NLP-based concrete pile body integrity grade intelligent determination method and system

Abstract

The application discloses a kind of based on NLP's concrete pile body integrity grade intelligent determination method and system, method includes: using acoustic wave transmission method detects to be measured concrete pile body, and obtains the acoustics parameter and received waveform signal data of to be measured concrete pile body;The acoustics parameter and received waveform signal data of to be measured concrete pile body are input into trained integrity grade determination model, and the integrity grade of to be measured concrete pile body is output.The application utilizes natural language processing technology and ultrasonic image recognition technology, solves the technical problems that based on acoustic wave transmission method concrete pile body integrity grade artificial discrimination is inefficient, cannot simultaneously determine the integrity of all to be measured piles under the same project, and the determination result is inconsistent.

Description

TECHNICAL FIELD

[0001] The present application relates to the application of natural language processing and ultrasonic image recognition in the field of pile foundation integrity level determination, and in particular to a concrete pile body integrity level intelligent determination method and system based on NLP. BACKGROUND

[0002] The acoustic transmission method is a method for detecting the integrity of a concrete pile body. The main principle is to measure the acoustic parameters of the measured concrete pile by using the diffraction, reflection and refraction characteristics of ultrasonic waves when encountering defects in the concrete, and then using the acoustic parameter characteristics and ultrasonic signal characteristics to determine the integrity of the concrete pile body and evaluate the position, range and degree of pile body defects. The acoustic transmission method for detecting the integrity of a concrete pile body is widely used in the fields of highway engineering pile detection, railway engineering pile foundation detection, building pile detection, etc., and relevant detection technical specifications have been issued by the state, local and provincial and municipal levels for the acoustic transmission method for detecting the integrity of a concrete pile body.

[0003] The requirements for pile body integrity level determination in the detection technical specifications of highway engineering piles, railway engineering pile foundations, building piles, etc. differ due to differences in application scenarios and regional soil structures. However, each detection technical specification includes a pile body integrity level determination table, which describes the comparison between the pile body integrity level and the acoustic parameter characteristics and ultrasonic signal characteristics in natural language.

[0004] The traditional way of determining the integrity level of a pile body based on the acoustic transmission method for detecting the integrity of a concrete pile body is manual determination by a detection personnel, who determines the integrity level of the measured concrete pile based on their understanding of the pile body integrity level determination table in the detection technical specifications issued by the state, local and provincial and municipal levels, combined with the acoustic parameter characteristics and ultrasonic signal characteristics of the measured concrete pile. However, as the state places increasing emphasis on the quality of construction engineering pile foundations, the workload of manual determination has increased dramatically, and the traditional method has individual subjective differences, leading to inconsistent determination results. SUMMARY

[0005] To overcome the shortcomings of the prior art, the present application provides a concrete pile body integrity level intelligent determination method and system based on NLP, which uses natural language processing technology and ultrasonic image recognition technology to solve the technical problems of low efficiency of manual determination of the integrity level of a concrete pile body based on the acoustic transmission method, the inability to simultaneously determine the integrity of all piles to be measured under the same project, and inconsistent determination results.

[0006] According to an aspect of the present application, a concrete pile body integrity level intelligent determination method based on NLP is provided, comprising:

[0007] Using the acoustic wave transmission method to detect the concrete pile body to be tested, and obtaining the acoustic parameters of the concrete pile body to be tested and receiving waveform signal data;

[0008] Inputting the acoustic parameters of the concrete pile body to be tested and the received waveform signal data into the trained integrity level determination model, and outputting the integrity level of the concrete pile body to be tested; wherein the training of the integrity level determination model includes:

[0009] Identify the corpus of several testing technical regulations corresponding to the integrity test of concrete piles, and obtain the acoustic parameters and waveform feature description information corresponding to the integrity level;

[0010] Extracting the judgment object and its characteristic keyword set based on the acoustic parameters and waveform feature description information corresponding to the integrity level, and simultaneously counting the fuzzy quantitative characteristic keywords;

[0011] Based on the pre-built quantitative discrimination formulas for state characteristics corresponding to acoustic parameters, the quantitative discrimination formulas for distribution characteristics corresponding to acoustic parameters, and the trained classification model for received waveform state characteristics, the extracted judgment objects and their characteristic keyword sets are quantified to form an integrity level judgment model.

[0012] A training sample set for integrity level determination is constructed, an integrity level determination model is trained using the training sample set, and fuzzy quantitative feature keywords are used as adjustment parameters to obtain a trained integrity level determination model.

[0013] As a further technical solution, the extracted judgment objects include: acoustic parameters, received waveform, sound speed range and first wave.

[0014] As a further technical solution, the pre-construction of the quantitative discrimination formula corresponding to the state characteristics of the acoustic parameters includes:

[0015] Obtain a historical data set of manually identified concrete pile integrity levels, and classify and statistically analyze the quantified values ​​of acoustic parameter state characteristics based on the integrity levels identified in the historical data set;

[0016] According to the classification statistical results, the characteristic quantitative limit values ​​of sound velocity, sound amplitude, sound time, main frequency and PSD in different states are calculated respectively;

[0017] According to the calculated characteristic quantization limit values ​​of different states, a characteristic quantization discrimination formula corresponding to the state of acoustic parameters is constructed.

[0018] As a further technical solution, the pre-construction of the quantitative discrimination formula corresponding to the distribution characteristics of the acoustic parameters includes:

[0019] Classifying and statistically analyzing the characteristic quantitative values ​​of the distribution state according to the integrity level determined in the historical data set;

[0020] According to the classification statistics, the characteristic quantization limit values of the longitudinal and transverse distribution directions in different distribution states are calculated respectively;

[0021] According to the calculated characteristic quantization limit values of different distribution states, an acoustic parameter corresponding distribution characteristic quantization value discrimination formula is constructed.

[0022] As a further technical solution, the training of the received waveform state feature classification model comprises:

[0023] Obtain historical data of concrete pile body detection;

[0024] According to the historical data, a training sample set and a test sample set are constructed, and each waveform data in the sample set is configured with a classification label;

[0025] A received waveform state feature classification model is constructed, including a CNN model for extracting high-level features of received waveform data, and an LSTM model for capturing time sequence information of received waveform data;

[0026] The received waveform state feature classification model is trained using the training sample set, and tested using the test sample set, to obtain a trained received waveform state feature classification model.

[0027] As a further technical solution, the acoustic wave transmission method is used to detect the to-be-detected concrete pile body, and the acoustic parameters and received waveform signal data of the to-be-detected concrete pile body are obtained, comprising:

[0028] The to-be-detected concrete pile body is detected using the acoustic wave transmission method to obtain a to-be-detected data set saved in a set format;

[0029] The acoustic parameter state feature quantization value of the to-be-detected data set is calculated, including the feature quantization values of the acoustic velocity, acoustic amplitude, acoustic time, main frequency and PSD respectively calculated based on the real-time values;

[0030] The acoustic parameter distribution feature quantization value of the to-be-detected data set is calculated, including the feature quantization values of the longitudinal distribution and transverse distribution respectively calculated based on the acoustic parameter state feature quantization values of the above to-be-detected data set;

[0031] All the measured point data of the to-be-detected data set are converted into time sequence signals and the corresponding state features are predicted using the trained received waveform state feature classification model to obtain the received waveform state feature quantization value.

[0032] As a further technical solution, the integrity level determination model comprises a plurality of determination levels, each determination level corresponding to one or more determination conditions, and when performing level determination, the determination is performed in a manner of gradually decreasing from a high level to a low level, and if a certain condition is met, the integrity level corresponding to the condition is determined; wherein the one or more determination conditions comprise: an acoustic parameter corresponding state feature quantization determination formula, an acoustic parameter corresponding distribution feature quantization determination formula, a trained received waveform state feature classification model, and a logical relationship of each acoustic parameter in the detection technical specification corresponding to the to-be-measured concrete pile body.

[0033] According to an aspect of the present application, an NLP-based concrete pile body integrity level intelligent determination system is provided, comprising:

[0034] A data acquisition module is configured to detect a to-be-measured concrete pile body by using an acoustic wave transmission method, and acquire acoustic parameters and received waveform signal data of the to-be-measured concrete pile body.

[0035] A level determination module is configured to input the acoustic parameters and received waveform signal data of the to-be-measured concrete pile body into a trained integrity level determination model, and output an integrity level of the to-be-measured concrete pile body; wherein the training of the integrity level determination model comprises:

[0036] Identifying a corpus of a plurality of detection technical specifications corresponding to concrete pile body integrity detection, and acquiring acoustic parameters and waveform feature description information corresponding to the integrity level;

[0037] Based on the acoustic parameters and waveform feature description information corresponding to the integrity level, a determination object and a feature keyword set thereof are extracted, and a fuzzy number feature keyword is counted;

[0038] Based on a pre-constructed acoustic parameter corresponding state feature quantization determination formula, an acoustic parameter corresponding distribution feature quantization determination formula, and a trained received waveform state feature classification model, the extracted determination object and the feature keyword set thereof are quantized to form an integrity level determination model;

[0039] A training sample set for integrity level determination is constructed, the training sample set is used to train the integrity level determination model, and the fuzzy number feature keyword is used as an adjustment parameter to obtain a trained integrity level determination model.

[0040] Compared with the prior art, the present application has the following advantages:

[0041] The application utilizes natural language processing technology and ultrasonic image recognition technology, and proposes an NLP-based concrete pile body integrity grade intelligent determination method and system, and solves the technical problems of low artificial discrimination efficiency and inconsistent determination results of the concrete pile body integrity grade based on the sound transmission method. BRIEF DESCRIPTION OF DRAWINGS

[0042] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings used in the embodiments or prior art description will be briefly introduced. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0043] Figure 1 The flowchart of the NLP-based concrete pile body integrity grade intelligent determination method provided by the embodiments of the present application is shown.

[0044] Figure 2 The training flowchart of the integrity grade determination model provided by the embodiments of the present application is shown.

[0045] Figure 3 The RSM ultrasonic general format provided by the embodiments of the present application is shown.

[0046] Figure 4 The structure diagram of the NLP-based concrete pile body integrity grade intelligent determination system provided by the embodiments of the present application is shown. DETAILED DESCRIPTION

[0047] It should be noted that:

[0048] The terms "include" and "have" and any variations thereof in the specification and claims of the present application and the above drawings are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device including a series of steps or units does not have to include all the steps or units clearly listed, but can include other steps or units not clearly listed or inherent to the process, method, product or device.

[0049] The block diagrams shown in the drawings are merely functional entities, and do not necessarily correspond to physically independent entities. That is, the functional entities can be implemented in the form of software, or in one or more hardware modules or integrated circuits, or in different networks and / or processor devices and / or microcontroller devices. The flowcharts shown in the drawings are merely exemplary illustrations, and do not necessarily include all contents and operations / steps, nor are they necessarily executed in the order described. For example, some operations / steps can be further decomposed, and some operations / steps can be combined or partially combined, so that the actual execution order can be changed according to actual conditions.

[0050] In order to make the purposes, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application. In addition, the technical features in each embodiment or in a single embodiment provided by the present application can be combined with each other at will to form new technical solutions, and such combination is not restricted by the order of steps and / or structure composition mode, but must be based on the fact that those skilled in the art can implement it. When the combination of technical solutions contradicts each other or cannot be implemented, it should be considered that such combination of technical solutions does not exist, nor is it within the scope of protection required by the present application.

[0051] The embodiments of the present application provide an NLP-based concrete pile body integrity grade intelligent determination method, as shown in Figure 1 The method comprises: detecting a to-be-detected concrete pile body by using an acoustic transmission method, and acquiring acoustic parameters and received waveform signal data of the to-be-detected concrete pile body; inputting the acoustic parameters and received waveform signal data of the to-be-detected concrete pile body into a trained integrity grade determination model, and outputting an integrity grade of the to-be-detected concrete pile body.

[0052] As shown in Figure 2 The training of the integrity grade determination model comprises:

[0053] Corpus of a plurality of detection technical regulations corresponding to the integrity detection of the concrete pile body is recognized, and acoustic parameters and waveform feature description information corresponding to the integrity grade are acquired;

[0054] Based on the acoustic parameters and waveform feature description information corresponding to the integrity grade, a determination object and a feature keyword set thereof are extracted, and a fuzzy number feature keyword is counted;

[0055] Based on the pre-constructed acoustic parameter corresponding state feature quantization discriminant formula, the acoustic parameter corresponding distribution feature quantization discriminant formula and the trained received waveform state feature classification model, the extracted judgment object and its feature keyword set are quantized to form a completeness level judgment model.

[0056] A training sample set for completeness level judgment is constructed, the training sample set is used to train the completeness level judgment model, and the fuzzy number feature keyword is used as an adjustment parameter to obtain the trained completeness level judgment model.

[0057] In the embodiment of the present application, before the integrity of the concrete pile body is judged, it also includes corpus collection in the field of concrete pile body detection and design of judgment object and its attribute keywords.

[0058] Specifically, the corpus in the field of concrete pile body integrity detection is collected and summarized, such as "Highway Engineering Pile Detection Technical Specification (JTG / T 3512-2020)", "Railway Engineering Pile Detection Technical Specification (TB10218-2019)", "Building Foundation Detection Specification (DB32T 3916-2020)" and the like. The judgment object and its attribute keyword dictionary {E|E has A number of characteristic attributes, B state characteristic attributes, C distribution state characteristic attributes} is designed, as shown in Table 1.

[0059] Table 1 Judgment object and its attribute keyword table

[0060]

[0061] In the embodiment of the present application, before the integrity of the concrete pile body is judged, it also includes designing the acoustic parameter state feature quantization value standard. Specifically, when V i ′<1, the sound velocity is judged as an abnormal value, A′ i <0, the sound amplitude is judged as an abnormal value, when T i ′<1, the sound time is judged as an abnormal value, when F i ′<1, the main frequency is judged as an abnormal value, when P i ′<1, the PSD is judged as an abnormal value.

[0062] The calculation of the acoustic parameter corresponding state feature quantization value is as follows:

[0063] The sound velocity state feature quantization value is: (wherein V i is the sound velocity of the measuring point, and V c is the critical value of the sound velocity calculated according to the detection specification);

[0064] The sound amplitude state feature quantization value is A′ i =A i -Ac (A i is the sound amplitude of the measuring point, A c is the critical value of the sound amplitude calculated according to the detection procedure);

[0065] Sound time state characteristic quantization value: (A i is the sound time of the measuring point, T c is the critical value of the sound time calculated according to the detection procedure);

[0066] Dominant frequency state characteristic quantization value: (A i is the dominant frequency of the measuring point, F c is the critical value of the dominant frequency calculated according to the detection procedure);

[0067] PSD state characteristic quantization value: (A i is the PSD of the measuring point, P c is the critical value of the PSD calculated according to the detection procedure).

[0068] In the embodiment of the present application, the construction of the state characteristic quantization discrimination formula of the acoustic parameter corresponds to:

[0069] The acoustic parameter state characteristic quantization limit value is calculated, the collected data set of the artificial discriminated concrete pile is classified and counted according to the integrity grade of the discrimination, the acoustic parameter state characteristic quantization value of the class IV pile is counted as a serious abnormal sample set, the acoustic parameter state characteristic quantization value of the class III pile is counted as an obvious abnormal sample set, the acoustic parameter state characteristic quantization value of the class II pile is counted as a slight abnormal sample set, and the acoustic parameter state characteristic quantization value of the class I pile is counted as a normal sample set; then the average value is calculated after the 3 times standard deviation is removed, the sound velocity, the sound amplitude, the sound time, the dominant frequency and the PSD four state characteristic quantization limit values are calculated in turn, and the discrimination formula is formulated, as shown in Table 2.

[0070] Table 2 Acoustic parameter state characteristic quantization discrimination formula

[0071] Acoustic parameter Severe anomaly Obvious anomaly Minor anomaly Normal Sound velocity [V'1 < V3] V3 < V' i V2 V2 < V' i V1 < V V1≤ V' i ]] Sound amplitude [A i <A3 [A3≤ A' i <A2]] [A2≤ A' i <A1]] [A1≤ A' i ]]> Sound time [T i <T3 T3≤ T' i <T2]] T2 < T' i <T1 T1≤ T' i ]] Dominant frequency F' i <F3]]> F3 < F' i <F2 <![CDATA[F2≤F′ i <F1]]> F1≤ F' i ]] PSD <![CDATA[P′ i <P3]]> [P3≤ P' i <P2]] P2 < P' i < P1 P i ≤ P' i ]]>

[0072] In Table 2, V1, V2, V3 are empirical values given with reference to the critical value of the intermediate sound velocity in the detection technical specification and they are all less than the critical value of the sound velocity and decrease in turn, A1, A2, A3 are empirical values given with reference to the critical value of the intermediate sound amplitude (wave amplitude) in the detection technical specification and they are all less than the critical value of the sound amplitude and decrease in turn, T1, T2, T3 are empirical values given with reference to the critical value of the intermediate first sound time (sound time) in the detection technical specification and they are all less than the critical value of the sound time and decrease in turn, PSD1, PSD2, PSD3 are empirical values given with reference to the critical value of the PSD (the product of the slope on the sound time depth curve and the sound time change) in the detection technical specification and they are all less than the critical value of the PSD and decrease in turn, F1, F2, F3 are empirical values given with reference to the critical value of the main frequency in the detection technical specification and they are all less than the critical value of the main frequency and decrease in turn.

[0073] In the embodiment of the present application, the construction of the acoustic parameter corresponding distribution feature quantization discriminant formula comprises:

[0074] The distribution state feature quantization limit value is calculated, wherein the longitudinal distribution abnormal state quantization value is the depth value (DN e ) of the continuous abnormal measuring point block, and the transverse distribution abnormal state quantization value is the abnormal profile number proportion (PP e ), the collection data of the type IV pile is taken as the distribution abnormal state quantization value sample set of a "very large area", the collection data of the type III pile is taken as the quantization value sample set of a "large area", the collection data of the type II pile is taken as the quantization value sample set of a "small area", and the collection data of the type I pile is taken as the quantization value sample set of a "small spatial distribution range"; then the average value is calculated by using the method of removing abnormal values by 3 times the standard deviation, and the longitudinal and transverse distribution state feature quantization limit values of the four kinds of "very large area", "large area", "small area" and "small spatial distribution range" are calculated in turn, and the discriminant formula is formulated, as shown in Table 3.

[0075] Further, the depth value of the continuous abnormal measuring point block is wherein i is continuously increased, that is, 1==Δi, ΔH is the measuring point moving step, and n represents the number of abnormal measuring points, represents an abnormal measuring point. When the number of abnormal types of the acoustic parameter of the measuring point reaches or is greater than the minimum value of the number of abnormal acoustic parameters in the corresponding integrity category level, the measuring point is recorded as an abnormal measuring point

[0076] Further, the abnormal profile number proportion is m represents the number of abnormal profiles, represents an abnormal profile, k represents the number of all detection profiles, and P represents a detection profile. When the number of abnormal measuring points of the profile If the total number of the abnormal measuring points in the corresponding integrity category level is equal to or greater than the minimum value, the profile is recorded as an abnormal profile

[0077] Table 3 longitudinal, lateral distribution state quantitative discrimination formula

[0078] Very large area Large area Smaller area Small spatial distribution range Longitudinal D3≤ DN e ]]> D2≤ DN e D3 D1≤ DN e D2]]> DN e <D1]]> Transverse P3 < PP e ]] [P2 < PP e <P3]]> [P1≤PP e <P2]]> PP e <P1]]>

[0079] In Table 3, D1, D2, D3 are the depth ranges of continuous abnormal measuring points summarized according to experience and are greater than the minimum calculated value of the depth range and sequentially increased, and P1, P2, P3 are the proportions of the number of abnormal detection profiles and sequentially increased.

[0080] The embodiment of the present application also includes designing a low limit value V low of the sound velocity of the data set collected by the type IV and III pile, and the average value calculated after removing the abnormal value by 3 times the standard deviation is used as the low limit value V low of the sound velocity V i at a measuring point is greater than or equal to V low , the corresponding keyword is “higher than the low limit value” or no keyword; and when the sound velocity V i at a measuring point is less than V low , the corresponding keyword is “lower than the low limit value”.

[0081] The embodiment of the present application also includes designing a low limit value T low of the first wave, and when the first arrival time T i at a measuring point is greater than or equal to T low , the corresponding keyword is “normal” or no keyword; and when the first arrival time T i at a measuring point is less than T low , the corresponding keyword is “undetectable”.

[0082] The embodiment of the present application also includes establishing a fuzzy quantity characteristic keyword value quantitative correction questionnaire (as shown in Table 4), such as “continuous”, “more”, “a certain depth range” and the like, which is difficult to find a reasonable sample set for derivation and estimation. Therefore, according to the experience of experts for many years, the minimum value is given, and then the optimal value is obtained according to the accuracy rate in the model evaluation stage. Specifically, Table 4 gives the minimum or maximum value of the fuzzy field according to the experience value, and the detection unit can appropriately adjust according to the actual situation.

[0083] Table 4 Fuzzy quantity characteristic keyword value quantitative correction questionnaire

[0084]

[0085] In the embodiment of the present application, the construction and training of the received waveform state feature classification model includes:

[0086] The received waveform state feature classification model is established, and the collected data of the concrete pile in the concrete pile integrity detection method of the acoustic wave transmission method is composed of a large number of measured point data of multiple profiles, one measured point data can be converted into a time sequence signal, which is also called a received waveform RW i , and the length of the single measured point time sequence signal is usually preset as 512, and the sampling period is usually preset as 1us. The received waveform state feature classification model utilizes the feature extraction capability of CNN and the time sequence modeling capability of LSTM to classify the state features of the collected time sequence signals of the concrete pile.

[0087] Specifically, the training process comprises: 1) data collection, collecting the collected data original data file of the concrete pile; 2) manual integrity discrimination, discriminating the collected original data file for integrity, and generating a verified test sample set; 3) data preprocessing, standardizing or normalizing the original waveform data; 4) data set division, dividing the preprocessed data into a training sample set and a test sample set, according to the manual discrimination result, assigning a classification label (normal-B1 w , slight distortion-B2 w , obvious distortion-B3 w , serious distortion-B4 w ) to each data; 5) establishing a CNN model, using CNN to extract high-level features of the received waveform data; 6) establishing an LSTM model, using LSTM to capture time sequence information of the received waveform data; 6) training the model, using the training sample set to train the combined model; 7) evaluating the model, using the test sample set to evaluate the performance of the model with precision rate as the evaluation standard; 8) saving the model.

[0088] Further, the received waveform feature information is extracted from the collected received waveform record, and the received waveform features are saved as state feature models B1 w , B2 w , B3 w , B4 w in turn according to the state level, to form a received waveform state feature model {B1 w , B2 w , B3 w , B4 w}.

[0089] In the embodiment of the application, the training of the integrity level determination model further comprises identification and training of the integrity determination table of the detection technical procedure.

[0090] Specifically, it includes: 1) using the table recognition function of OCR technology to recognize the picture of the "integrity judgment table of detection technical regulations", such as the integrity judgment table of "Highway Engineering Foundation Pile Detection Technical Regulations (JTG / T 3512-2020)", to obtain the information of acoustic parameters and waveform feature description corresponding to the integrity category; 2) using Jieba word segmentation open source library for word segmentation processing; 3) taking the punctuation mark ";" as the end mark of the conditional statement to divide each integrity description information into multiple conditional statements; 4) extracting the judgment object and its feature keyword set {E_r|A_r, B_r, C_r} from the conditional statement one by one, such as the feature keyword set of "Highway Engineering Foundation Pile Detection Technical Regulations (JTG / T 3512-2020)", the judgment object has multiple state attribute keywords, and the most serious state is used as the criterion; 5) counting the fuzzy number feature keywords, such as "continuous", "a certain depth range", "all", "more" and the like.

[0091] Table 5 Integrity judgment table of "Highway Engineering Foundation Pile Detection Technical Regulations (JTG / T 3512-2020)"

[0092]

[0093] Table 6 Feature keyword set of "Highway Engineering Foundation Pile Detection Technical Regulations (JTG / T 3512-2020)"

[0094]

[0095] In the embodiment of the present application, the concrete pile body integrity grade determination model specifically includes: 1) quantizing the extracted detection specification integrity category evaluation object and its feature keyword set using the "Table 2 acoustic parameter corresponding state feature quantization discrimination formula", "Table 3 transverse and longitudinal distribution state quantization discrimination formula", and the received waveform state feature classification model, and the quantization results of Table 6 are as shown in Table 7 "Highway Engineering Foundation Pile Detection Technical Specification (JTG / T 3512-2020)" quantization table. 2) constructing an integrity grade determination model, as shown in Table 8 "Highway Engineering Foundation Pile Detection Technical Specification (JTG / T 3512-2020)" integrity grade determination model, wherein the concrete pile body integrity grade determination rule is to determine the higher grade first, and if the higher grade determination condition is not met, the previous grade is determined. In the determination of each specific grade, the logical relationship of each acoustic parameter corresponding to the detection technology specification of the concrete pile body to be tested, such as logical and, logical or, etc., is also needed, and the acoustic parameter corresponding state feature quantization discrimination formula, the acoustic parameter corresponding distribution feature quantization discrimination formula, and the trained received waveform state feature classification model are used as the criterion to determine the grade of the concrete pile body to be tested. 3) evaluation model, using the collected data for artificial determination of integrity grade as test sample set to evaluate the performance of the model with precision rate as evaluation standard, wherein the "fuzzy number feature keyword value quantization correction questionnaire" is used as adjustment parameter. 4) save the model.

[0096] Table 7 "Highway Engineering Foundation Pile Detection Technical Specification (JTG / T 3512-2020)" quantization table

[0097]

[0098] Table 8 "Highway Engineering Foundation Pile Detection Technical Specification (JTG / T 3512-2020)" integrity grade determination model

[0099]

[0100]

[0101]

[0102]

[0103]

[0104]

[0105] As an embodiment, when the concrete pile body integrity grade is detected by using the aforementioned method of the present application, the received waveform data collected by using the acoustic wave transmission method to detect the integrity of the concrete pile is preprocessed, and then the preprocessed data is input into the received waveform state feature classification model to determine the integrity category of the concrete pile body. Figure 3The acquisition data saved in the shown "RSM ultrasonic general format" are subjected to integrity level prediction to determine the corresponding integrity levels, including: 1) calculating the critical value of the acoustic parameter of the to-be-tested data set according to the detection procedure; 2) calculating the quantized value of the acoustic parameter of the to-be-tested data set according to the aforementioned "standard of quantized value of state of acoustic parameter"; 3) converting all the data of the measurement points of the to-be-tested data set into time series signals and predicting the corresponding state features by using the "state feature classification model of received waveform"; and 4) predicting and determining the corresponding integrity levels of the processed to-be-tested data set by using the "pile concrete integrity level determination model".

[0106] The implementation basis of each embodiment of the present application is achieved by programmed processing of a device with processor function. Therefore, in engineering practice, the technical solutions and functions of each embodiment of the present application are packaged into various modules. Based on this actual situation, on the basis of the above-mentioned embodiments, the embodiments of the present application provide an NLP-based intelligent concrete pile body integrity level determination system, which is used to execute the NLP-based intelligent concrete pile body integrity level determination method in the above-mentioned method embodiments.

[0107] Referring to Figure 4 The system comprises a data acquisition module configured to detect a to-be-tested concrete pile body by using the acoustic wave transmission method and acquire acoustic parameters and received waveform signal data of the to-be-tested concrete pile body; and a level determination module configured to input the acoustic parameters and received waveform signal data of the to-be-tested concrete pile body into a trained integrity level determination model and output the integrity level of the to-be-tested concrete pile body. The training of the integrity level determination model comprises: identifying the corpus of a plurality of detection technical procedures corresponding to the integrity detection of the concrete pile body, acquiring acoustic parameters and waveform feature description information corresponding to the integrity level, extracting the determination object and its feature keyword set based on the acoustic parameters and waveform feature description information corresponding to the integrity level, and simultaneously counting the fuzzy number feature keywords; quantizing the extracted determination object and its feature keyword set based on a pre-constructed state feature quantization discrimination formula corresponding to the acoustic parameter, a distribution feature quantization discrimination formula corresponding to the acoustic parameter, and a trained received waveform state feature classification model, to form the integrity level determination model; constructing a training sample set for the integrity level determination, training the integrity level determination model with the training sample set, and taking the fuzzy number feature keywords as the adjustment parameters to obtain the trained integrity level determination model.

[0108] The NLP-based intelligent concrete pile body integrity level determination system provided by the embodiments of the present application adopts Figure 4Several modules in the system utilize natural language processing technology and ultrasonic image recognition technology to solve technical problems such as low efficiency of manual discrimination of concrete pile integrity grade based on sound wave transmission method, inability to simultaneously discriminate integrity of all piles to be measured in the same project, and inconsistent discrimination results.

[0109] It should be noted that the system embodiments provided by the present application are used to implement the methods in the above method embodiments, and are also used to implement the methods in other method embodiments provided by the present application. The difference is only that the corresponding function modules are set, and the principle is basically the same as that of the above system embodiments provided by the present application. As long as the person skilled in the art improves the modules in the above system embodiments on the basis of the above system embodiments, refers to the specific technical solutions in other method embodiments, obtains the corresponding technical means by combining technical features, and the technical solutions composed of these technical means, as long as the technical solutions have practicality, the corresponding system class embodiments are obtained, which are used to implement the methods in other method class embodiments. For example:

[0110] Based on the content of the above system embodiments, as a preferred embodiment, the NLP-based concrete pile integrity grade intelligent discrimination system provided in the embodiments of the present application further comprises:

[0111] The determined discrimination object is: acoustic parameters, received waveforms, sound velocity range and first wave.

[0112] Based on the content of the above system embodiments, as a preferred embodiment, in the NLP-based concrete pile integrity grade intelligent discrimination system provided in the embodiments of the present application, the pre-construction of the acoustic parameter corresponding state feature quantization discrimination formula further comprises:

[0113] Obtain a historical data set of manual discrimination of concrete pile integrity grade, and classify and statistically analyze the state feature quantization values of acoustic parameters according to the integrity grade discriminated in the historical data set;

[0114] According to the classification and statistical results, the characteristic quantization limit values of sound velocity, sound amplitude, sound time, main frequency and PSD in different states are calculated respectively;

[0115] According to the calculated characteristic quantization limit values of different states, the acoustic parameter corresponding state feature quantization discrimination formula is constructed.

[0116] Based on the content of the above system embodiments, as a preferred embodiment, in the NLP-based concrete pile integrity grade intelligent discrimination system provided in the embodiments of the present application, the pre-construction of the acoustic parameter corresponding distribution feature quantization discrimination formula further comprises:

[0117] According to the integrity level distinguished in the historical data set, the feature quantization values of the distribution state are classified and counted;

[0118] According to the classification counting result, the feature quantization limit values of the longitudinal and transverse distribution directions in different distribution states are calculated respectively;

[0119] According to the calculated feature quantization limit values of different distribution states, the acoustic parameter corresponding distribution feature quantization value discrimination formula is constructed.

[0120] Based on the content of the above system embodiment, as a preferred embodiment, in the NLP-based concrete pile body integrity level intelligent judgment system provided in the embodiment, the training of the received waveform state feature classification model further includes:

[0121] Obtaining historical data of concrete pile body detection;

[0122] According to the historical data, a training sample set and a test sample set are constructed, and each waveform data in the sample set is configured with a classification label;

[0123] A received waveform state feature classification model is constructed, including a CNN model for extracting high-level features of received waveform data, and an LSTM model for capturing time sequence information of received waveform data;

[0124] The received waveform state feature classification model is trained using the training sample set, and tested using the test sample set, to obtain a trained received waveform state feature classification model.

[0125] Based on the content of the above system embodiment, as a preferred embodiment, in the NLP-based concrete pile body integrity level intelligent judgment system provided in the embodiment, the acoustic wave transmission method is used to detect the to-be-detected concrete pile body, and the acoustic parameters and received waveform signal data of the to-be-detected concrete pile body are obtained, and further includes:

[0126] The to-be-detected concrete pile body is detected by using the acoustic wave transmission method to obtain a to-be-detected data set saved in a set format;

[0127] The acoustic parameter state feature quantization values of the to-be-detected data set are calculated, including the feature quantization values of each calculated based on the real-time values of sound velocity, sound amplitude, sound time, main frequency and PSD;

[0128] The acoustic parameter distribution feature quantization values of the to-be-detected data set are calculated, including the feature quantization values of longitudinal distribution and transverse distribution calculated based on the acoustic parameter state feature quantization values of the above to-be-detected data set respectively;

[0129] The all-measurement point data of the to-be-tested data set is converted into a time sequence signal, and a corresponding state feature is predicted by using the trained received waveform state feature classification model to obtain a received waveform state feature quantitative value.

[0130] Based on the above system embodiment, as a preferred embodiment, in the NLP-based concrete pile body integrity grade intelligent judgment system provided in the embodiment, the integrity grade judgment model includes multiple judgment grades, each judgment grade corresponds to one or more judgment conditions, and when performing grade discrimination, the judgment is performed in a manner of gradually decreasing from a high grade to a low grade, and if a condition is met, the integrity grade corresponding to the condition is determined.

[0131] The system embodiments described above are merely illustrative, wherein the units described as separate components are or are not physically separated, and the components shown as units are or are not physical units, located in one place, or distributed to multiple network units. The purpose of the embodiment scheme is achieved by selecting part or all of the modules. Those skilled in the art can understand and implement without creative labor.

[0132] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer usable program code.

[0133] The present application is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of flows and / or blocks in the flowcharts and / or block diagrams can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing apparatus to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing apparatus produce a machine that implements the functions specified in the flowcharts and / or block diagrams. Figure 1 The apparatus for implementing each flow or multiple flows and / or blocks Figure 1 The apparatus for implementing each flow or multiple flows and / or blocks

[0134] These computer program instructions can also be stored in a computer readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer readable memory produce an article of manufacture including instructions which implement the flow Figure 1 of the flow or flows and / or blocks Figure 1 of the block or blocks specified in the flow.

[0135] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the flow Figure 1 of the flow or flows and / or blocks Figure 1 of the block or blocks specified in the flow.

[0136] Finally, it should be noted that the above-mentioned embodiments are merely used to illustrate the technical solutions of the present application, rather than limit the present application; even though the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that the technical solutions recorded in the foregoing embodiments can be modified, or some or all of the technical features thereof can be substituted with equivalent replacements; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the technical solutions of the embodiments of the present application.

Claims

1. The NLP-based intelligent determination method for concrete pile integrity level is characterized by: include: Using the acoustic wave transmission method to detect the concrete pile body to be tested, and obtaining the acoustic parameters of the concrete pile body to be tested and receiving waveform signal data; Inputting the acoustic parameters of the concrete pile body to be tested and the received waveform signal data into the trained integrity level determination model, and outputting the integrity level of the concrete pile body to be tested; wherein the training of the integrity level determination model includes: Identify the corpus of several testing technical regulations corresponding to the integrity test of concrete piles, and obtain the acoustic parameters and waveform feature description information corresponding to the integrity level; Extracting the judgment object and its characteristic keyword set based on the acoustic parameters and waveform feature description information corresponding to the integrity level, and simultaneously counting the fuzzy quantitative characteristic keywords; Based on the pre-constructed quantitative discrimination formula for state characteristics corresponding to acoustic parameters, the quantitative discrimination formula for distribution characteristics corresponding to acoustic parameters, and the trained reception waveform state characteristic classification model, the extracted judgment objects and their characteristic keyword sets are quantified to form an integrity level judgment model; the training of the reception waveform state characteristic classification model includes: obtaining historical data of concrete pile body detection; constructing a training sample set and a test sample set based on the historical data, and configuring a classification label for each waveform data in the sample set; constructing a reception waveform state characteristic classification model, including a CNN model for extracting high-level features of the reception waveform data, and an LSTM model for capturing the timing information of the reception waveform data; using the training sample set to train the reception waveform state characteristic classification model, and using the test sample set to test it, to obtain a trained reception waveform state characteristic classification model; A training sample set for integrity level determination is constructed, an integrity level determination model is trained using the training sample set, and fuzzy quantitative feature keywords are used as adjustment parameters to obtain a trained integrity level determination model.

2. The NLP-based intelligent determination method for concrete pile integrity level according to claim 1 is characterized in that: The extracted judgment objects include: acoustic parameters, received waveform, sound speed range and first wave.

3. The NLP-based intelligent determination method for concrete pile integrity level according to claim 1 is characterized in that: The pre-construction of the quantitative discrimination formula corresponding to the state characteristics of the acoustic parameters includes: Obtain a historical data set of manually identified concrete pile integrity levels, and classify and statistically analyze the quantified values ​​of acoustic parameter state characteristics based on the integrity levels identified in the historical data set; According to the classification statistical results, the characteristic quantitative limit values ​​of sound velocity, sound amplitude, sound time, main frequency and PSD in different states are calculated respectively; According to the calculated characteristic quantization limit values ​​of different states, a characteristic quantization discrimination formula corresponding to the state of acoustic parameters is constructed.

4. The NLP-based intelligent determination method for concrete pile integrity level according to claim 3 is characterized in that: The pre-construction of the quantitative discrimination formula corresponding to the distribution characteristics of the acoustic parameters includes: Classifying and statistically analyzing the characteristic quantitative values ​​of the distribution state according to the integrity level determined in the historical data set; According to the classification statistics, the characteristic quantitative limit values ​​of the longitudinal and transverse distribution directions in different distribution states are calculated respectively; According to the calculated characteristic quantization limit values ​​of different distribution states, a discriminant formula for the distribution characteristic quantization value corresponding to the acoustic parameter is constructed.

5. The NLP-based intelligent determination method for concrete pile integrity level according to claim 1 is characterized in that: The acoustic wave transmission method is used to detect the concrete pile body to be tested, and the acoustic parameters of the concrete pile body to be tested and the received waveform signal data are obtained, including: The concrete pile to be tested is detected by using the acoustic wave transmission method to obtain a test data set saved in a set format; Calculate the quantized characteristic values ​​of the acoustic parameter state of the data set to be tested, including the quantized characteristic values ​​of each of the parameters calculated based on the real-time values ​​of sound velocity, sound amplitude, sound time, main frequency and PSD; Calculating the acoustic parameter distribution characteristic quantization value of the data set to be measured, including the characteristic quantization values ​​of the longitudinal distribution and the transverse distribution respectively calculated based on the acoustic parameter state characteristic quantization value of the data set to be measured; All the measurement point data of the data set to be measured are converted into time series signals and the corresponding state features are predicted using the trained receiving waveform state feature classification model to obtain the quantized value of the receiving waveform state feature.

6. The NLP-based intelligent determination method for concrete pile integrity level according to claim 1 is characterized in that: The integrity level determination model includes multiple determination levels, each of which corresponds to one or more determination conditions. When performing level determination, the determination is made in a step-by-step manner from high to low levels. If a certain condition is met, the integrity level corresponding to this condition is determined. Among them, one or more of the determination conditions include: a quantitative determination formula for state characteristics corresponding to acoustic parameters, a quantitative determination formula for distribution characteristics corresponding to acoustic parameters, a trained classification model for received waveform state characteristics, and the logical relationship between various acoustic parameters in the inspection technical regulations corresponding to the concrete pile body to be tested.

7. The NLP-based intelligent determination system for concrete pile integrity level is characterized by: include: The data acquisition module is used to detect the concrete pile body to be tested by using the acoustic wave transmission method, and obtain the acoustic parameters of the concrete pile body to be tested and receive waveform signal data; The grade determination module is used to input the acoustic parameters of the concrete pile body to be tested and the received waveform signal data into the trained integrity grade determination model, and output the integrity grade of the concrete pile body to be tested; wherein the training of the integrity grade determination model includes: Identify the corpus of several testing technical regulations corresponding to the integrity test of concrete piles, and obtain the acoustic parameters and waveform feature description information corresponding to the integrity level; Extracting the judgment object and its characteristic keyword set based on the acoustic parameters and waveform feature description information corresponding to the integrity level, and simultaneously counting the fuzzy quantitative characteristic keywords; Based on the pre-constructed quantitative discrimination formula for state characteristics corresponding to acoustic parameters, the quantitative discrimination formula for distribution characteristics corresponding to acoustic parameters, and the trained reception waveform state characteristic classification model, the extracted judgment objects and their characteristic keyword sets are quantified to form an integrity level judgment model; the training of the reception waveform state characteristic classification model includes: obtaining historical data of concrete pile body detection; constructing a training sample set and a test sample set based on the historical data, and configuring a classification label for each waveform data in the sample set; constructing a reception waveform state characteristic classification model, including a CNN model for extracting high-level features of the reception waveform data, and an LSTM model for capturing the timing information of the reception waveform data; using the training sample set to train the reception waveform state characteristic classification model, and using the test sample set to test it, to obtain a trained reception waveform state characteristic classification model; A training sample set for integrity level determination is constructed, an integrity level determination model is trained using the training sample set, and fuzzy quantitative feature keywords are used as adjustment parameters to obtain a trained integrity level determination model.

Images