Flexible strain sensor and measurement method for measuring deformation of oil and gas pipeline

Abstract

This invention discloses a flexible strain sensor and measurement method for measuring the deformation of oil and gas pipelines, belonging to the field of pipeline structural health monitoring technology. The flexible strain sensor includes a flexible printed circuit board (FPC), four strain gauges, two thermistors, and a ribbon cable connector. The FPC uses a polyimide substrate. The four strain gauges have pre-drilled holes arranged at equal intervals along the circumference, and the two thermistor pre-drilled holes are symmetrically distributed in the center. Each sensor is electrically connected to the FPC via laser welding. The thermistors measure temperature in real time, and temperature compensation is achieved through a polynomial algorithm. The FPC can integrate a MEMS accelerometer, an acoustic emission sensor, and an ASIC chip, enabling local anomaly detection and data compression in conjunction with an edge AI terminal. This flexible strain sensor is adhered to the anti-corrosion layer of the pipeline's outer wall without affecting the anti-corrosion effect. It is easy to install and can be connected to a static strain gauge via a ribbon cable connector or used with a wireless acquisition module for remote intelligent monitoring.

Description

Technical Field

[0001] The present invention belongs to the technical field of pipeline structural health monitoring, and particularly relates to a flexible strain sensor for measuring the deformation of oil and gas transmission pipelines and a measurement method. Background Art

[0002] The means for pipeline strain monitoring mainly rely on installing various rigid or semi-rigid sensors on the outer wall of the pipeline, such as vibrating wire sensors, fiber Bragg grating (FBG) sensors, and resistance strain gauges, etc. Although these traditional methods are relatively mature in technology, they have shown significant limitations in engineering practice. The installation process of vibrating wire sensors is complex, usually requiring welding or fixture fixation, which will damage the original anti-corrosion layer of the pipeline, and its volume is relatively large, with poor adaptability in complex parts such as elbows and welds. Although fiber Bragg grating sensors have advantages such as anti-electromagnetic interference and suitability for distributed measurement, their system cost is extremely high, and they have extremely high requirements for installation processes (such as welding and gluing) and protection measures. Moreover, there is a cross-sensitivity problem between strain and temperature, and additional compensation gratings need to be arranged, increasing the system complexity and failure risk. In addition, in order to reconstruct the complete strain field of the pipeline cross-section, traditional point-type or quasi-distributed measurement methods often need to arrange multiple sensors (usually in a "pin" shape at 90° or 120°) circumferentially in the same cross-section. This not only doubles the material and installation costs but also poses greater challenges to data synchronization and system reliability. Therefore, developing a flexible strain sensing solution that can fit the pipeline surface, does not affect the integrity of the anti-corrosion layer, has high reliability, low cost, and is convenient for installation and maintenance has become a clear and urgently needed technical direction in this field. Summary of the Invention

[0003] In order to solve the above technical problems, the present invention proposes a flexible strain sensor for measuring the deformation of oil and gas transmission pipelines, comprising:

[0004] A flexible printed circuit board, on which four strain gauge reserved holes and two thermistor reserved holes are provided;

[0005] A strain acquisition strip, the strain acquisition strip includes a plurality of strain gauges, and the plurality of strain gauges are respectively connected to the solder joints on the corresponding strain gauge reserved holes by welding. Each of the strain gauges includes a plurality of strain sensitive grids, and a part of the strain sensitive grids are arranged along the axial direction of the pipeline, and another part of the strain sensitive grids are arranged along the circumferential direction of the pipeline;

[0006] A plurality of thermistors, the plurality of thermistors are connected to the solder joints on the thermistor reserved holes by welding and are symmetrically distributed about the center line of the flexible printed circuit board;

[0007] The flexible printed circuit board (FPC) has a flexible printed circuit board cable inside. The FPC cable connects the strain gauges and the thermistors at various locations, and then converges in the middle of the flexible printed circuit board before connecting to the cable connector.

[0008] Furthermore, the flexible printed circuit board integrates a microprocessor or application-specific integrated circuit for preliminary signal amplification, filtering, real-time temperature compensation, and sensor health self-diagnosis.

[0009] Furthermore, the sensor health self-diagnosis includes: detecting the open or short circuit status of the strain gauge bridge, monitoring the open circuit fault of the thermistor, judging the integrity of the flexible printed circuit board cable connection, and outputting a fault code through the cable connector when a fault occurs.

[0010] Furthermore, the flexible printed circuit board uses a polyimide substrate with a thickness of 0.2 mm, a length customized according to the pipe diameter, and a width of 50 mm.

[0011] Furthermore, the plurality of strain gauges are evenly distributed on the flexible printed circuit board, and the position of each strain gauge corresponds to a monitoring point on the outer wall of the pipe.

[0012] Furthermore, the plurality of strain gauges are arranged at equal intervals on the flexible printed circuit board.

[0013] Furthermore, there are four strain gauges. When the flexible strain sensor is deployed on the pipeline, the four corresponding monitoring points are located at the 0° outer arc side, 90° neutral plane, 180° inner arc side and 270° neutral plane positions in the circumferential direction of the pipeline, respectively.

[0014] This invention also proposes a method for measuring the deformation of oil and gas transmission pipelines using a flexible strain sensor, comprising the following steps:

[0015] The flexible strain sensor is attached to the anti-corrosion layer of the outer wall of the pipe by adhesive bonding, so that the multiple strain gauges on the strain acquisition band correspond one-to-one with the multiple monitoring points on the outer wall of the pipe.

[0016] The data acquisition device is connected through the ribbon cable connector to acquire the resistance change signal of the strain gauge due to pipe deformation, as well as the temperature signal acquired by the thermistor.

[0017] Based on the temperature signal, a temperature compensation algorithm is used to correct the resistance change signal;

[0018] Based on the corrected resistance change signal, the true strain at the pipeline monitoring point is calculated.

[0019] Furthermore, the temperature compensation algorithm calculates the true strain based on the following formula. :

[0020] ;

[0021] in, To measure strain, K is the temperature coefficient of resistance of the strain gauge material, and K is the strain gauge sensitivity coefficient. The coefficient of thermal expansion of the pipe material. The coefficient of thermal expansion of the substrate material for flexible printed circuit boards. The temperature change measured by the thermistor is the base shear transfer coefficient.

[0022] Furthermore, the attention level is triggered when the strain at a single point exceeds the historical average; the warning level is triggered when the reconstruction error exceeds the threshold or the trend is abnormal; the danger level is triggered when the multi-point coordination is abnormal or the predicted failure probability is greater than 30%; and the emergency level is triggered when the strain mutation exceeds 50% of the yield strength.

[0023] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0024] Replacing traditional strain sensors with a 0.1-0.5mm flexible FPC substrate, these sensors are embedded into the pipe's anti-corrosion layer, achieving zero-intrusion installation and avoiding damage to the anti-corrosion structure. Installation is convenient and costs are reduced by over 50%. A four-point distributed arrangement simultaneously acquires axial, circumferential strain, and temperature data, combined with real-time temperature compensation using thermistors to eliminate thermal drift errors. Accelerated aging verification shows a service life exceeding 10 years with an accuracy maintained at ±2%, significantly superior to the 3-5 year lifespan of conventional sensors.

[0025] Optional MEMS vibration / acoustic emission sensors can be equipped to achieve multi-physics field monitoring, and integrated ASIC chips or edge AI terminals can achieve a local anomaly detection rate of >95% and a data compression ratio of 600:1; combined with cloud-based digital twins and LSTM prediction models, fatigue life warnings of 30-90 days can be achieved.

[0026] Through ultra-thin flexible structure, multi-point distributed layout, real-time temperature compensation and optional edge AI integration, long-term, high-precision and intelligent monitoring of pipeline axial and circumferential strain, temperature and dynamic response is achieved without affecting the pipeline anti-corrosion layer, which is significantly better than the 3-5 year lifespan of traditional vibrating wire sensors. Attached Figure Description

[0027] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0028] Figure 1 This is a schematic diagram of the device structure of the present invention;

[0029] Figure 2 For strain gauge data acquisition, see physical image;

[0030] Figure 3 This is a schematic diagram of the flexible printed circuit board of the present invention;

[0031] Figure 4 This is a flowchart of the measurement method of the present invention;

[0032] Figure 5 These are the results of the cyclic loading-unloading test of the present invention;

[0033] Figure 6 This is a fitting graph of the stress and voltage simulation signal changes according to the present invention;

[0034] Figure 7 The strain change curves of the strain monitoring device and strain gauge of the present invention are shown.

[0035] Figure 8 This is a schematic diagram of the on-site installation and equipment connection method of strain gauges. Detailed Implementation

[0036] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0037] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0038] Example 1: Refer to Figure 1 The right figure shows a side view of a section of a cylindrical long-distance pipeline. Four monitoring points are arranged on each pipeline section. Four-point strain gauges are attached to the anti-corrosion layer of the pipeline's outer wall using an adhesive method. Strain monitoring is performed at these four monitoring points. The actual four-point strain gauges are shown below. Figure 2 As shown.

[0039] The pipeline four-point strain acquisition band includes: a flexible printed circuit board, several strain gauges, several thermistors, and ribbon cable connectors.

[0040] The flexible printed circuit board (FPC) further includes: four strain gauge pre-drilled holes, two thermistor pre-drilled holes, and an FPC cable. The strain gauge pre-drilled holes are distributed in four equal parts on the strain acquisition strip. Each strain gauge is connected to a solder joint on each strain gauge pre-drilled hole via soldering, with the position of each strain gauge corresponding sequentially to the position of each monitoring point. The thermistor pre-drilled holes are located in the central region of the FPC, symmetrically distributed about the centerline of the FPC. Each thermistor is connected to a solder joint on each thermistor pre-drilled hole via soldering. Each strain gauge contains three strain-sensitive grids, two arranged axially along the pipe and one circumferentially along the pipe. The FPC cable is evenly distributed throughout the FPC to connect the strain gauges and thermistors at various points, converging in the central part of the FPC, with exposed cable connectors directly connected to each other. The cable connectors are used in conjunction with a traditional static strain gauge for measuring the strain at four points in the pipe, or with a wireless strain acquisition module for transmitting strain and temperature data, ultimately achieving the purpose of strain data acquisition.

[0041] Reference Figure 1 As shown in the left figure, which is a cross-sectional view of a cylindrical long-distance pipeline, the four-point strain gauges on the pipeline are attached to the anti-corrosion layer on the outer wall of the pipeline by adhesive bonding. The four strain gauges on the four-point strain gauges correspond one-to-one with the four monitoring points on the pipeline, making positioning convenient and allowing for integrated direct installation. Compared with traditional methods, the installation is more convenient.

[0042] like Figure 3 As shown, the flexible printed circuit board (FPC) internally includes: strain gauge pre-drilled holes, thermistor pre-drilled holes, and FPC cables. The FPC cables are evenly distributed throughout the FPC to connect the strain gauges and thermistors at various locations, converging in the center of the FPC and exposing the cable connectors. The strain gauges and thermistors are connected to solder joints on the strain gauge and thermistor pre-drilled holes, respectively, by soldering. The strain gauge method is used to measure the magnitude of the strain generated by the component or structure under stress, and the thermistor method is used to measure the temperature of the component or structure during operation.

[0043] Strain gauge measurement is an important method in engineering used to measure the magnitude of strain in components or structures under stress. Its measurement principle is as follows: the resistance of a metal wire depends not only on the properties of the material but also on the length and cross-sectional area of ​​the wire. When the wire is attached to a component, its length and cross-sectional area change along with the component when the component deforms under stress, resulting in a change in resistance.

[0044] ;

[0045] in, The sensitivity coefficient of the material is the rate of change of resistance per unit strain, which indicates whether the resistance strain gauge effect of this type of wire material is significant. Strain at the measuring point is a dimensionless quantity, but it is conventionally given as unit microstrain, and is commonly represented by the symbol [symbol missing]. Therefore, we can calculate the deformation of the strain gauge by measuring its resistance. The measurement is simple; the strain gauge is simply attached to the surface of the test object.

[0046] ;

[0047] in, Measurement error caused by temperature changes The temperature coefficient of resistance of the material. The change in temperature The initial resistance of the strain gauge. The strain gauge sensitivity coefficient, For temperature strain, and These are the coefficients of thermal expansion of the strain gauge and the measuring material, respectively.

[0048] Therefore, it can be seen that the measurement error caused by temperature is linearly related to the amount of temperature change. This invention uses a thermistor to measure the temperature of a component or structure during operation, thereby correcting for measurement errors caused by temperature changes and further improving the accuracy and stability of strain monitoring.

[0049] This invention takes a pipeline with a diameter of 1016 mm as an example: its flexible printed circuit board (PCB) uses a polyimide (PI) substrate with a thickness of 0.2 mm, a length of 2500 mm, and a width of 50 mm, possessing excellent temperature resistance and chemical stability; the ribbon cable connector is 50 mm long, with a ribbon cable spacing of 1 mm at the connector; the strain gauge has a pre-drilled hole of 14 mm × 14 mm, and the thermistor has a pre-drilled hole of 6 mm × 20 mm. The strain gauge and thermistor are connected to the PCB by soldering and placed in the pre-drilled holes; the PCB is then pasted inside the anti-corrosion layer on the outer wall of the long-distance pipeline; the PCB is connected to the ribbon cable connector; and strain data is acquired by connecting a traditional static strain gauge or a wireless strain acquisition module through the ribbon cable connector.

[0050] The FPC surface is covered with a 0.3mm thick silicone potting protective layer, forming a flexible sealing structure that effectively blocks the corrosion of oil and gas media. Strain gauges and thermistors are connected to pre-drilled solder joints via laser welding or ultrasonic welding. The solder joints employ a gold-nickel composite plating process to ensure reliable electrical connections under long-term stress cycling and corrosive environments. The FPC surface is also covered with a two-component addition-type silicone potting protective layer, 0.3mm ± 0.05mm thick.

[0051] For pipes of different diameters, flexible printed circuit boards of different lengths can be manufactured. Strain gauges are placed in four equal parts in the strain gauge pre-drilled holes, and thermistors are placed in the central thermistor pre-drilled hole.

[0052] The technical parameters of the device can all be adapted to the pipe diameter and pipe type.

[0053] In a preferred embodiment, a low-power ASIC chip is integrated in the central region of the FPC. The chip incorporates a programmable gain amplifier (PGA), a 16-bit ADC, a digital filter, and a temperature compensation algorithm module. The ASIC connects to strain gauges and thermistors via I2C / SPI interfaces to achieve data acquisition at a sampling rate of 1kHz.

[0054] The temperature compensation algorithm is based on the strain gauge heat output mechanism and the theory of thermal expansion coefficient mismatch. Temperature errors in pipeline strain measurement mainly originate from three aspects.

[0055] (1) Temperature effect of strain gauge resistance

[0056] The resistivity of the sensitive gate material changes with temperature, resulting in a temperature coefficient of resistance. False strain caused :

[0057] ;

[0058] Where K is the strain gauge sensitivity coefficient, which is typically 2.0 ± 1%.

[0059] (2) Thermal expansion coefficient mismatch effect

[0060] False strain caused by the mismatch of thermal expansion coefficients between the strain gauge and the pipe steel :

[0061] ;

[0062] coefficient of linear expansion ≈11×10 -6 / °C, coefficient of thermal expansion ≈12×10 -6 / °C.

[0063] (3) Shear hysteresis effect between substrate and adhesive layer: Temperature change causes nonlinear shear strain transmission between polyimide substrate and steel. ≈28×10 -6 / °C, this effect exhibits second-order characteristics at large temperature differences (|ΔT|>40°C), and requires correction with quadratic terms.

[0064] The total temperature error is calculated by combining the above effects. Represented as:

[0065] ;

[0066] in, is the base shear transfer coefficient, with a value ranging from 0.02 to 0.05.

[0067] The substrate shear transfer coefficient λ was determined by the following method: a composite specimen consisting of FPC, adhesive layer, and tubing was prepared and placed in a high and low temperature test chamber. Temperature cycling was performed at 10°C intervals within the range of -20°C to 60°C, while simultaneously measuring the apparent strain caused by pure thermal expansion. Theoretical thermal strain of pipe Through formula Calculations were performed. For the 0.2mm polyimide substrate and epoxy adhesive layer used in this invention, actual measurements showed… The value is 0.03 to 0.04. This coefficient is pre-stored in the temperature compensation algorithm of the ASIC chip.

[0068] The health self-diagnosis function is achieved by monitoring the change rate of the bridge resistance value: when the resistance value changes by more than 50%, it is determined to be an open circuit; when it is lower than the normal value by 30%, it is determined to be a short circuit. The diagnosis cycle is 1 second, and the fault information is output through the reserved pin of the ribbon cable connector.

[0069] Preferably, the health self-diagnosis function can also be implemented through a dedicated integrated circuit, which includes a constant current source and an analog-to-digital converter to continuously monitor the resistance value R of each strain gauge bridge. When R is found to be greater than 150% of the initial value R0 and remains so for more than 100ms, an open circuit is identified, and fault code E1 is output; when R is less than 50% of R0, a short circuit is identified, and fault code E2 is output. The fault codes are output as serial digital signals through the reserved data pins of the ribbon cable connector.

[0070] Specifically, the bridge resistance values ​​of each strain gauge measured after sensor installation, under initial stress-free conditions or minimum load conditions within one pressure cycle, are used as the initial reference value R0. During subsequent operation, the system continuously records and updates the minimum resistance value R0 of each strain gauge within each acquisition cycle. minThis corresponds to the minimum strain state of the pipeline within that cycle, and is used as the dynamic reference benchmark for judging short circuits in the next diagnostic cycle.

[0071] When the bridge resistance R monitored in real time satisfies R>1.5 * R0 and continues for more than 3 diagnostic cycles, it is determined to be an open circuit, and fault code E1 is output. This condition mainly applies to physical disconnection of the line.

[0072] When the bridge resistance R monitored in real time satisfies R < 0.5 * R min If the short circuit persists for more than 3 diagnostic cycles, it is determined to be a short circuit, and fault code E2 is output. Here, R... min By using dynamically updated values, it is ensured that even if the pipeline undergoes significant normal deformation leading to a change in resistance, the current lowest value can still serve as a reasonable reference for judging short circuits, thus avoiding false alarms.

[0073] To differentiate between extreme strain and faults, when a short-circuit or open-circuit warning is triggered, the system simultaneously checks whether the changes in strain gauge readings at other locations along the same pipeline are consistent with the current pipeline pressure change trend. If only a single point of data is abnormal while the responses at other points are normal, the confidence level of the fault determination is increased.

[0074] The temperature compensation algorithm calculates the true strain based on the following formula. :

[0075] ;

[0076] in, To measure strain, K is the temperature coefficient of resistance of the strain gauge's sensitive grid material, and K is the strain gauge's sensitivity coefficient. The coefficient of thermal expansion of the pipe material. The coefficient of thermal expansion of the substrate material for flexible printed circuit boards. The temperature change measured by the thermistor. is the base shear transfer coefficient.

[0077] In summary, this device can be directly installed on the outer wall of the pipeline. It can be used with a traditional static strain gauge via a ribbon connector, or connected to an integrated wireless strain acquisition module to achieve simultaneous monitoring of the pipeline's axial, circumferential strain, and temperature. Compared with traditional methods, this invention is simpler to install, has lower manufacturing costs, and is ultra-thin, thus not affecting the coverage of the pipeline's external anti-corrosion layer. Furthermore, this device supports two installation modes, catering to the monitoring needs of the pipeline throughout its entire lifecycle—it can be pre-embedded on the pipe surface during the pipe manufacturing stage, constructed simultaneously with the anti-corrosion layer; or, during pipeline operation, for areas with potential safety hazards to be assessed, the anti-corrosion layer can be partially removed for rapid installation and data acquisition, enabling the assessment of the strain status of the operational pipeline.

[0078] To verify the linear response characteristics of the strain band, tensile tests were conducted. Standard specimens (7 mm thick, 32 mm wide tensile zone) were fabricated using Q235 steel plate. The specimens were loaded to 150 MPa at a rate of 0.5 MPa / s using a tensile testing machine, and the loading-unloading process was repeated four times. The obtained voltage-time curves are shown below. Figure 5 As shown, the linear fitting results of stress and voltage simulation signal changes are as follows: Figure 6 As shown. The fitting correlation coefficient R 2 The value >0.999 indicates a very strong linear correlation between the two, confirming that the strain band has a stable linear response and sensitivity, which can meet the requirements of high-precision strain measurement in engineering.

[0079] The strain change curves of strain monitoring equipment and strain gauges are as follows: Figure 7 As shown in Table 1, the relative error analysis table of strain monitoring data shows that when the stress is below 60 MPa, the relative error is less than 5%; when the stress rises to 75~150 MPa, the relative error drops to less than 1%, verifying that the strain band has high accuracy and reliability in actual measurement.

[0080] Table 1

[0081]

[0082] Example 2: Based on the flexible strain sensor in Example 1, this example integrates low-power wireless sensing, edge computing nodes and cloud-based intelligent analysis platform to build an intelligent diagnostic monitoring system.

[0083] The intelligent strain acquisition tape is attached to the anti-corrosion layer on the outer wall of the pipeline. Its four strain gauges continuously monitor the axial and circumferential strain of the pipeline, while two thermistors simultaneously monitor the temperature.

[0084] At the ribbon cable connector end, an ultra-low power edge AI chip (such as Nordicn RF5340 or Ambiq Apollo4 BlueLite, with a built-in ARM Cortex-M33 core and TinyML acceleration engine) is integrated to form an intelligent acquisition terminal.

[0085] MEMS accelerometers and acoustic emission sensors are used as edge AI chips, integrated onto a single flexible printed circuit board (FPC) via surface mount technology (SMT) or soldering. An ASIC chip serves as the intelligent acquisition terminal, connecting and managing these diverse sensors through onboard FPC cables. This terminal runs a quantized and optimized TinyML model (such as a lightweight neural network deployed using the TensorFlow Lite for Microcontrollers framework) locally, achieving: ① Real-time anomaly detection: Millisecond-level inference of strain time-series data using IsolationForest or 1D-CNN models, with a local anomaly detection rate >95%; ② Intelligent data compression: Only feature values ​​and statistics of abnormal events are uploaded, reducing data upload volume by over 90% under normal conditions; ③ Adaptive sampling: The sampling frequency is dynamically adjusted according to pipeline operating conditions (adjustable from 0.1Hz to 100Hz), further reducing power consumption. The model size is controlled within 100KB, and the inference latency is <10ms, meeting real-time requirements. As an edge computing node, this terminal can perform temperature compensation calculations on raw strain data in real time (using thermistor data to correct measurement errors caused by temperature changes in the strain gauge), and perform filtering, compression, and eigenvalue extraction (such as mean, variance, and peak value), significantly reducing the amount of data that needs to be uploaded.

[0086] Data synchronization can be achieved through unified time base and sampling control implemented using an ASIC chip. This ASIC chip enables data acquisition at a sampling rate of 1kHz. It possesses multi-channel acquisition capabilities and can provide a synchronized sampling clock for strain, temperature, vibration, and acoustic emission signals, ensuring data timestamp consistency at the hardware level. Data fusion is completed at the intelligent acquisition terminal or in the cloud, using algorithms to correlate multiple physical quantities at the same moment to achieve more accurate anomaly diagnosis.

[0087] The determination of strain gauge spacing needs to comprehensively consider the pipe diameter, stress distribution characteristics, and monitoring accuracy requirements. For cylindrical pressure pipes, stress concentration exists in the bend area, while the stress distribution is relatively uniform in the straight pipe section, requiring a differentiated spacing design.

[0088] In a specific embodiment, the radius of curvature R of the bend region is ≤ 6D, where D is the outer diameter of the pipe.

[0089] Four strain gauges are arranged at equal intervals along the circumference of the pipe, with a spacing L = πD / 4;

[0090] Taking a pipe with a diameter of 1016mm as an example: L=3.14×1016 / 4≈798mm;

[0091] The coverage angles are: each piece is spaced 90° apart, located at 0° (outer arc side), 90° (neutral plane), 180° (inner arc side), and 270° (neutral plane).

[0092] In the straight pipe region R>6D, non-equal spacing is adopted, and the strain gauges at both ends are close to the weld or geometric discontinuity area.

[0093] Spacing formula: L1 = 0.15L total L2 = 0.35L total L3 = 0.35L total L4 = 0.15L total .

[0094] Where L total The total length of the monitoring section is typically 1000-3000 mm.

[0095] It should be noted that the spacing ratio is determined based on finite element parametric analysis of typical oil and gas pipelines (such as X65 and X80 steel grades, with diameters ranging from 406mm to 1219mm). A finite element model of the pipeline-soil interaction is established to simulate the axial stress distribution under design pressure, temperature gradient, and typical uneven foundation settlement conditions. The impact of different sensor spacing schemes on the accuracy of reconstructed axial stress field is analyzed. Simulation results show that using the stated ratio can optimize the number of sensors and achieve effective coverage of stress concentration areas at both ends while ensuring overall monitoring accuracy. This ratio is suitable for monitoring general straight pipe sections. For specific geological conditions or special load situations, it can be adaptively adjusted based on specialized finite element analysis.

[0096] Data preprocessed at the edge is wirelessly transmitted to the cloud platform via a low-power wide-area network or cellular network. This solves the problems of complex deployment and difficulty in widespread deployment along long-distance pipelines associated with traditional wired transmission, enabling remote, real-time, and low-power backhaul of monitoring data.

[0097] The cloud platform receives and stores data from numerous intelligent data collection zones along the pipeline, forming spatiotemporal big data on pipeline strain.

[0098] By leveraging cloud computing power, machine learning and deep learning models are deployed to perform in-depth analysis of massive strain data. The models can automatically identify abnormal strain patterns, assess the health status of pipeline structures, and predict potential risks.

[0099] Based on pipeline GIS information, BIM models, and monitoring data, a high-fidelity digital twin of the pipeline is constructed in the cloud. This twin not only achieves dynamic visualization of strain and temperature fields, but more importantly, it integrates a finite element analysis (FEA) simulation engine (such as ANSYS or Abaqus cloud interface), using real-time strain data as boundary condition input for online stress reconstruction and fatigue life prediction. The system employs a surrogate model based on Physical Information Neural Network (PINN), reducing FEA calculation time from hours to seconds, achieving: ① Real-time stress field reconstruction: inverting the circumferential stress distribution of the pipeline based on four-point strain data; ② Fatigue life prediction: combining Miner's linear cumulative damage theory to predict the remaining life under different pressure and temperature cyclic loads, with the warning time window extended to 30-90 days; ③ Virtual sensing: in areas not covered by physical sensors, the strain state is calculated through the twin model, achieving full-pipeline digital monitoring. This realizes a leap from condition monitoring to predictive maintenance, guiding precise maintenance decisions.

[0100] Predicting the remaining life under cyclic loading at different pressures and temperatures relies on a finite element analysis (FEA) simulation engine within a digital twin. Real-time strain data from four monitoring points are used as boundary conditions and input into the FEA model to reconstruct the complete stress field at key pipe components, particularly stress concentration points. Rainflow counting analysis is performed on the reconstructed stress time history data at key points to extract cyclic information on stress amplitude and mean, thereby generating a load spectrum for fatigue calculations. A surrogate model based on a Physical Information Neural Network (PINN) accelerates this reconstruction process, enabling it to meet the timeliness requirements of online prediction.

[0101] Validation of the predictive model: Using historical monitoring data from known failure cases, verify whether the model can provide early warnings. Apply known cyclic loads to the experimental pipeline until fatigue damage occurs, and compare the model's predicted life with the actual life.

[0102] The system has set multiple warning thresholds. When AI analysis identifies anomalies or predicts risks exceeding the threshold, it automatically issues tiered warnings to managers via SMS, app push notifications, and other means. Managers can then formulate precise inspection or maintenance strategies based on cloud-based analysis reports and visualization results.

[0103] Specific Implementation: Intelligent Strain Monitoring in Pipeline Bends Based on Edge AI Terminals

[0104] A hot work operation of an X80 oil and gas pipeline in Southwest China was selected for on-site deployment and continuous monitoring. This case study focuses on examining the strain response characteristics, signal stability, and multi-channel synchronous monitoring performance of strain gauges under complex thermal disturbances to evaluate their feasibility and reliability for engineering applications. Strain gauges were deployed upstream and downstream of the target welding point, encircling the pipeline and covering four typical points (0°, 3°, 6°, and 9°, corresponding to the previously mentioned 0°, 90°, 180°, and 270°), simultaneously acquiring axial and circumferential strain signals at each point. Sensors were connected to a multi-channel strain acquisition device via leads to achieve real-time strain monitoring throughout the construction process. The on-site deployment of the strain gauges and equipment connection methods are as follows: Figure 8 As shown, the fatigue damage risk under periodic internal pressure fluctuations and seasonal temperature changes is assessed.

[0105] During on-site installation, the specific procedures are as follows:

[0106] Four monitoring points were selected on the outer wall of the pipe bend area. The anti-corrosion layer was partially peeled off, and the pipe surface was cleaned. Four strain gauges were then tightly adhered to the pipe surface using a special adhesive, ensuring precise alignment between the four strain gauges and the monitoring points. Each strain gauge contains two axial and one circumferential sensitive grids for comprehensive multi-directional strain capture. The data acquisition cable connectors were then connected to an IP67-rated intelligent data acquisition terminal, which incorporates an edge AI chip, an NB-IoT communication module, and a battery. Finally, the anti-corrosion layer was restored, and the terminal device was waterproofed and explosion-proof.

[0107] After the intelligent data acquisition terminal is powered on, it collects raw signals from each sensor according to a set cycle, uses pre-stored temperature compensation coefficients for real-time correction, extracts characteristic values ​​such as mean, variance, and peak value, and compares them with preset normal operating condition threshold ranges. If a characteristic value exceeds the threshold three times consecutively, the data segment is marked as abnormal and uploaded to the cloud platform wirelessly.

[0108] The cloud platform collects monitoring data from the initial phase (e.g., one month), combines it with historical pipeline operating pressure and temperature data, and trains an LSTM autoencoder model to learn the strain sequence pattern of the bend under normal operating conditions. After training, the model is deployed as an online anomaly detector.

[0109] The system runs continuously, with edge terminals uploading temperature-corrected strain characteristic data every 10 minutes. Upon receiving the data, the cloud platform synchronously updates the visualization status of the digital twin and inputs the continuous strain characteristic sequence into the LSTM autoencoder model for real-time reconstruction error calculation. If the reconstruction error continuously exceeds a set threshold, the system determines that the current strain mode deviates from the historical normal mode, automatically triggering a medium-level warning and indicating possible causes, such as a weakened correlation with pressure fluctuations or suspected local stiffness changes. Management personnel can access the historical strain curve, temperature curve, and AI diagnostic report for that point to decide whether to initiate a special on-site inspection.

[0110] Example 3: Intelligent Measurement Method for Pipeline Bends Based on Edge AI and Digital Twin

[0111] This embodiment addresses the fatigue damage risk faced by critical bend areas in oil and gas pipelines during long-term operation, providing an intelligent measurement method for pipeline bend areas based on edge AI and digital twins. This measurement method employs a four-layer technical architecture, deeply integrating current hot technologies in the fields of IoT and AI. The intelligent sensing layer (end-side) utilizes MEMS strain sensing technology, employing a high-precision foil strain gauge array. Each monitoring point is equipped with two axial sensitive grids and one circumferential sensitive grid to achieve multi-directional strain field capture; an integrated thermistor is used for synchronous temperature monitoring, providing a data foundation for temperature compensation; the acquisition strip uses a flexible substrate to adapt to the pipeline surface and has an IP68 protection rating.

[0112] The edge computing layer (edge ​​side) deploys intelligent acquisition terminals integrating NPUs (Neural Processing Units), achieving a computing power of 1-4 TOPS. TinyML technology is used to run lightweight machine learning models at the edge, enabling millisecond-level data preprocessing. Edge nodes have local model update capabilities, supporting distributed intelligent evolution. The wireless transmission layer (connectivity) adopts LPWAN technology with NB-IoT / LoRa dual-mode communication, balancing coverage and power consumption control. 5GuRLLC slicing technology can be enabled in critical pipe sections to meet ultra-low latency requirements. Huffman coding and differential compression algorithms are used to reduce the amount of transmitted data by more than 85%. The cloud intelligence layer (cloud side) uses InfluxDB / TDengine time-series databases to store massive amounts of strain time-series data. LSTM and Transformer time-series prediction models are built based on the TensorFlow / PyTorch deep learning framework. A digital twin engine is built by integrating GIS, BIM, and real-time simulation technologies to form a virtual pipeline mapping.

[0113] Specifically, such as Figure 3 As shown, the specific implementation steps of this measurement method are as follows:

[0114] Step 1: Deployment of monitoring points and installation of hardware.

[0115] First, site selection and planning were conducted. Based on historical pipeline failure data and stress analysis, four key monitoring points were selected in the bend area, covering the area of ​​maximum stress concentration and transition zone. Next, surface treatment was performed, with partial removal of the anti-corrosion layer and sandblasting or grinding to Sa2.5 cleanliness level to ensure reliable adhesion. Then, the data acquisition strips were installed, tightly adhering the four-point strain gauges to the pipeline surface using a special epoxy adhesive. This ensured that the positional error between the four strain gauges and the preset monitoring points was less than 2 mm, and the parallelism deviation between the axial sensitive grid and the pipeline axis was less than 1 degree, achieving precise alignment and directional calibration. Following this, the intelligent terminal was connected, with the data acquisition strip connector interfaced with the intelligent acquisition terminal. This terminal is equipped with a 480MHz MCU, integrating an ARM Cortex-M7 core and an AI acceleration unit. The power supply system uses a 10Ah lithium battery combined with a flexible solar film, supporting 7 days of continuous operation in cloudy or rainy weather. Finally, protection restoration was performed, restoring the anti-corrosion layer to ensure compatibility with the original anti-corrosion system. The terminal equipment uses an explosion-proof aluminum alloy shell, meeting the ExdIICT6 explosion-proof rating.

[0116] The power supply system adopts a multi-source energy harvesting and intelligent management architecture: ① Main power supply: 10Ah lithium thionyl chloride battery, operating in a wide temperature range of -40℃ to +85℃; ② Energy harvesting: Flexible amorphous silicon solar thin film (conversion efficiency >12%), combined with pipeline thermoelectric generators (TEG) to continuously generate electricity using the temperature difference between the pipeline transport medium and the environment (typically >20℃), with an output power >50mW; ③ Power management: Employing the MPPT algorithm to optimize energy harvesting efficiency, combined with a supercapacitor (10F / 5.5V) to achieve instantaneous power compensation, ensuring more than 7 days of continuous operation in rainy weather. The terminal supports low-power wide-area network (LPWAN) dual-mode communication: It defaults to using NB-IoT (Band 5 / 8) to achieve wide-coverage data transmission, automatically switching to LoRa (470MHz band, spread factor SF7-SF12 adaptive) in signal dead zones, and using self-organizing Mesh networking technology to achieve multi-node data relay, solving communication problems in complex terrain along the pipeline.

[0117] Step 2: Configure the edge intelligent preprocessing algorithm.

[0118] Edge computing algorithms are deployed on intelligent acquisition terminals to enable on-site data processing. A temperature compensation algorithm establishes a strain-temperature coupling model, acquiring ΔT in real-time via a thermistor, achieving a compensation accuracy of ±1 micro-strain. A digital filtering algorithm uses Kalman filtering to eliminate high-frequency noise, with a moving average window of 10 sampling points, improving the signal-to-noise ratio by 15 dB after filtering. A feature extraction algorithm calculates statistical features such as mean, standard deviation, peak value, and peak-to-peak value every 10 minutes, extracting the dominant frequency component in the 0.01-1 Hz band through FFT transformation, achieving a data compression ratio of 600:1 between the original and uploaded data. A lightweight IsolationForest quantized anomaly detection model is deployed for edge AI inference, with a model size of less than 50KB, inference latency of less than 5 milliseconds, and a local anomaly detection rate greater than 90%, effectively reducing invalid data transmission.

[0119] Step 3: Wireless data transmission and cloud aggregation.

[0120] The transmission protocol uses MQTT over NB-IoT, with a QoS level of 1 to ensure at least one delivery. Data packets are encapsulated in JSON format, containing device ID, timestamp, feature value, battery status, and other information. Normal operation is reported every 10 minutes, while abnormal operation is reported in real time. Network optimization employs an adaptive retransmission mechanism, enabling buffering and retransmission when the signal strength is below -110dBm. CRC32 checksum ensures data integrity, and the bit error rate is controlled below 10^-6. Cloud access adopts a microservice architecture, supporting tens of thousands of concurrent devices. Data stream processing uses an Apache Kafka message queue, with a peak processing capacity of 100,000 TPS.

[0121] The transmission protocol adopts an MQTT-SN over NB-IoT / LoRa dual-mode protocol stack, optimized for LPWAN networks: ① Protocol optimization: MQTT-SN simplifies the message header, reducing transmission overhead by 30%; supports QoS 0 / 1 / 2 three-level service quality, with critical early warning information using QoS 2 to ensure exactly-once delivery. ② Network adaptation: Real-time monitoring of RSSI and SNR; when the NB-IoT signal strength is <-110dBm or the packet loss rate is >5%, it automatically switches to LoRa mode and adjusts the spreading factor; supports edge node relay: when a node cannot directly connect to the base station, it uses multi-hop relay through neighboring nodes, with a network latency of <2s. ③ Secure transmission: DTLS 1.3 lightweight encryption and PSK pre-shared key authentication are used to ensure the confidentiality and integrity of data transmission. ④ Extremely low power consumption design: terminal deep sleep current <5μA, average data transmission current <50mA, achieving theoretical unlimited battery life in conjunction with the energy harvesting system.

[0122] Step 4: Cloud-based intelligent analysis and digital twins.

[0123] Data storage and management utilize TDengine time-series data partitioning, automatically sharded by monitoring point, and PostgreSQL for storing relational data such as device information and pipeline basic data. Raw feature data is stored for 5 years, and compressed archived data is stored for 10 years. AI model training and deployment are divided into two phases. Phase A is baseline model training, where strain time-series data under normal operating conditions is collected in the initial 1-3 months to construct an LSTM autoencoder network. The input layer consists of 24-dimensional features (4 points × 6-dimensional features / point), and the hidden layers are two LSTM layers with 128 units each. The output layer reconstructs the input sequence. The training objective is to minimize the reconstruction error MSE. Model validation is performed using an 80 / 20 split, with the validation set reconstruction error threshold set to μ+3σ. Phase B is online learning and optimization, where new data is used monthly to fine-tune model parameters. An attention mechanism is introduced to improve long-sequence modeling capabilities. When the validation error continues to rise, model drift detection is triggered, and retraining is performed. The anomaly diagnosis algorithm uses CNN to extract strain image features to identify typical failure modes such as corrosion, cracking, and buckling. It employs the Prophet algorithm to predict strain development trends, with an early warning window of 7-30 days. It also correlates pressure, temperature, and geological data and uses a Bayesian network to infer the cause of the anomaly. A digital twin visualization constructs a 1:1 3D model of the pipeline and integrates GIS coordinate information. The strain field is dynamically rendered using blue-green-yellow-red cloud maps to correspond to safety, caution, warning, and danger states. Based on a finite element model, the stress response under different internal pressures and temperature loads is simulated. WebGL 3D visualization technology is used, supporting a timeline playback and multi-view switching interactive interface.

[0124] Step 5: Intelligent early warning and decision feedback.

[0125] The system employs a multi-level early warning mechanism. The blue alert level is triggered when a single-point strain exceeds the historical average of 2σ, recording the event and increasing the monitoring frequency to 1 minute. The yellow warning level is triggered when the reconstruction error exceeds a threshold or shows an abnormal trend, automatically pushing notifications to on-duty personnel and generating a preliminary diagnostic report. The orange danger level is triggered when multiple points exhibit abnormal coordination or the predicted failure probability exceeds 30%, notifying management via SMS and telephone and initiating an emergency consultation. The red emergency level is triggered when a sudden strain change exceeds 50% of the yield strength, automatically triggering SCADA system linkage and preparing for an emergency shutdown. The decision support system generates an intelligent diagnostic report containing the location of the anomaly, possible causes, recommended measures, and historical similar cases. Based on a risk matrix, it outputs maintenance priorities and resource allocation suggestions, and uses a knowledge graph to link pipeline design parameters, construction records, and historical maintenance data to assist in root cause localization.

[0126] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made by those skilled in the art to the technical solutions of the present invention without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A method for measuring the deformation of an oil and gas transmission pipeline using a flexible strain sensor, characterized in that: The flexible strain sensor includes: A flexible printed circuit board, wherein the flexible printed circuit board is provided with four strain gauge reserved holes and two thermistor reserved holes; The strain acquisition band includes multiple strain gauges, which are respectively connected to the welding points on the corresponding pre-reserved holes of the strain gauges by welding. Each strain gauge includes multiple strain-sensitive grids, some of which are arranged along the axial direction of the pipe and others are arranged circumferentially along the pipe. Multiple thermistors are connected to solder joints on the thermistor pre-drilled holes by soldering and are symmetrically distributed about the centerline of the flexible printed circuit board; The flexible printed circuit board has a flexible printed circuit board ribbon cable inside, which connects the strain gauges and the thermistors at various locations and converges in the middle of the flexible printed circuit board before connecting to the ribbon cable connector. The flexible strain sensor is attached to the anti-corrosion layer of the outer wall of the pipe by adhesive bonding, so that the multiple strain gauges on the strain acquisition band correspond one-to-one with the multiple monitoring points on the outer wall of the pipe. The data acquisition device is connected via a ribbon cable connector to acquire the resistance change signal of the strain gauge caused by the pipe deformation, as well as the temperature signal acquired by the thermistor. Based on the temperature signal, a temperature compensation algorithm is used to correct the resistance change signal; The true strain at the pipeline monitoring point is calculated based on the corrected resistance change signal; The temperature compensation algorithm calculates the true strain based on the following equation : ； wherein, is the total temperature error, is the measured strain, is the temperature coefficient of resistance of the strain sensitive grid material, K is the strain gage sensitivity factor, is the coefficient of thermal expansion of the pipe material, is the coefficient of thermal expansion of the flexible printed circuit board base material, is the change in temperature measured by the thermistor, is the base shear transfer factor; Total temperature error It stems from the following three effects: (1) Temperature effect of strain gauge resistance: The resistivity of the strain-sensitive grid material changes with temperature, resulting in a temperature coefficient of resistance. False strain caused : ； (2) Thermal expansion coefficient mismatch effect: spurious strain caused by the mismatch of thermal expansion coefficients between the strain gauge and the pipe material. : ； (3) Shear hysteresis effect between substrate and adhesive layer: Temperature changes cause nonlinear shear strain transmission between the substrate and the pipe material, which is reflected in the thermal expansion coefficient of the flexible printed circuit board substrate material. .

2. The method according to claim 1, characterized in that, The attention level is triggered when the strain at a single point exceeds the historical average; the warning level is triggered when the reconstruction error exceeds the threshold or the trend is abnormal; the danger level is triggered when the multi-point coordination is abnormal or the predicted failure probability is greater than 30%; and the emergency level is triggered when the strain mutation exceeds 50% of the yield strength.

3. The method according to claim 1, characterized in that, The flexible printed circuit board integrates a microprocessor or application-specific integrated circuit for preliminary signal amplification, filtering, real-time temperature compensation, and sensor health self-diagnosis.

4. The method according to claim 3, characterized in that, The sensor health self-diagnosis includes: detecting strain gauge bridge open circuit or short circuit status, monitoring thermistor open circuit fault, judging the integrity of flexible printed circuit board ribbon cable connection, and outputting fault code through ribbon cable connector when a fault occurs.

5. The method according to claim 1, characterized in that, The flexible printed circuit board is made of polyimide substrate with a thickness of 0.2mm, a length customized according to the pipe diameter, and a width of 50mm.

6. The method according to claim 1, characterized in that, The multiple strain gauges are evenly distributed on the flexible printed circuit board, and the position of each strain gauge corresponds to a monitoring point on the outer wall of the pipe.

7. The method according to claim 6, characterized in that, The multiple strain gauges are arranged at equal intervals on the flexible printed circuit board.

8. The method according to claim 7, characterized in that, The plurality of strain gauges consists of four. When the flexible strain sensor is deployed on the pipeline, the four corresponding monitoring points are located at the 0° outer arc side, 90° neutral plane, 180° inner arc side and 270° neutral plane of the pipeline circumferential direction, respectively.

Images