculate the priority flow activation intensity index.

[0026] It should be noted that in step S100, the minute-by-minute rainfall data at the engineering site and the matrix saturated permeability coefficient corresponding to the retaining structure material of each monitoring section are obtained. The instantaneous external water load intensity is compared with the saturated permeability coefficient step by step. When the external water load intensity continuously exceeds the saturated permeability coefficient to the preset time window, the over-permeability event identifier and the over-permeability start time are output. The over-permeability ratio is determined as the priority flow activation intensity index based on the ratio of the external water load intensity to the saturated permeability coefficient.

[0027] For example, the minute-by-minute rainfall data collected by tipping bucket rain gauges deployed at the engineering site is converted into units to obtain the instantaneous external water load intensity. (Unit: m / s), where This is the current time step. Obtain the pre-calibrated matrix saturated permeability coefficient of the retaining structure material at this monitoring section. (Unit: m / s). At each time step, compare and The magnitude relationship is determined, and a continuous overperfusion duration counter is maintained. ;when hour, Accumulate the current time step, otherwise Reset to zero.

[0028] In one embodiment of the present invention, a preset time window is used. It was set to 30 minutes. When When the over-permeation event is active, the over-permeation start time is recorded. And calculate the superosmosis ratio:

[0029] in, From the moment of hyperosmosis onset Up to the current moment The time-average value of external water load intensity (unit: m / s). The matrix saturation permeability coefficient (unit: m / s). The superpermeability ratio is an indicator of the preferential flow activation intensity; the numerator and denominator have the same dimensions. It is a dimensionless number. It characterizes the degree to which external water load exceeds the saturated permeability of the matrix.

[0030] For example, in one instance, the saturated permeability coefficient of the matrix material of the retaining structure in a deep foundation pit project is... The instantaneous external water load intensity, calculated from minute-by-minute rainfall data during the rainstorm, remained at [m / s] for 40 consecutive minutes. If the speed exceeds m / s and the preset time window is exceeded by 30 minutes, the system will output an over-permeability event flag and calculate the over-permeability ratio. .

[0031] Step S200: Obtain historical data sequences from multiple depth sensors at the same monitoring section, and generate seepage time delay characteristic parameters and their seasonal benchmark ranges between layers based on the sliding window cross-correlation calculation results of adjacent layer sequences.

[0032] It should be noted that in step S200, historical data sequences of multiple depth sensors deployed stratified along the depth direction in the same monitoring section are acquired. A sliding window cross-correlation calculation is performed on the data sequences of adjacent layers. The input to this sliding window cross-correlation calculation is the pore water pressure time series of two adjacent depth layer sensors within the same time window, and the output is the time offset corresponding to the peak value of the cross-correlation function. Based on the time offsets extracted from multiple historical time windows, seepage time lag characteristic parameters between each layer are fitted and generated. and its seasonally varying baseline range, among which and These are the depth values ​​of two adjacent layers, serving as the baseline for anomaly detection under normal operating conditions.

[0033] For example, for adjacent depth layers m and m, in historical data, slides along the time axis with a preset window length, performs cross-correlation calculations on the two-layer sequences within each window position, and extracts the peak time offset. The median of the extracted results for all window positions is used as the seepage time lag characteristic parameter between that layer. Simultaneously, the upper and lower limits of the seepage time lag characteristic parameters are statistically divided into intervals according to the season, serving as a seasonal benchmark range. The interlayer seepage time lag characteristic parameters and baseline response amplitudes generated in step S200 are... Simultaneously, it serves as an input parameter for the exponential decay depth allocation function in step S300, where Depth under conventional matrix seepage conditions The baseline response amplitude.

[0034] In one embodiment of the present invention, the baseline time lag for conventional seepage between 2 m and 6 m layers, calculated by sliding window cross-correlation, is 120 minutes, i.e. The seepage time lag characteristic parameter has a baseline range of 100-140 minutes in the current season.

[0035] Step S300: Based on the priority flow activation intensity index and the fissure rate parameter of the building envelope, generate the expected shortening time delay value and the expected response amplitude envelope of each depth layer under the priority flow mode according to the exponential decay depth allocation function.

[0036] It should be noted that in step S300, the priority flow activation intensity index is based on the output of step S100. and the pre-acquired crack rate parameters of the building envelope The arrival time and response range of the expected priority flow volume at each depth layer during the over-permeability period are calculated using the exponential decay depth allocation function, and the expected shortened time delay value and expected response range envelope of each layer under the priority flow mode are generated.

[0037] It should be further explained that the exponential decay depth allocation function is a calculation method established for the nonlinear decay and transport characteristics of preferential flow in fractured channels. The exponential decay depth allocation function uses the product of the over-permeability ratio and the fracture rate as the decay driving factor, characterizing the exponential decay of preferential flow with depth. The exponential decay depth allocation function consists of two parts: arrival time allocation calculation and response amplitude allocation calculation. The arrival time allocation calculation determines the expected shortening time lag value between layers during over-permeability, while the response amplitude allocation calculation determines the upper and lower bounds of the expected response amplitude of each layer.

[0038] For example, for a monitoring section with a depth of Sensor layer depth and shallow reference layer depth The arrival time allocation calculation of the exponentially decaying depth allocation function is expressed as follows:

[0039] in, Depth in priority flow mode To the depth The expected shortening time lag value (unit: min). The baseline time lag value between the same layers under conventional matrix seepage conditions (unit: min, obtained from step S200). The crack rate parameter of the building envelope (dimensionless). The permeability ratio is dimensionless. The depth difference between the two layers (unit: m). This represents the maximum monitoring depth span of the monitoring section (unit: m). This represents an exponential function, where the exponent term is the independent variable. The dimensions of the intermediate depth difference and the maximum depth span cancel each other out, resulting in a dimensionless number that characterizes the coupled attenuation effect of fracture rate and over-permeability intensity on the degree of time lag shortening.

[0040] The response amplitude distribution calculation of the exponentially decaying depth distribution function is expressed as follows:

[0041] in, For depth The median of the expected response magnitude Depth under conventional matrix seepage conditions The baseline response amplitude (obtained by step S200). Amplitude depth attenuation coefficient (unit: m) ), This is the fracture rate parameter (dimensionless). The permeability ratio is dimensionless. The shallow reference layer depth (unit: m) This represents an exponential function, where the exponent term is the independent variable. middle (m) )and (m) The dimensions cancel each other out, resulting in a dimensionless number. It should be noted that when... At that time, the exponent term Gain factor Take the maximum value. Median expected response magnitude This serves as the denominator for the vertical dimension substitution estimation formula in step S600. It is based on the median of the expected response magnitude. Upper bound of the expected response amplitude envelope and the lower realm :

[0042]

[0043] in, The envelope tolerance coefficient (dimensionless) characterizes the allowable fluctuation range of the preferred flow response amplitude near the expected median.

[0044] In one embodiment of the invention, sensors monitoring the cross-section are arranged at five depth levels. m、 m、 m、 m、 m, maximum monitoring depth span m, fracture ratio parameter Envelope tolerance coefficient The magnitude of the preferential flow events in the historical data of the cross-section is determined based on the degree of dispersion. The expected time delay value was calculated for each depth layer. and expected response magnitude envelope ,in For depth level bit index, Depth difference between layers The values ​​are 4 m, 8 m, 12 m, and 16 m, respectively. Substituting these values ​​into the above arrival time allocation calculation and response amplitude allocation calculation, we obtain the expected shortening time delay value and expected response amplitude envelope corresponding to each layer.

[0045] Step S400: During an over-permeability event, based on the comparison between the measured response of the deep sensor and the expected response envelope of the preferred flow mode, determine the time delay deviation standardized score and the amplitude deviation standardized score to generate a two-dimensional anomaly index.

[0046] It should be noted that in step S400, when the over-permeability event flag output in step S100 is in an active state, a two-dimensional comparison is performed on the measured response of each deep sensor: the measured response arrival time of each deep sensor is compared with the expected shortened time delay value of the priority flow mode generated in step S300, and the measured response amplitude is compared with the expected response amplitude envelope. The time delay deviation standardized score and the amplitude deviation standardized score are calculated respectively, and the two standardized scores together constitute the two-dimensional anomaly index under the priority flow condition.

[0047] For example, for depth The sensor's response arrival time was measured relative to the shallow reference layer. The measured time delay is The measured response amplitude is The standardized scores for time delay deviation and amplitude deviation are calculated as follows:

[0048]

[0049] in, For depth Standardized score of time delay deviation The measured time delay is in minutes. The expected shortened time delay value (unit: min) generated in step S300. This represents the standard deviation of the interlayer historical time lag (unit: min), with the numerator and denominator having the same dimensions. It is a dimensionless number; For depth The standardized score of amplitude deviation To measure the actual response amplitude, The median of the expected response magnitude generated in step S300. The upper bound of the expected response amplitude envelope, This is the lower bound of the expected response amplitude envelope. Using the envelope half-width as the normalization benchmark for amplitude deviation, both the numerator and denominator are amplitude units. It is a dimensionless number.

[0050] In one embodiment of the present invention, when Exceeding the preset time delay tolerance threshold or Exceeding the preset allowable amplitude threshold At that time, the deep sensor was marked as being in a two-dimensional anomaly indicator triggered state. For example, in one instance, the 18 m deep sensor measured a response 15 minutes after the onset of hyperperfusion, with its time-lag deviation standardized score... The time delay tolerance threshold was not exceeded, but the measured response amplitude exceeded the upper bound of the expected response amplitude envelope, and the amplitude deviation standardized score was [not specified]. If the amplitude exceeds the allowable threshold, the sensor is flagged as triggered in the amplitude dimension of the two-dimensional anomaly indicators.

[0051] Step S500: For deep sensors whose dual-dimensional abnormal indicators exceed the allowable range, determine them as true priority flow penetration events or sensor failures based on the cross-correlation shape similarity coefficient between their response waveforms and those of shallow sensors.

[0052] It should be noted that in step S500, for the deep sensor whose dual-dimensional anomaly index exceeds the corresponding allowable threshold in step S400, the response waveform sequence of the sensor within a certain time period after the triggering of the over-permeability event is extracted. At the same time, the response waveform sequence of the shallow reference sensor at the same cross section within the same time period is extracted. Normalized cross-correlation calculation is performed on the two sequences. The input of the normalized cross-correlation calculation is the response waveform sequence of the deep sensor and the response waveform sequence of the shallow sensor, and the output is the cross-correlation shape similarity coefficient. Based on the cross-correlation shape similarity coefficient Similarity threshold Based on the comparison results and the directional characteristics of time delay and amplitude deviation, it was determined that the sensor's response belonged to a true priority flow intrusion event or a sensor malfunction.

[0053] For example, the decision logic includes the following two cases: Scenario 1: If the time delay deviation of the deep sensor is standardized... Within the allowable range or with offset but cross-correlation shape similarity coefficient If the response waveform of the deep sensor maintains the same rising edge and exponential decay characteristic as that of the shallow sensor, then the response of the sensor is determined to be a normal fast response caused by a real priority flow penetration event, and no data correction is performed.

[0054] Scenario 2: If the amplitude deviation of the deep sensor is standardized... Exceeding the allowable threshold and cross-correlation shape similarity coefficient If the response waveform of the deep sensor exhibits monotonic drift characteristics or step change characteristics that are not similar to those of the shallow sensor, it is determined that the sensor has experienced a concurrent physical fault during the over-permeation period. The fault type is further marked according to the waveform characteristics: monotonic drift characteristics correspond to probe contact degradation faults, and step change characteristics correspond to range saturation drift faults.

[0055] In one embodiment of the present invention, a similarity threshold It is set to 0.7. For example, the cross-correlation shape similarity coefficient is obtained by normalizing the cross-correlation calculation of the response waveforms of a 18 m depth sensor and a 2 m shallow depth sensor. The similarity was below the threshold of 0.7, and the response waveform of the 18 m depth sensor showed a monotonic drift characteristic. Based on this, it was determined that the 18 m depth sensor experienced a physical failure of probe contact degradation during hyperperfusion.

[0056] Step S600: For nodes determined to be sensor faults, perform weighted fusion based on the measured values ​​of shallow sensors, the expected deep response values ​​of the priority flow depth allocation function, and the spatial interpolation estimates of the horizontal neighboring sensors in the same layer, to generate fused and corrected deep monitoring data.

[0057] It should be noted that in step S600, for the deep sensor nodes that were determined to be sensor faults in step S500, alternative estimates are obtained from the vertical and horizontal dimensions respectively, and the corrected monitoring data is generated by weighted fusion.

[0058] For example, the process of obtaining the vertical dimension substitute estimate is as follows: obtain a shallow reference sensor (depth) from the same cross-section. ) measured value ,in For the current moment, the measured value of the shallow reference sensor is... Substitute the response amplitude allocation into the exponential decay depth allocation function in step S300 to determine the fault node depth. At the present moment Expected deep response value :

[0059] in, Replace the estimated value with the vertical dimension. This is the current measured value from the shallow reference sensor. The depth generated in step S300 Median expected response magnitude The depth generated in step S300 Median expected response magnitude (when) hour, ),ratio This is a dimensionless scaling factor, and... Dimensions consistent Maintain the same dimensions as the measured value.

[0060] The process of obtaining the horizontal dimension substitute estimate is as follows: obtain the values ​​at the same depth as the fault node. The measured values ​​of other monitoring section sensors in the horizontal neighborhood are used to perform spatial interpolation calculations on these measured values ​​(the input of this spatial interpolation calculation is the measured values ​​and spatial coordinates of each section sensor in the same horizontal neighborhood at the same depth, and the output is the interpolated estimate of the spatial location of the fault node) to obtain a substitute estimate of the horizontal dimension. .

[0061] Alternative estimates based on vertical dimension and horizontal dimension substitute estimates Weighted fusion is performed to generate corrected deep monitoring data:

[0062] in, To integrate the corrected deep-sea monitoring data, To integrate weights in the vertical dimension, To fuse weights along the horizontal dimension, and , and Same dimensions, weighted fusion result Maintain the same dimensions as the measured value.

[0063] It should be noted that the weights are fused along the vertical dimension. Weights are integrated with horizontal dimensions. The vertical dimension fusion weights are determined jointly by the vertical cross-correlation strength and the reliability of the horizontal spatial variability function. Weights are integrated with horizontal dimensions. The calculation method is as follows:

[0064]

[0065] in, The vertical cross-correlation intensity (dimensionless) between the shallow reference sensor and the fault node depth layer under historical normal operating conditions. A larger value indicates a higher reliability of the vertical dimension substitution estimate; This is a dimensionless reliability index (SRI) for the spatial variogram fitted based on measured values ​​from sensors in the same horizontal neighborhood at the fault node location. A larger value indicates higher reliability of the horizontal space interpolation estimate; and Since they have the same dimensions, their sum is used as the normalized denominator to ensure... Vertical dimension fusion weights Weights are integrated with horizontal dimensions. Each is proportional to the reliability of its respective dimension, achieving reliability-driven adaptive fusion.

[0066] In one embodiment of the present invention, the vertical cross-correlation strength of a fault node at a depth of 18 m is... Reliability index of horizontal spatial variogram Thus determined , The vertical dimension and horizontal dimension replacement estimates are fused according to the above weights to generate fused and corrected deep monitoring data with a depth of 18 m.

[0067] Step S700: Combine the corrected deep data with the measured data determined to be normal preferential flow response and output them. Then, add the measured time delay data during the over-permeability period to the preferential flow condition-specific sample set to update the seasonal baseline range of the seepage time delay characteristic parameters.

[0068] It should be noted that in step S700, the fused and corrected deep monitoring data generated in step S600 is merged with the sensor measured data determined as the real priority flow penetration event in step S500, and the multi-depth monitoring data correction sequence after over-permeability anomaly identification is output as the input data of the downstream engineering analysis and early warning module.

[0069] For example, for all depth layers of the same monitoring section, the sensors that are determined to have normal priority flow response in step S500 retain their original measured data, and the fused and corrected deep monitoring data in step S600 replaces the original abnormal data. The two parts of data are merged into a complete multi-depth monitoring data correction sequence according to the time series and depth layer index.

[0070] In one embodiment of the present invention, the interlayer time delay data measured by sensors at each layer during the current over-permeability event (including the interlayer measured time delay determined as a normal preferential flow response) are added to the preferential flow condition-specific sample set of seepage time delay characteristic parameters in step S200. The preferential flow condition-specific sample set and the conventional condition sample set are statistically analyzed separately and are not mixed. The preferential flow condition-specific sample set is used to statistically analyze the distribution characteristics of interlayer time delays under over-permeability conditions, providing time delay reference information under over-permeability conditions for the exponential decay depth allocation function in step S300; the conventional condition sample set continues to be used to update the baseline time delay values ​​in step S200. This ensures that the baseline under normal operating conditions is not affected by the time lag shortening effect during over-permeability, and completes the closed-loop output of sensor data acquisition and processing.

[0071] In one embodiment of the present invention, a sensor data acquisition and processing system for digital engineering includes an over-permeability event identification module, a conventional baseline acquisition module, an expected response envelope generation module, a two-dimensional anomaly index calculation module, a fault discrimination module, a data fusion correction module, and a closed-loop update module.

[0072] The over-permeability event identification module is configured to acquire minute-by-minute rainfall data at the engineering site and the matrix saturated permeability coefficient corresponding to the retaining structure material at each monitoring section. It compares the instantaneous external water load intensity with the saturated permeability coefficient at each time step. When the over-permeability event continuously exceeds a preset time window, it outputs an over-permeability event identifier and calculates the over-permeability ratio as a priority flow activation intensity index. For a detailed description of the over-permeability event identification module, please refer to the relevant description of step S100 in the above-described sensor data acquisition and processing method embodiment.

[0073] The conventional baseline acquisition module is configured to acquire historical data sequences from multiple depth sensors at the same monitoring section, perform sliding window cross-correlation calculations on adjacent layer sequences, extract the time offsets corresponding to the cross-correlation peaks, and fit and generate seepage time lag characteristic parameters and their seasonal baseline ranges between each layer. This corresponds to the relevant description of step S200 in the above-described sensor data acquisition and processing method embodiment.

[0074] The expected response envelope generation module is configured to calculate the expected arrival time and response amplitude range of each depth layer during hyperpermeability based on the priority flow activation intensity index and the retaining structure fracture rate parameter, using an exponential decay depth allocation function, and generate the expected shortening time delay value and the expected response amplitude envelope. This corresponds to the relevant description of step S300 in the above sensor data acquisition and processing method.

[0075] The dual-dimensional anomaly index calculation module is configured to compare the measured response arrival time of each deep sensor with the expected shortened time delay value and the response amplitude with the expected response amplitude envelope during an over-permeability event, calculate the time delay deviation standardized score and the amplitude deviation standardized score, and generate a dual-dimensional anomaly index. See the relevant description for step S400.

[0076] The fault identification module is configured to extract the cross-correlation shape similarity coefficient between the response waveform of a deep sensor and the response waveform of a shallow sensor at the same cross-section for deep sensors whose dual-dimensional abnormal indicators exceed the allowable range. It then combines this coefficient with time delay and amplitude deviation direction characteristics to determine the true priority flow penetration event or sensor fault and marks the fault type. This corresponds to the relevant description of step S500.

[0077] The data fusion and correction module is configured to perform weighted fusion of the expected deep response value calculated by combining the measured values ​​of shallow sensors with the preferred flow depth allocation function and the spatial interpolation estimate of the horizontal neighboring sensors at the same layer for nodes determined to be sensor faults. The fusion weights are jointly determined by the vertical cross-correlation strength and the reliability of the horizontal spatial variability function, generating fused and corrected deep monitoring data. See the relevant description for step S600.

[0078] The closed-loop update module is configured to merge the fused and corrected deep monitoring data with the measured data determined to be normal preferential flow response, outputting a multi-depth monitoring data correction sequence after over-permeability anomaly identification. It also appends the measured time-delay data of each layer during the over-permeability period to the preferential flow condition-specific sample set, and statistically analyzes it separately with the regular condition sample set to update the seasonal baseline range of the seepage time-delay characteristic parameters. This corresponds to the relevant description of step S700.

[0079] The technical effects of this sensor data acquisition and processing system are the same as those of the sensor data acquisition and processing methods described above, and will not be repeated here.

[0080] In a deep foundation pit monitoring project, the monitoring area was a subway station foundation pit with an excavation depth of approximately 22m. The retaining structure was a diaphragm wall, and the matrix saturation permeability coefficient of the wall material, as determined by indoor permeability tests, was [value missing]. m / s. The monitoring system deployed five pore water pressure sensor layers along the depth direction in the soil outside the pit, located at depths of 2 m, 6 m, 10 m, 14 m, and 18 m, respectively, corresponding to section M1. On a certain day during the flood season of year A, the on-site tipping bucket rain gauge recorded a continuous heavy rainfall event. The system converted the minute-by-minute rainfall data to obtain the instantaneous external water load intensity. Sequence, and with Compare at each time step.

[0081] In this application scenario, the system calculates the rainfall data from the minute-by-minute rainfall data collected by the rain gauge. The sequence detected activity starting from the 37th minute. Continue to exceed m / s, continuous overperfusion duration counter The preset time window was reached in the 67th minute. At min, the system outputs an overperfusion event marker indicating an active state and records the start time of the overperfusion. It is the 37th minute. The time-averaged external water load intensity during the interval from the onset of infiltration to the current time is... m / s, based on which the excess permeability ratio is calculated:

[0082] Table 1 Key parameters and output results for identifying over-permeability events

[0083] After identifying the over-permeability event in step S100, the system retrieves historical data sequences accumulated by pore water pressure sensors at five depths in section M1 under normal operating conditions before the flood season in year A. It then performs sliding window cross-correlation calculations on the sequences of adjacent layers, setting the window length to 240 minutes and the sliding step size to 30 minutes. For each window location, the system extracts the time offset corresponding to the peak value of the cross-correlation function, and calculates the median of all window location extraction results as the interlayer seepage time lag characteristic parameter. Simultaneously, the upper and lower limits of the flood season subset are statistically analyzed as the seasonal baseline range. Baseline response amplitude between layers is also considered. The statistics are also completed synchronously in this step, serving as input for the response amplitude allocation calculation in step S300.

[0084] Table 2. Characteristic parameters of seepage time lag between adjacent layers in section M1 and seasonal baseline range

[0085] Based on the output of step S100 And the crack rate parameters of the M1 section of the enclosure structure With depth m represents the shallow reference layer and the maximum monitoring depth span. m, amplitude depth attenuation coefficient Envelope tolerance coefficient For each depth layer m、 m、 m、 m performs arrival time allocation calculation and response amplitude allocation calculation respectively.

[0086] by m( Taking m as an example, in the arrival time allocation calculation, The time delay between inter-story baselines ranges from 2 m to 18 m, and is obtained by summing the time delays of adjacent inter-story baselines. min, then:

[0087] The median expected response magnitude at point m is:

[0088] The median expected response magnitude at point m is:

[0089] The upper and lower bounds of the envelope are:

[0090] Table 3 Expected response parameters of preferential flow modes at various depths in section M1

[0091] After the hyperpermeability event was activated, the system performed a two-dimensional comparison of the measured responses of the sensors at each depth in section M1. In this hyperpermeability event, the deviations between the response arrival times of the sensors at depths of 6 m, 10 m, and 14 m and the expected shortened time delay values ​​were all within the allowable range. However, the sensor at depth of 18 m (sensor number S5) responded approximately 15 minutes after the onset of hyperpermeability, with a measured time delay of [missing value]. min, measured response amplitude The standard deviation of the interlayer historical time delay is 2.54 kPa, exceeding the upper limit of the expected response amplitude envelope. Taking a depth of 18 m as an example, the standard deviation of the interlayer historical time delay is... min, the standardized scores of time delay bias and amplitude bias are respectively:

[0092]

[0093] Standardized score of time delay deviation Far exceeding the allowable time delay threshold Standardized score of amplitude deviation Exceeding the allowable amplitude threshold Sensor S5 was marked as a dual-dimensional anomaly indicator with both dimensions simultaneously triggered.

[0094] Table 4 Calculation results of dual-dimensional anomaly indicators of various deep sensors at section M1

[0095] For sensor S5, which is triggered by a dual-dimensional anomaly index, its response waveform sequence within 60 minutes after the hyperpermeability event is extracted. Simultaneously, the response waveform sequence of the shallow reference sensor S1 (depth 2 m) within the same time period is extracted. Normalized cross-correlation is then performed on the two sequences. The calculation results show the cross-correlation shape similarity coefficient. Below the similarity threshold Further examination of the response waveform of sensor S5 revealed a monotonic drift characteristic relative to the onset of hyperperfusion, which was significantly inconsistent with the exponential decay waveform characteristic of shallow sensor S1 after the rising edge. According to the scenario two determination logic in step S500, sensor S5 was determined to have experienced a concurrent physical fault during hyperperfusion, and the monotonic drift characteristic corresponded to a probe contact degradation fault type. The original measured data of S5 was marked as invalid and proceeded to step S600 for fusion correction.

[0096] Table 5 Key parameters for sensor S5 fault diagnosis process

[0097] For sensor S5 (depth 18 m), which was determined to have probe contact degradation failure, surrogate estimates were obtained from both the vertical and horizontal dimensions. These estimates were obtained at a typical moment during an over-permeability event. The measured value of the shallow reference sensor S1 (depth 2 m) is kPa, step S300 has been generated. kPa kPa, the alternative estimate for the vertical dimension is:

[0098] Horizontal dimension substitute estimate The horizontal dimension of section M1 was obtained by spatial interpolation from the sensor measurements at horizontal neighboring sections M2 and M3 at the same depth of 18 m. The measured value of M2 at a depth of 18 m was 6.35 kPa, and that of M3 was 6.55 kPa. After spatial interpolation, the horizontal dimension of section M1 was estimated. kPa.

[0099] Vertical cross-correlation strength Reliability index of horizontal spatial variogram The fusion weights are:

[0100] The fused and corrected deep monitoring data are as follows:

[0101] Table 6 Key data of sensor S5 fusion calibration process

[0102] The 18 m depth fusion correction data generated in step S600 The original measured data of sensors S2, S3, and S4, which were determined to be in normal priority flow response in step S500, are merged according to time series and depth layer index to generate a complete multi-depth monitoring data correction sequence during this hyperpermeability event at section M1, and output to the downstream engineering early warning module.

[0103] In this over-permeability event, the interlayer time delay data measured by sensors S2, S3, and S4 (108 min, 368 min, and 551 min, respectively) were added to the priority flow condition-specific sample set of seepage time delay characteristic parameters in step S200. These samples were statistically stored separately from the conventional condition sample set and were not mixed. The conventional condition sample set continued to be used to update the baseline time delay values. This ensures that the baseline of normal operating conditions is not affected by the time lag shortening effect during this period of over-infiltration.

[0104] Table 7. Output results of the merged correction sequence of multi-depth monitoring data from section M1 The embodiments of the present invention have been described above. However, the embodiments are not limited to the specific implementation methods described above. The specific implementation methods described above are merely illustrative and not restrictive. Those skilled in the art can make more equivalent embodiments under the guidance of the present embodiments, and all of them are within the protection scope of the present embodiments.

Claims

1. A sensor data acquisition and processing method for digital engineering, characterized in that, Includes the following steps: The over-permeability event identifier is determined and the preferential flow activation intensity index is calculated based on the comparison between the instantaneous external water load intensity and the saturated permeability coefficient of the building envelope matrix. The seepage time delay characteristic parameters between layers are generated by sliding window cross-correlation calculation based on the historical data sequence of multiple depth sensors at the same monitoring section; Based on the priority flow activation intensity index and the fracture rate parameter, the expected shortening time delay value and the expected response amplitude envelope of each depth layer are generated by the exponential decay depth allocation function. During an over-permeability event, a two-dimensional anomaly index is generated based on the deviation between the measured response of the deep sensor and the expected shortened time delay value and the expected response amplitude envelope. For deep sensors whose dual-dimensional abnormal indicators exceed the allowable range, the cross-correlation shape similarity coefficient between the deep sensor and the shallow sensor response waveform is used to determine whether it is a real priority flow penetration event or a sensor failure. For nodes identified as sensor faults, the expected deep response value obtained by the shallow sensor measured value through the exponential decay depth allocation function is weighted and fused with the spatial interpolation estimate of the same-layer horizontal neighboring sensor to generate corrected deep monitoring data. The corrected deep data is combined with the measured data that are determined to be normal preferential flow response and output. The measured time delay data during the over-permeability period is added to the preferential flow condition-specific sample set to update the seasonal baseline range of the seepage time delay characteristic parameters.

2. The sensor data acquisition and processing method for digital engineering according to claim 1, characterized in that, The process of determining the over-permeability event identifier and calculating the preferential flow activation intensity index based on the comparison between the instantaneous external water load intensity and the saturated permeability coefficient of the building envelope matrix includes: Instantaneous external water load intensity at each time step With matrix saturation permeability coefficient Compare and maintain the continuous over-permeability duration counter ; when Reaching the preset time window The system outputs an over-permeation event marker indicating it is active and records the start time of the over-permeation. ; Based on the time of hyperosmosis initiation Up to the current moment Time-average value of external water load intensity and The ratio is used to calculate the superosmosis ratio. As the priority flow activation intensity index, where , It is a dimensionless number.

3. The sensor data acquisition and processing method for digital engineering according to claim 1, characterized in that, The arrival time allocation of the exponential decay depth allocation function is calculated as follows: ， in, For shallow reference layer depth in priority flow mode To the depth The expected shortening of the time lag between them, This represents the baseline time lag value between layers under conventional matrix seepage conditions. The parameter is the crack rate of the building envelope. The permeability ratio, The maximum monitoring depth span of the monitoring section is represented by the depth difference in the independent variable of the exponent, which cancels out the dimension of the maximum depth span, making the whole number dimensionless.

4. The sensor data acquisition and processing method for digital engineering according to claim 3, characterized in that, The response amplitude allocation of the exponentially decaying depth allocation function is calculated as follows: ， in, For depth The median of the expected response magnitude Depth under conventional matrix seepage conditions The baseline response amplitude, This is the amplitude depth attenuation coefficient. This is the shallow reference layer depth; The upper and lower bounds of the expected response amplitude envelope are respectively and ,in This is the envelope tolerance coefficient.

5. The sensor data acquisition and processing method for digital engineering according to claim 4, characterized in that, The two-dimensional anomaly indicators include the standardized score of time lag deviation. and amplitude deviation standardized score : ， ， in, For the actual time delay measured by the deep sensor, This represents the standard deviation of the interlayer historical time lag. To measure the actual response amplitude, The envelope half-width is used as the normalized reference for amplitude deviation.

6. The sensor data acquisition and processing method for digital engineering according to claim 1, characterized in that, The determination of a true priority flow intrusion event or sensor failure based on the cross-correlation shape similarity coefficient includes: When the cross-correlation shape similarity coefficient Not lower than the preset similarity threshold Furthermore, when the response waveform of the deep sensor maintains the same exponential decay characteristic after the rising edge as that of the shallow sensor, it is determined to be a true priority flow penetration event. When the cross-correlation shape similarity coefficient Below the preset similarity threshold Furthermore, when the response waveform of the deep sensor exhibits monotonic drift characteristics or step change characteristics, it is determined to be a sensor fault. Among them, monotonic drift characteristics correspond to probe contact degradation faults, and step change characteristics correspond to range saturation drift faults.

7. The sensor data acquisition and processing method for digital engineering according to claim 1, characterized in that, The vertical dimension substitution estimate in the weighted fusion is ,in This is the current measured value from the shallow reference sensor. and These are the median expected response amplitudes of the corresponding layers generated by the exponentially decaying depth allocation function; The monitoring data after fusion correction is ,in This is a spatial interpolation estimate of the sensor space in the same horizontal neighborhood at the same depth. , , For vertical cross-correlation strength, It is a reliability index for the horizontal spatial variability function.

8. The sensor data acquisition and processing method for digital engineering according to claim 1, characterized in that, Also includes: The corrected deep monitoring data is combined with the sensor measured data determined to be true priority flow penetration events and output as a multi-depth monitoring data correction sequence. The interlayer time delay data measured by sensors at each layer during the over-permeability event are added to the priority flow condition-specific sample set. The priority flow condition-specific sample set and the regular condition sample set are statistically analyzed separately. The regular condition sample set is used to update the baseline time delay values ​​of the seepage time delay characteristic parameters, and the priority flow condition-specific sample set is used to provide time delay reference information under over-permeability conditions for the exponential decay depth allocation function.

9. An electronic device, characterized in that, include: At least one memory stores computer-executable instructions non-transiently; At least one processor, configured to run the computer-executable instructions, The computer-executable instructions are executed by the processor to implement the method according to any one of claims 1 to 8.

10. A sensor data acquisition and processing system for digital engineering, used to execute the sensor data acquisition and processing method for digital engineering according to any one of claims 1 to 8, characterized in that, include: The over-permeability event identification module is configured to determine the over-permeability event identifier and calculate the priority flow activation intensity index based on the comparison between the instantaneous external water load intensity and the saturated permeability coefficient of the building envelope matrix. The conventional baseline acquisition module is configured to generate interlayer seepage time delay characteristic parameters based on the sliding window cross-correlation calculation of historical data sequences from multiple depth sensors at the same monitoring section. The expected response envelope generation module is configured to generate the expected shortening time delay value and expected response amplitude envelope of each depth layer based on the priority flow activation intensity index and the fracture rate parameter and through an exponential decay depth allocation function. The dual-dimensional anomaly index calculation module is configured to generate dual-dimensional anomaly indexes based on the deviation between the measured response of the deep sensor and the expected shortened time delay value and the expected response amplitude envelope. The fault identification module is configured to determine whether a real priority flow penetration event or a sensor fault is determined based on the cross-correlation shape similarity coefficient of the response waveforms of abnormal deep and shallow sensors. The data fusion correction module is configured to perform weighted fusion of the expected deep response value obtained by the shallow sensor measured value through the exponential decay depth allocation function and the spatial interpolation estimate of the same-layer horizontal neighboring sensor for nodes determined to be sensor faults, and generate corrected deep monitoring data.

Images