A method for detecting integrity of a cast-in-place pile based on a heat conduction model

Abstract

The application discloses a cast-in-place pile integrity detection method based on a heat conduction model, and relates to the technical field of cast-in-place pile detection. The method comprises the following steps: firstly, a cast-in-place pile heat conduction characteristic template library containing standard heat conduction curves corresponding to various integrity states is constructed, and each standard curve contains multiple key temperature nodes; secondly, a comprehensive heat conduction curve is generated by fusing standard curves of the same integrity state; thirdly, actual temperature monitoring data of a cast-in-place pile to be detected are acquired, a corresponding comprehensive heat conduction curve is selected, and a reference conduction curve is extracted; fourthly, an abnormal conduction node in the reference conduction curve is labeled to obtain a modified conduction curve; fifthly, a heat conduction abnormality analysis model is established to check key temperature nodes of the modified conduction curve, a defect node is determined, and an integrity report containing defect node position information is generated to complete the detection. The application can cover various quality conditions of cast-in-place piles, integrate the advantageous features of the curves, and reduce detection errors and unnecessary links.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of cast-in-place pile detection technology, and particularly relates to a cast-in-place pile integrity detection method based on a heat conduction model. BACKGROUND

[0002] As a commonly used foundation component in construction engineering, the integrity of the cast-in-place pile is directly related to the stability and safety of the entire building structure, so the detection of the integrity of the cast-in-place pile has always been an important content in the control of engineering quality. At present, the commonly used cast-in-place pile integrity detection methods in the industry mainly include the ultrasonic transmission method, the low-strain reflected wave method and the core drilling method.

[0003] The ultrasonic transmission method needs to pre-bury a sounding pipe during the construction of the cast-in-place pile, and transmits and receives ultrasonic waves in the sounding pipe to determine whether there is a defect in the pile body according to the ultrasonic wave propagation speed, amplitude and waveform parameters. However, this method has high requirements for the installation accuracy of the sounding pipe, and if the sounding pipe is inclined, curved or blocked, it will directly affect the accuracy of the detection results; at the same time, for large-diameter cast-in-place piles, the ultrasonic waves are prone to attenuation during propagation, resulting in the appearance of a detection blind area and the inability to fully reflect the integrity of the pile body.

[0004] The low-strain reflected wave method is to apply a transient excitation on the top of the pile to generate a stress wave propagating along the pile body, and when the stress wave meets a defect in the pile body or the pile bottom, a reflected wave is generated, and the defect position and nature of the pile body are determined by analyzing the propagation time, amplitude and waveform characteristics of the reflected wave. However, this method is greatly affected by the performance of the excitation device, the uniformity of the pile material and the soil constraint conditions around the pile, and has weak recognition ability for shallow defects and small deep defects, often resulting in misjudgment or omission; in addition, when the cast-in-place pile has multiple defects, the reflected wave signals are superimposed on each other, making it difficult to accurately distinguish the specific position and size of each defect.

[0005] The core drilling method is to drill a core sample on the cast-in-place pile by a drilling device, and to determine the pile body quality by directly observing the integrity of the core sample. This method has intuitive and reliable detection results. However, the core drilling method is a destructive detection method, which will cause a certain degree of damage to the structural integrity of the cast-in-place pile, and has low detection efficiency and high cost, so it is only suitable for sampling detection and cannot be used for comprehensive detection of the entire cast-in-place pile; at the same time, the selection of the core drilling position is random, and if the defective part is not drilled, the detection result cannot truly reflect the overall integrity of the pile body. SUMMARY

[0006] The technical problems to be solved by the present application are: in order to solve the detection limitations, low efficiency, detection errors and other problems caused by single reference data in the prior art, a bored pile integrity detection method based on a heat conduction model is provided, which can cover all possible quality conditions of bored piles, effectively integrate the advantages of different standard curves under similar conditions, greatly improve the efficiency of preliminary identification of defects, and ensure the structural safety and engineering quality of bored piles.

[0007] In order to solve the above technical problems, the present application adopts the following technical solutions:

[0008] A bored pile integrity detection method based on a heat conduction model, comprising the following steps:

[0009] S1, a bored pile heat conduction feature template library is constructed, the bored pile heat conduction feature template library includes standard heat conduction curves corresponding to various integrity states, and each standard heat conduction curve includes a plurality of key temperature nodes.

[0010] S2, the standard heat conduction curves corresponding to similar integrity states are fused to generate a comprehensive heat conduction curve, and the comprehensive heat conduction curve includes a main conduction path and an auxiliary conduction path.

[0011] S3, actual temperature monitoring data of a bored pile to be detected is obtained, a corresponding comprehensive heat conduction curve is selected based on a preset integrity level, and a reference conduction curve is extracted in the comprehensive heat conduction curve based on temperature change characteristics.

[0012] S4, an abnormal conduction node is marked in the reference conduction curve to obtain a corrected conduction curve, and the abnormal conduction node is a key temperature node with a deviation between the actual temperature monitoring data and the reference conduction curve.

[0013] S5, a heat conduction anomaly analysis model is established, the key temperature nodes in the corrected conduction curve are checked using the model to obtain a defect node, an integrity report including position information of the defect node is generated, and the detection of the integrity of the bored pile is completed.

[0014] Preferably, in step S2, generating a comprehensive heat conduction curve includes the following contents:

[0015] The key temperature nodes in the standard heat conduction curve are provided with a conduction relationship identifier, the conduction relationship identifier assigns a serial number to each key temperature node, and the standard heat conduction curve is divided into a plurality of conduction intervals according to the serial number.

[0016] The standard heat conduction curves corresponding to similar integrity states are taken as fusion objects, based on the fusion objects, the standard heat conduction curve with the most conduction intervals is taken as a main curve, and the key temperature nodes in the main curve are taken as main nodes.

[0017] A reference node is determined in the other fusion objects except the main curve, and a key temperature node of each conduction interval of the other fusion objects is correspondingly fused into the main curve starting from the reference node, if the key temperature node and the main node satisfy a first correlation condition, the key temperature node is merged into the main node, otherwise the key temperature node is accessed into the main curve as an auxiliary node, and a comprehensive heat conduction curve is obtained.

[0018] Preferably, the determination of the reference node comprises the following contents:

[0019] The conduction interval with the smallest sequence number is located in the other fusion objects except the main curve, a key temperature node in the conduction interval is taken as an initial node, if a main node satisfying the first correlation condition with the initial node is found in the main curve, the initial node is determined as the reference node, otherwise, the key temperature nodes are sequentially selected from the other conduction intervals in the other fusion objects in the order from small to large sequence number, if a main node satisfying the first correlation condition with the key temperature node is found in the main curve, the key temperature node is determined as the reference node.

[0020] Preferably, the judgment of whether the first correlation condition is satisfied comprises the following contents:

[0021] A conduction parameter is configured for each key temperature node, and the conduction parameter comprises a temperature change gradient, a temperature change gradient of a previous node and a temperature change gradient of a subsequent node.

[0022] When the temperature change gradient of the main node is the same as that of the key temperature node, and the temperature change gradient of at least one of the previous node and the subsequent node is the same, it is determined that the main node and the key temperature node satisfy the first correlation condition.

[0023] Preferably, in step S3, the extraction of the reference conduction curve comprises the following contents:

[0024] The actual temperature monitoring data are processed in time sequence, discrete data are converted into continuous curves by using data fitting means, a target conduction curve is obtained, and a starting temperature node and an ending temperature node in the target conduction curve are located.

[0025] Key temperature nodes satisfying a second correlation condition with the starting temperature node and the ending temperature node in the comprehensive heat conduction curve are respectively located as a top node and a bottom node.

[0026] A plurality of candidate conduction curves are generated in the comprehensive heat conduction curve with the top node and the bottom node as end points, the goodness of fit of the target conduction curve and each candidate conduction curve is calculated, and the candidate conduction curve with the highest goodness of fit is taken as the reference conduction curve.

[0027] Preferably, the location of the starting temperature node and the ending temperature node comprises the following contents:

[0028] The target conduction curve comprises a plurality of characteristic temperature nodes, and whether each characteristic temperature node is a core characteristic point is determined in turn from both ends of the target conduction curve as a starting point.

[0029] If the key temperature node and the characteristic temperature node in the comprehensive heat conduction curve satisfy the second correlation condition, the characteristic temperature node is determined as a core characteristic point.

[0030] The core characteristic points determined first at both ends of the target conduction curve are defined as a starting temperature node and a terminal temperature node respectively.

[0031] Preferably, whether the second correlation condition is satisfied includes the following contents:

[0032] For each key temperature node, a conduction parameter is configured, and when the temperature variation gradient of a temperature node is the same as that of a key temperature node, and at least one of the temperature variation gradients of the previous node and the subsequent node is the same, it is determined that the temperature node and the key temperature node satisfy the second correlation condition.

[0033] Preferably, the calculation of the fitting degree includes the following contents:

[0034] Step 1, according to a preset characteristic sorting rule, the temperature variation gradient of the temperature node is integrated into a structured data form to obtain a corresponding feature vector; when calculating the feature vector similarity between the characteristic temperature nodes in the target conduction curve, the characteristic temperature node adjacent to the starting temperature node is taken as the first adjacent node; when calculating the feature vector similarity between the target conduction curve and the candidate conduction curve, the key temperature node adjacent to the top node is taken as the first adjacent node.

[0035] Step 2, similarity calculation is performed on the feature vectors of the first adjacent node and the starting temperature node to obtain a feature index of the starting temperature node; similarity calculation is performed on the feature vectors of the first adjacent node and the top node to obtain a feature index of the top node.

[0036] Step 3, any other adjacent temperature node of the first adjacent node is defined as a second adjacent node, and similarity calculation is performed on the feature vectors of the first adjacent node and the second adjacent node to obtain a feature index of the first adjacent node.

[0037] Step 4, repeat step 3 until all temperature nodes in the candidate conduction curve and the target conduction curve are traversed to obtain feature indexes of all temperature nodes.

[0038] Step 5, based on the feature indexes obtained in step 4, a first feature set and a second feature set corresponding to the target conduction curve and the candidate conduction curve are generated respectively, and the fitting degree is obtained according to the number of same feature indexes in the first feature set and the second feature set.

[0039] Preferably, in step S5, the conduction parameters of the key temperature nodes in different integrity states of the cast-in-place pile are taken as training samples, the training samples are preprocessed to obtain preprocessed training samples, a heat conduction anomaly analysis model is constructed based on a deep learning framework, the preprocessed training samples are used to train the model, and a trained heat conduction anomaly analysis model is obtained; the conduction parameters of the key temperature nodes in the corrected conduction curve are input into the trained heat conduction anomaly analysis model, the sequence relationship of the parameters is processed, and an anomaly index of the key temperature nodes is obtained; when the anomaly index exceeds a preset threshold value, the key temperature node corresponding to the anomaly index is determined as a defect node.

[0040] Preferably, the method further comprises a real-time monitoring step:

[0041] The temperature field distribution data of the cast-in-place pile to be detected is continuously collected, and the real-time monitoring mechanism is triggered when the temperature variation amplitude of the temperature field distribution data exceeds a set limit value; the heat conduction characteristics of the abnormal change region are obtained and updated to the corresponding standard heat conduction curve in the cast-in-place pile heat conduction characteristic template library; steps S1-S5 are re-executed based on the updated cast-in-place pile heat conduction characteristic template library to generate an updated integrity report.

[0042] Compared with the prior art, the above technical scheme has the following technical effects:

[0043] 1. The cast-in-place pile heat conduction characteristic template library containing standard heat conduction curves corresponding to multiple integrity states is constructed, providing rich and comprehensive reference for subsequent detection work. The standard heat conduction curves under different integrity states can cover various quality conditions of the cast-in-place pile, avoiding the detection limitations caused by single reference data, and making the detection process more targeted and scientific.

[0044] 2. The standard heat conduction curves corresponding to similar integrity states are fused to generate a comprehensive heat conduction curve containing a main conduction path and an auxiliary conduction path, effectively integrating the advantages of different standard curves under similar states. The main conduction path can reflect the main law of heat conduction under this type of integrity state, and the auxiliary conduction path can supplement the detailed changes in the heat conduction process. The combination of the two makes the comprehensive heat conduction curve more accurately represent the heat conduction characteristics of the cast-in-place pile under the corresponding integrity state, reduces the deviation of a single standard curve, and lays a more reliable foundation for subsequent comparison and analysis with actual monitoring data.

[0045] 3、The application can realize accurate adaptation of the detection process after obtaining actual temperature monitoring data of the to-be-detected bored pile, based on the preset integrity level selection corresponding comprehensive heat conduction curve, and combining the temperature change characteristic to extract the reference conduction curve. The division of the preset integrity level makes the detection work quickly locate to the corresponding reference curve according to the preliminary judgment result of the bored pile, avoids unnecessary curve screening link, and improves the detection efficiency; at the same time, the reference conduction curve is extracted combined with the actual temperature change characteristic, which further shortens the distance between the reference curve and the actual situation of the to-be-detected bored pile, and reduces the detection error caused by the inconsistency between the reference curve and the actual situation.

[0046] 4、The application marks the abnormal conduction node in the reference conduction curve, and clearly shows the key temperature node of the deviation between the actual temperature monitoring data and the reference curve, so that the identification of the abnormal condition of the bored pile in the detection process is more targeted. Through accurate positioning of the abnormal node, the area where the defect may exist can be quickly locked, avoiding the tedious process of full investigation of the whole pile body, greatly improving the efficiency of preliminary identification of defects, and providing a clear direction for subsequent in-depth analysis of defect conditions.

[0047] 5、The application establishes a heat conduction abnormality analysis model, obtains a corrected conduction curve after adjusting the abnormal conduction node, and checks each key temperature node in the corrected conduction curve, marks the nodes that fail the check as defect nodes, and finally generates an integrity report containing defect node position information, forming a complete and rigorous detection process. The application of the heat conduction abnormality analysis model can analyze the nature and influence range of the abnormal node, obtain the corrected conduction curve by adjusting the abnormal node, and further improve the accuracy of the detection result; the checking link of the key temperature node can effectively filter out abnormal data caused by accidental factors, and ensure that the marked defect node is real and reliable; the integrity report containing defect position information finally generated can provide clear and intuitive detection results for engineering personnel, so that the engineering personnel can take targeted treatment measures in time, and ensure the structural safety and engineering quality of the bored pile. BRIEF DESCRIPTION OF DRAWINGS

[0048] Figure 1 is the overall implementation flowchart of the application.

[0049] Figure 2 is the implementation flowchart of the application to obtain a comprehensive heat conduction curve.

[0050] Figure 3 is the implementation flowchart of the application to obtain a reference conduction curve.

[0051] Figure 4 is a heat conduction analysis comparison chart in the embodiment of the application. DETAILED DESCRIPTION

[0052] The application will be further described below with reference to the drawings. The following examples are only used to more clearly illustrate the technical solutions of the application, and cannot be used to limit the protection scope of the application.

[0053] To achieve the above object, the application provides a cast-in-place pile integrity detection method based on a heat conduction model, as shown in the following steps: Figure 1

[0054] S1, a heat conduction feature template library of cast-in-place piles is constructed, the heat conduction feature template library of cast-in-place piles includes standard heat conduction curves corresponding to multiple integrity states, and each standard heat conduction curve includes multiple key temperature nodes, which represent important turning points in the temperature change process.

[0055] S2, as shown in the following, standard heat conduction curves corresponding to similar integrity states are fused to generate a comprehensive heat conduction curve, and the comprehensive heat conduction curve includes a main conduction path and an auxiliary conduction path. Specifically: Figure 2

[0056] The key temperature nodes in the standard heat conduction curve are provided with a conduction relationship identifier, the conduction relationship identifier is established by analyzing the temperature gradient change, time interval and spatial position relationship between the nodes, for example, the sequence of the nodes is defined by a time stamp sequence or a temperature change inflection point, the conduction relationship identifier assigns a sequence number to each key temperature node, the sequence number reflects the sequence and position relationship of the nodes in the heat conduction process, and the standard heat conduction curve is divided into multiple conduction intervals according to the continuity of the sequence number and the conduction characteristics between the nodes, each conduction interval represents a continuous heat conduction stage, and the nodes in the interval have consistent conduction characteristics and temperature change modes, for example, a conduction interval can correspond to a uniform heating or cooling stage.

[0057] The standard heat conduction curves corresponding to similar integrity states are taken as fusion objects, based on the fusion objects, the standard heat conduction curve with the most conduction intervals is taken as a main curve, the main curve contains the most abundant heat conduction modes and the most key temperature nodes, and can provide a more comprehensive reference framework, and the key temperature nodes in the main curve are taken as main nodes.

[0058] ​​locating the conduction interval with the smallest sequence number in the other fusion objects except the main curve, taking the key temperature node in the conduction interval as an initial node, if a main node satisfying the first correlation condition with the initial node is found in the main curve, the initial node is determined as a reference node; when the initial node cannot be matched, key temperature nodes are selected from other conduction intervals in the other fusion objects in ascending order of sequence number, if a main node satisfying the first correlation condition with the key temperature node is found in the main curve, the key temperature node is determined as the reference node. Taking the reference node as a starting point, the key temperature nodes of each conduction interval of the other fusion objects are fused into the main curve, if the key temperature node satisfies the first correlation condition with the main node, the key temperature node is merged into the main node, otherwise the key temperature node is accessed to the main curve as an auxiliary node, to obtain a comprehensive heat conduction curve.

[0059] The judgment of whether the first correlation condition is satisfied includes the following contents:

[0060] A conduction parameter is configured for each key temperature node, the conduction parameter including a temperature change gradient, a temperature change gradient of a previous node, and a temperature change gradient of a subsequent node.

[0061] When the temperature change gradient of the main node is the same as that of the key temperature node, and the temperature change gradient of at least one of the previous node and the subsequent node is the same, it is determined that the main node and the key temperature node satisfy the first correlation condition.

[0062] The main conduction path is formed based on the main curve, the main nodes constitute the core framework of the main conduction path, and can reflect the main rules and typical characteristics of heat conduction of the cast-in-place pile under the integrity state, and is the most representative conduction path in the heat conduction process; the auxiliary conduction path is a supplementary conduction path formed after the auxiliary nodes are accessed to the main curve, and can supplement the non-typical and detailed conduction modes that may appear in the heat conduction process under the integrity state, and together with the main conduction path, covers various possible heat conduction conditions of the cast-in-place pile under the integrity state.

[0063] The comprehensive heat conduction curve forms a directed graph structure, in which the main nodes constitute the core conduction path, and the auxiliary nodes serve as the supplementary path, enhancing the expression ability and adaptability of the curve. This fusion method avoids information loss, while maintaining the logical consistency and structural integrity of the curve, and provides a reliable basis for subsequent extraction of the reference conduction curve and anomaly detection.

[0064] The entire fusion process can be automated through algorithms. These algorithms match and merge nodes according to preset rules and parameters, ensuring the consistency and repeatability of the results. During the fusion process, conflicts and inconsistencies between nodes also need to be addressed. For example, when multiple fusion objects propose different fusion schemes for the same master node, a weighted average or majority voting mechanism is used to resolve conflicts, ensuring the rationality and representativeness of the fusion result. Furthermore, for special transmission patterns or abnormal nodes, additional processing rules can be set, such as separate marking or priority retention, to handle complex situations.

[0065] S3, such as Figure 3 As shown, the actual temperature monitoring data of the cast-in-place pile to be tested is obtained. Based on a preset integrity level, a corresponding comprehensive heat conduction curve is selected. A benchmark conduction curve is extracted from the comprehensive heat conduction curve based on temperature change characteristics. Specifically:

[0066] Temperature sensors are installed at preset intervals along the body of the pile to be tested, covering different depth areas of the pile. Actual temperature monitoring data inside the pile concrete is continuously collected. This data is preprocessed to remove abnormal data (such as instantaneous temperature jumps) caused by temporary sensor malfunctions or environmental interference, retaining valid data that reflects the temperature change over time or depth to ensure the smoothness and representativeness of the curve. The preprocessed actual temperature monitoring data is then processed according to time series or depth series, and data fitting techniques are used to transform discrete data into a continuous curve, obtaining the target conduction curve. The starting and ending temperature nodes in the target conduction curve are then located. Specifically:

[0067] The target conduction curve includes multiple characteristic temperature nodes. Starting from both ends of the target conduction curve, we sequentially determine whether each characteristic temperature node is a core characteristic point.

[0068] If a key temperature node and a characteristic temperature node in the comprehensive heat conduction curve satisfy the second correlation condition, then the characteristic temperature node is determined as the core characteristic point.

[0069] The core feature points first determined at both ends of the target conduction curve are defined as the starting temperature node and the ending temperature node, respectively.

[0070] In the integrated heat conduction curve, the key temperature nodes that satisfy the second correlation condition with the starting temperature node and the ending temperature node are respectively located and designated as the top node and the bottom node.

[0071] The determination of whether the second association condition is met includes the following:

[0072] The conduction parameters are configured for each key temperature node, and when the temperature node has the same temperature change gradient as the key temperature node and at least one of the preceding node and the subsequent node has the same temperature change gradient, it is determined that the temperature node and the key temperature node meet the second correlation condition.

[0073] A plurality of candidate conduction curves are generated in the comprehensive heat conduction curve with the top node and the bottom node as end points, the generation of the candidate conduction curve is to find all possible conduction paths by a path search algorithm using the graph structure characteristics of the comprehensive heat conduction curve. The goodness of fit of the target conduction curve and each candidate conduction curve is calculated, and the goodness of fit is evaluated based on the similarity comparison of the feature vectors. The candidate conduction curve with the highest goodness of fit is taken as the reference conduction curve; the specific content is:

[0074] Step 1, according to the preset feature sorting rule, the temperature change gradient of the temperature node is integrated into a structured data form to obtain the corresponding feature vector, the feature vector includes temperature gradient, time attribute and spatial position information; when calculating the feature vector similarity between the feature temperature nodes in the target conduction curve (quantified by cosine similarity or Euclidean distance algorithm), the feature temperature node adjacent to the starting temperature node is taken as the first adjacent node; when calculating the feature vector similarity between the candidate conduction curve and the target conduction curve, the key temperature node adjacent to the top node is taken as the first adjacent node;

[0075] Step 2, the feature vectors of the first adjacent node and the starting temperature node are calculated to obtain the feature index of the starting temperature node; the feature vectors of the first adjacent node and the top node are calculated to obtain the feature index of the top node;

[0076] Step 3, define any other adjacent temperature node of the first adjacent node as the second adjacent node, and calculate the feature vectors of the first adjacent node and the second adjacent node to obtain the feature index of the first adjacent node;

[0077] Step 4, repeat step 3 until all temperature nodes in the candidate conduction curve and the target conduction curve are traversed to obtain the feature index of all temperature nodes;

[0078] Step 5, based on the feature index obtained in step 4, the first feature set and the second feature set corresponding to the target conduction curve and the candidate conduction curve are generated respectively, and the goodness of fit is obtained according to the number of the same feature index in the first feature set and the second feature set.

[0079] S4, marking the abnormal conduction node in the reference conduction curve to obtain a modified conduction curve, the abnormal conduction node is a key temperature node with deviation between the actual temperature monitoring data and the reference conduction curve.

[0080] S5, a thermal conduction anomaly analysis model is established, the model is used to check the key temperature nodes in the corrected conduction curve, and a defect node is obtained; a completeness report including the position information of the defect node is generated, and the detection of the integrity of the cast-in-place pile is completed. Specifically:

[0081] The conduction parameters of the key temperature nodes under different integrity states (including normal state and various defect states) of the cast-in-place pile are taken as training samples, the training samples are preprocessed to ensure the effectiveness and consistency of the data, and preprocessed training samples are obtained; a thermal conduction anomaly analysis model is constructed based on a deep learning framework, the preprocessed training samples are used to train the model, the model parameters are iteratively optimized, the model can accurately learn the characteristic rules of the conduction parameters of the key temperature nodes under different integrity states, and the performance of the model is verified using independent verification samples to ensure that the model can accurately output the anomaly index of the key temperature nodes, meet the determination requirements of the key temperature nodes in the cast-in-place pile integrity detection, and obtain a trained thermal conduction anomaly analysis model; the conduction parameters of the key temperature nodes in the corrected conduction curve are input into the trained thermal conduction anomaly analysis model, and the sequence relationship of the parameters is processed to realize effective analysis of the abnormal conditions of the key temperature nodes, and the anomaly index of the key temperature nodes is obtained; when the anomaly index exceeds a preset threshold value, the key temperature node corresponding to the anomaly index is determined as a defect node.

[0082] S6, the temperature field distribution data of the cast-in-place pile to be detected is continuously collected, the sensor records the temperature values at different depths in the concrete at a specific sampling frequency, and the data is sent to the central processing unit through wired or wireless transmission. When abnormal changes in the temperature field distribution data are detected, a real-time monitoring mechanism is triggered, and the determination of abnormal changes is based on temperature change rate threshold and spatial continuity analysis, for example, the temperature change amplitude of a certain region in a unit time exceeds the set limit value and the adjacent sensors show similar trends; the temperature gradient distribution, heat conduction rate and temperature change mode of the abnormal change region are obtained and updated to the corresponding standard heat conduction curve in the cast-in-place pile heat conduction characteristic template library; based on the updated cast-in-place pile heat conduction characteristic template library, steps S1-S5 are re-executed to generate an updated integrity report.

[0083] Embodiment 1:

[0084] The present application detects the integrity of a cast-in-place pile with a diameter of 1.0 meters and a depth of 22 meters in a certain bridge engineering, collects 5 standard heat conduction curves of the same type and the same integrity state (preliminarily determined as a slight defect state), each curve includes multiple key temperature nodes and has a conduction relationship identifier between the nodes.

[0085] Firstly, the serial numbers are assigned to each key temperature node based on the conduction relationship, and each standard heat conduction curve is divided into different conduction intervals according to the numbers. It is found through statistics that one of the standard heat conduction curves contains 8 conduction intervals, which is more than the other 4 curves (containing 5, 6, 7, and 6 conduction intervals respectively), so the standard heat conduction curve containing 8 conduction intervals is determined as the main curve, and the key temperature nodes thereof are determined as the main nodes.

[0086] When determining the reference nodes in the remaining 4 standard heat conduction curves, the conduction interval with the smallest serial number in each standard heat conduction curve is first located, and the key temperature node in the interval is extracted as the initial node. The initial node is subjected to a first association condition judgment with the main nodes of the main curve. The temperature change gradient of the initial node of one of the fusion objects is the same as that of a certain main node of the main curve, and the temperature change gradient of the previous node is also consistent, satisfying the first association condition. Therefore, the initial node is determined as the reference node. Taking the reference node as the starting point, the key temperature nodes of each conduction interval in the standard heat conduction curve are correspondingly fused into the main curve. For the key temperature nodes that satisfy the first association condition with the main nodes, they are merged into the main nodes; otherwise, they are connected to the main curve as auxiliary nodes. In this way, the fusion of the remaining 3 standard heat conduction curves with the main curve is sequentially completed, and finally a comprehensive heat conduction curve containing the main conduction path and the auxiliary conduction path is generated, which is used for the subsequent reference conduction curve extraction of the to-be-detected cast-in-place pile.

[0087] Embodiment 2:

[0088] As shown in Figure 4 , the present application carries out integrity detection on a cast-in-place pile with a diameter of 1.2 meters and a depth of 28 meters in a certain bridge engineering. Twenty temperature sensors are embedded in the pile body of the cast-in-place pile, and the temperature sensors are uniformly distributed along the pile body at an interval of 2 meters. Temperature monitoring data is collected 72 hours after the completion of concrete pouring, and an actual monitored target conduction curve is obtained. The curve shows that there is abnormal temperature fluctuation at a depth of 16 meters. The conduction interval with the smallest serial number is located, and the interval corresponds to the depth range of 0-4 meters at the top of the pile. The key temperature node T0 is taken as the initial node, and the key temperature node T0 has a temperature change gradient of 0.12℃ / h and a time constant of 4.5 hours. In the main curve, a key temperature node with similar characteristics to the key temperature node T0 is found. The temperature change gradient of the starting node M0 of the main curve is 0.11℃ / h, and the time constant is 4.6 hours. The difference between the two is within the allowable error range, so the key temperature node T0 is determined as the reference node.

[0089] The target conduction curve contains 25 characteristic temperature nodes, which are located by scanning from both ends of the pile top and pile bottom: when scanning from the pile top downward, a characteristic temperature node P1 is found at a depth of 2.1 meters, with a temperature value of 28.5°C and a temperature change gradient of 0.13°C / h, which matches the key temperature node C1 in the comprehensive thermal conduction curve; when scanning from the pile bottom upward, a characteristic temperature node P25 is located at a depth of 26.8 meters, with a temperature value of 31.2°C and a temperature change gradient of 0.09°C / h, which matches the key temperature node C20 in the comprehensive thermal conduction curve. Therefore, the characteristic temperature node P1 and the characteristic temperature node P25 are determined as the starting temperature node and the ending temperature node, and the key temperature node C1 and the key temperature node C20 are defined as the top node and the bottom node.

[0090] Three alternative conduction curves are generated in the comprehensive thermal conduction curve with the key temperature nodes C1 and C20 as endpoints: the first alternative conduction curve passes through nodes C1-C4-C7-C10-C13-C16-C20, the second alternative conduction curve passes through C1-C3-C6-C9-C12-C15-C18-C20, and the third alternative conduction curve passes through C1-C2-C5-C8-C11-C14-C17-C19-C20.

[0091] Taking the first alternative conduction curve as an example, when calculating the goodness of fit between the target conduction curve and each alternative conduction curve, starting from the key temperature node C1, the similarity of the feature vector between it and the starting temperature node P1 is calculated as 0.87, the similarity of the feature vector between the key temperature node C1 and its adjacent key temperature node C4 is calculated as 0.92, and the calculation continues for the key temperature node C4 and its adjacent key temperature node C7, and so on until the bottom node C20. In the second alternative conduction curve, the key temperature node C15 shows a lower similarity of the feature vector with the corresponding characteristic temperature node in the target conduction curve, only 0.63, because the temperature gradient value at this key temperature node has a large deviation from the actual monitoring data. In the third alternative conduction curve, the time stamp feature of the key temperature node C19 does not match the actual monitoring data, resulting in a decrease in the overall goodness of fit of this path. After comprehensive comparison, the overall goodness of fit of the first alternative conduction curve reaches 0.89, which is higher than the 0.76 and 0.71 of the other two alternative conduction curves, so the first alternative conduction curve is determined as the reference conduction curve.

[0092] In the process of determining the reference node, when the initial node does not meet the first correlation condition, the search will continue in the direction of increasing sequence number of the conduction interval. For example, if the key temperature node T0 fails to match successfully, the key temperature node T1 in the 4-8 meter interval will be checked. The node temperature variation gradient is 0.15℃ / h, and the time constant is 5.2 hours. In the main curve, the key temperature node M1 has similar characteristics, with a temperature variation gradient of 0.14℃ / h and a time constant of 5.1 hours. Therefore, the key temperature node T1 is determined as the reference node. This process continues until a suitable matching node is found, ensuring that the reference node for the fusion operation is accurate and reliable.

[0093] In actual monitoring, the target conduction curve may contain abnormal points caused by measurement noise or temporary environmental factors. These data are processed by sliding window average filtering, and the window size is set to 5 data points according to the sampling frequency and sensor characteristics. When positioning the starting temperature node and the ending temperature node, the tolerance range of feature matching is set, with a temperature value deviation of ±0.3℃, a temperature variation gradient deviation of ±0.02℃ / h, and a time parameter deviation of ±0.5 hours, to accommodate unavoidable measurement errors.

[0094] The entire processing process establishes a log record, saving all intermediate calculation results and decision basis for subsequent traceability analysis and optimization. Through Example 2, the temperature abnormal area at a depth of 16 meters is successfully identified.

[0095] Example 3:

[0096] The present application carries out integrity detection on a bored pile with a diameter of 1.5 meters and a depth of 30 meters in a certain bridge engineering. When acquiring actual temperature monitoring data of the bored pile to be detected, 20 temperature sensors are arranged along the pile body at an interval of 1.5 meters, and temperature data is continuously collected for 72 hours. After pre-processing and removing abnormal data, a target conduction curve is generated by fitting. Then, the starting temperature node and the ending temperature node of the target conduction curve are located, and it is judged in turn from the inside of the curve at both ends whether the characteristic temperature node is a core feature point. The first characteristic temperature node that meets the second correlation condition with the key temperature node of the comprehensive heat conduction curve is found at a distance of 2.1 meters from the top of the pile, and the characteristic temperature node is determined as the starting temperature node. The first characteristic temperature node that meets the second correlation condition with the key temperature node of the comprehensive heat conduction curve is found at a distance of 1.8 meters from the bottom of the pile, and the characteristic temperature node is determined as the ending temperature node. Then, the key temperature nodes that meet the second correlation condition with the starting temperature node and the ending temperature node in the comprehensive heat conduction curve are located respectively as the top node and the bottom node. Taking the top node and the bottom node as endpoints, four alternative conduction curves are generated in the comprehensive heat conduction curve. By calculating the degree of coincidence of the target conduction curve and each alternative conduction curve, it is found that one of the alternative conduction curves has the highest degree of coincidence, which is determined as the reference conduction curve, laying a foundation for subsequent abnormal conduction node labeling.

[0097] In the process of locating the starting temperature node and the ending temperature node, the multiple characteristic temperature nodes contained in the target conduction curve need to be systematically analyzed. These characteristic temperature nodes are significant change points extracted from actual monitoring data, including extreme points, inflection points or mutation points.

[0098] The determination of the second correlation condition adopts a parameterized comparison method. The temperature change gradient is obtained by the ratio of the temperature difference between adjacent nodes to the time difference. The calculation formula is:

[0099]

[0100] Among them, represents the temperature change gradient, represents the adjacent node temperature difference, represents the corresponding time interval.

[0101] The temperature change gradient of the previous node refers to the gradient calculation value between the current key temperature node and the previous key temperature node, and the temperature change gradient of the subsequent node refers to the gradient calculation value between the current key temperature node and the next key temperature node. In the comparison process, the temperature change gradient difference between the main node and the reference node needs to be less than the error tolerance , and at least one of the temperature change gradient difference of the previous node and the subsequent node needs to be less than the error tolerance , in order to meet the second correlation condition. The error tolerance The value of the gradient is determined according to the sensor accuracy and environmental fluctuation characteristics, and is usually set by statistically determining the distribution range of the gradient value in historical data to ensure matching robustness.

[0102] After the start temperature node and the end temperature node are located, the position information of the nodes is recorded and used for subsequent candidate conduction curve generation. The position information of the nodes includes depth position and timestamp information, ensuring consistency in spatial and temporal dimensions.

[0103] The entire positioning process is automatically executed by an algorithm, and the algorithm has a built-in verification mechanism to prevent false positioning, such as checking the continuity and logical reasonableness of the node sequence. All matching results and parameter settings are saved to a configuration file for subsequent auditing and optimization. The management of conduction parameters uses a structured storage method, and each node's parameter set includes gradient value, timestamp, spatial coordinates, and adjacent node reference, forming a complete node relationship network. The parameter update mechanism allows dynamic adjustment of the gradient value based on new monitoring data, maintaining the timeliness of the parameter set. The logic of the association judgment is encapsulated as an independent module, which can adapt to the needs of different engineering scenarios by adjusting threshold values and algorithm parameters. The accuracy of the positioning results directly affects the quality of the reference conduction curve, so multiple verification steps are set, including node sequence integrity check, conduction logic consistency verification, and historical data backtracking comparison. The final output of the start temperature node and the end temperature node serves as a key anchor point, providing a stable reference framework for the entire integrity detection process.

[0104] Example 4:

[0105] The present application carries out integrity detection on a bored pile with a diameter of 0.8 meters and a depth of 18 meters in a certain residential district construction project. In the process of locating the start temperature node and the end temperature node, the top end of the pile (one end of the curve) of the target conduction curve is first analyzed, and each characteristic temperature node is analyzed in turn. Take the first characteristic temperature node, check if there is a key temperature node in the comprehensive heat conduction curve that meets the second association condition with the characteristic temperature node. After comparison, the temperature change gradient of the characteristic temperature node is different from that of a certain key temperature node in the comprehensive heat conduction curve, and the second association condition is not met. Then the next characteristic temperature node is analyzed. The temperature change gradient of the characteristic temperature node is 0.13℃ / h, and the temperature change gradient of a certain key temperature node in the comprehensive heat conduction curve is also 0.13℃ / h, and the temperature change gradients of the subsequent nodes of both are 0.11℃ / h, which meets the second association condition. The characteristic temperature node is determined as the start temperature node. Then, starting from the bottom end of the pile (the other end of the curve) of the target conduction curve, the characteristic temperature nodes are analyzed in the same way, and the first characteristic temperature node that meets the second association condition is found and determined as the end temperature node, providing an accurate endpoint for subsequent reference conduction curve extraction. The feature vector of the key temperature node is obtained, and the specific data is shown in Table 1.

[0106] Table 1 Eigen vector of key temperature nodes

[0107]

[0108] In the embodiment, the generation of the eigen vector adopts a standardization processing mode, the temperature change gradient is normalized to the range of 0-1, the timestamp is converted into a relative time difference value, and the node type is represented by classification coding. The eigen vector similarity is calculated using an improved cosine similarity algorithm, taking into account numerical similarity and sequence correlation. The comparison of the first eigen vector set and the second eigen vector set adopts a fast matching algorithm based on hash values to improve the calculation efficiency.

[0109] In the training process of the heat conduction anomaly analysis model, the preprocessed training samples are input into the heat conduction anomaly analysis model, transmitted through the input layer, processed in multiple convolution layers and fully connected layers, and finally output the anomaly index through the sigmoid activation function. The cross-entropy loss function and the Adam optimizer are used to optimize the model parameters through multiple iterations.

[0110] The entire processing process of the defect node is performed using a parallel computing architecture, and the eigen vector calculation and similarity comparison of multiple nodes are processed simultaneously. A node feature database is maintained to store historical calculation results for accelerating subsequent processing. For newly detected node types, their eigen vectors are added to the node feature database to continuously improve the reference data set. The model inference result is verified by a post-processing module, including node sequence integrity check, spatial consistency verification and timing logic audit; only the defect nodes that pass all the verifications will be finally confirmed and included in the integrity report. A visualization interface is also provided to display the intermediate results in the calculation process, including eigen vector distribution, similarity matrix and anomaly index curve, to facilitate manual review and verification.

[0111] Example 5:

[0112] The present application carries out integrity detection on a bored pile with a diameter of 1.2 meters and a depth of 25 meters in a certain water conservancy project, 36 temperature sensors are arranged on the bored pile, and the sensors are distributed along the pile body at equal intervals to form a monitoring network. Taking the starting temperature node of the target conduction curve as an example, the temperature change gradient is 0.15℃ / min, the time stamp is 6.3min, and the node type is the starting node, and the corresponding feature vector is constructed according to the characteristic index; meanwhile, the feature vectors of the top node and each key temperature node of the alternative conduction curve are constructed. The adjacent feature temperature node of the starting temperature node is used as the first adjacent node, the similarity between the feature vectors of the first adjacent node and the starting temperature node is calculated, and the characteristic index of the starting temperature node is obtained; the similarity between the feature vectors of the first adjacent node and the top node is calculated, and the characteristic index of the top node is obtained. Then find the second adjacent node (exclude the starting temperature node) of the first adjacent node, calculate the similarity between the feature vectors of the first adjacent node and the second adjacent node, and obtain the characteristic index of the first adjacent node. Repeat until all temperature nodes in the target conduction curve and the alternative conduction curve are traversed, and obtain the characteristic indexes of all temperature nodes. Based on the characteristic indexes, the first feature set and the second feature set corresponding to the target conduction curve and the alternative conduction curve are generated respectively, the coincidence degree is obtained according to the number of same characteristic indexes in the first feature set and the second feature set, and the alternative conduction curve with the highest coincidence degree is selected as the reference conduction curve. Label the abnormal conduction node in the reference conduction curve to obtain the modified conduction curve, and the abnormal conduction node is the key temperature node with deviation between the actual temperature monitoring data and the reference conduction curve. A thermal conduction anomaly analysis model is established, and the key temperature nodes in the modified conduction curve are checked by using the model to obtain the defect node; generate an integrity report including the position information of the defect node to complete the integrity detection of the bored pile.

[0113] The temperature sensors record sensor data at a frequency of once per minute. After 72 hours of continuous monitoring, it is found that the temperature change rate of sensor group S12-S15 reaches 0.8°C / min within 3 minutes, exceeding the preset threshold of 0.5°C / min. The area is automatically marked as an abnormal area, triggering a real-time monitoring mechanism to increase the sampling frequency to once per second, while calling the data of adjacent sensors S10-S17 for collaborative analysis. The acquired heat conduction characteristics show that the area has an abnormal heat conduction pattern, with an asymmetric temperature gradient distribution and a heat conduction rate that is about 40% faster than normal areas. When updating the heat conduction feature template library, first find the standard heat conduction curve closest to the current abnormal feature in the template library, and use the feature vector similarity algorithm in the matching process; after finding the corresponding curve, integrate the new heat conduction feature as supplementary data into the original curve, update the curve parameters using a weighted average algorithm, and retain the historical version for retrospective analysis. Based on the updated heat conduction feature template library, the integrity detection process is re-executed, including curve fusion, benchmark extraction, abnormal labeling, and model verification, etc. Finally, an integrity report containing the latest analysis results is generated.

[0114] The determination of abnormal changes uses a multi-level early warning mechanism, the first level monitors the instantaneous change of a single sensor, the second level analyzes the spatial correlation of a sensor group, and the third level evaluates the distribution rule of the overall temperature field; each level of early warning triggers different response strategies, from increasing sampling frequency to starting auxiliary detection means. The sensor network uses a ring topology, each sensor works independently and participates in network computing, improving fault tolerance and monitoring coverage. Data transmission uses encryption protocols and redundancy checks to ensure data integrity and security. The feature update process establishes a version management system, each update generates a new curve version and records the update time, update reason and update content; maintain a feature change log to record the modification history of all template libraries, support data backtracking and comparative analysis at any time point. When re-executing integrity detection, prioritize processing data related to abnormal areas, use incremental calculation to reduce processing time, and only re-fuse and verify the affected part of the curve. After generating the integrity report, automatically compare the differences between the new and old reports, label the newly added defect nodes and change areas; the report output uses a standardized format, including data collection time, analysis parameters, detection results and confidence evaluation, etc. At the same time, generate visual charts to show temperature field distribution, defect location and change trend, support multi-dimensional data query and comparative analysis. Set up a self-diagnosis mechanism to check sensor status, data quality and algorithm performance regularly, and automatically generate a health status report. In the long-term running process, continuously optimize the temperature change threshold and feature extraction parameters through machine learning algorithms to adapt to the impact of environmental changes and material aging, maintain monitoring accuracy and reliability.

[0115] It is to be understood that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting; it is not intended to exclude myriad other embodiments of the present application that other inventors can develop based on the same general inventive concepts embodied by the described embodiments. That is, although the present application is described in terms of particular embodiments and illustrative figures, it should be apparent that the scope of the present application is not limited to these specific embodiments.

[0116] While the embodiments of the application have been shown and described herein, it will be understood by those skilled in the art that many changes, modifications, substitutions and alterations to these embodiments can be made without departing from the principles and spirits of the application, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for detecting the integrity of cast-in-place piles based on a heat conduction model, characterized in that, include: S1. Construct a template library of thermal conduction characteristics for cast-in-place piles. The template library includes standard thermal conduction curves corresponding to various integrity states. Each standard thermal conduction curve includes multiple key temperature nodes. S2. Merge the standard heat conduction curves corresponding to the same integrity state to generate a comprehensive heat conduction curve; specifically: The standard heat conduction curve has conduction relationship identifiers between key temperature nodes. Each key temperature node is assigned a sequence number, and the standard heat conduction curve is divided into multiple conduction intervals according to the sequence number. The standard heat conduction curves corresponding to the same integrity state are used as the fusion objects. Based on the fusion objects, the standard heat conduction curve containing the most conduction intervals is used as the master curve, and the key temperature nodes in the master curve are used as master nodes. In addition to the main curve, a reference node is determined among the other fusion objects. Starting from the reference node, the key temperature nodes of each conduction interval of the other fusion objects are fused into the main curve. If the key temperature node and the main node meet the first association condition, the key temperature node is merged into the main node. Otherwise, the key temperature node is connected to the main curve as an auxiliary node to obtain the comprehensive heat conduction curve. The determination of the baseline node includes the following: In the other fusion objects besides the main curve, locate the conduction interval with the smallest sequence number, and take the key temperature node in the conduction interval as the initial node. If a main node that satisfies the first association condition with the initial node is found in the main curve, then the initial node is determined as the reference node; otherwise, in the other fusion objects, select key temperature nodes from the other conduction intervals in ascending order of sequence number. If a main node that satisfies the first association condition with the key temperature node is found in the main curve, then the key temperature node is determined as the reference node. S3. Obtain the actual temperature monitoring data of the cast-in-place pile to be tested, select the corresponding comprehensive heat conduction curve based on the preset integrity level, and extract the benchmark conduction curve from the comprehensive heat conduction curve based on the temperature change characteristics; specifically: The actual temperature monitoring data is processed according to the time series, and the discrete data is transformed into a continuous curve using data fitting methods to obtain the target conduction curve, and the starting temperature node and the ending temperature node in the target conduction curve are located. In the comprehensive heat conduction curve, the key temperature nodes that satisfy the second correlation condition with the starting temperature node and the ending temperature node are respectively located and designated as the top node and the bottom node. Multiple alternative heat conduction curves are generated from the top and bottom nodes in the composite heat conduction curve. The degree of fit between the target heat conduction curve and each alternative heat conduction curve is calculated, and the alternative heat conduction curve with the highest degree of fit is taken as the benchmark heat conduction curve. The location of the starting temperature node and the ending temperature node includes the following: The target conduction curve includes multiple characteristic temperature nodes. Starting from both ends of the target conduction curve, we sequentially determine whether each characteristic temperature node is a core characteristic point. If a key temperature node and a characteristic temperature node in the comprehensive heat conduction curve satisfy the second correlation condition, then the characteristic temperature node is determined as the core characteristic point. The core feature points first determined at both ends of the target conduction curve are defined as the starting temperature node and the ending temperature node, respectively. S4. Mark the abnormal conduction nodes in the baseline conduction curve to obtain the corrected conduction curve. The abnormal conduction nodes are the key temperature nodes where the actual temperature monitoring data deviates from the baseline conduction curve. S5. Establish a heat conduction anomaly analysis model, use this model to verify the key temperature nodes in the modified conduction curve, obtain the defect nodes, generate an integrity report including the location information of the defect nodes, and complete the integrity detection of the cast-in-place pile.

2. The method for detecting the integrity of cast-in-place piles based on a heat conduction model according to claim 1, characterized in that, Determining whether the first association condition is met includes the following: Each key temperature node is configured with a conduction parameter, which includes the temperature change gradient, the temperature change gradient of the preceding node, and the temperature change gradient of the subsequent node. When the temperature change gradients of the master node and the critical temperature node are the same, and at least one of the preceding node and the subsequent node has the same temperature change gradient, the master node and the critical temperature node are determined to satisfy the first association condition.

3. The method for detecting the integrity of cast-in-place piles based on a heat conduction model according to claim 1, characterized in that, Determining whether the second association condition is met includes the following: A conduction parameter is configured for each critical temperature node. When the temperature change gradient of a temperature node is the same as that of a critical temperature node, and at least one of the preceding and subsequent nodes has the same temperature change gradient, the temperature node is determined to satisfy the second association condition with the critical temperature node.

4. The method for detecting the integrity of cast-in-place piles based on a heat conduction model according to claim 3, characterized in that, The calculation of the fit includes the following: Step 1: According to the preset feature sorting rules, integrate the temperature change gradients of the temperature nodes into structured data form to obtain the corresponding feature vectors; when calculating the feature vector similarity between feature temperature nodes in the target conduction curve, the feature temperature nodes adjacent to the starting temperature node are taken as the first neighboring nodes; when calculating the feature vector similarity between the candidate conduction curve and the target conduction curve, the key temperature nodes adjacent to the top node are taken as the first neighboring nodes. Step 2: Calculate the similarity between the feature vectors of the first neighboring node and the starting temperature node to obtain the feature index of the starting temperature node; calculate the similarity between the feature vectors of the first neighboring node and the top node to obtain the feature index of the top node. Step 3: Define any other adjacent temperature node of the first neighboring node as the second neighboring node, calculate the similarity between the feature vectors of the first neighboring node and the second neighboring node, and obtain the feature index of the first neighboring node. Step 4: Repeat step 3 until all temperature nodes in the candidate conduction curve and the target conduction curve are traversed to obtain the characteristic indicators of all temperature nodes. Step 5: Based on the feature indicators obtained in Step 4, generate a first feature set and a second feature set corresponding to the target transmission curve and the alternative transmission curve, respectively, and obtain the degree of fit based on the number of identical feature indicators in the first feature set and the second feature set.

5. The method for detecting the integrity of cast-in-place piles based on a heat conduction model according to claim 1, characterized in that, In step S5, the conduction parameters of key temperature nodes under different integrity states of the cast-in-place pile are used as training samples. The training samples are preprocessed to obtain preprocessed training samples. A heat conduction anomaly analysis model is constructed based on a deep learning framework. The model is trained using preprocessed training samples to obtain the trained heat conduction anomaly analysis model. The conduction parameters of key temperature nodes in the corrected conduction curve are input into the trained heat conduction anomaly analysis model, and the sequence relationship of the parameters is processed to obtain the anomaly index of the key temperature nodes. When the anomaly index exceeds a preset threshold, the key temperature node corresponding to the anomaly index is determined to be a defect node.

6. The method for detecting the integrity of cast-in-place piles based on a heat conduction model according to claim 1, characterized in that, It also includes real-time monitoring steps: The temperature field distribution data of the cast-in-place pile under test is continuously collected. When the temperature change of the temperature field distribution data exceeds the set limit, the real-time monitoring mechanism is triggered. The heat conduction characteristics of the abnormal change area are obtained and updated to the corresponding standard heat conduction curve in the cast-in-place pile heat conduction characteristic template library. Based on the updated cast-in-place pile heat conduction characteristic template library, steps S1-S5 are re-executed to generate an updated integrity report.

Images