Pile driving resistance inversion and stratum mutation identification method based on sparse dynamics identification

Abstract

The application discloses a method for pile sinking resistance inversion and stratum mutation identification based on sparse dynamics identification, which obtains pile sinking force or hammering energy, penetration, penetration speed, pile top resilience, pore water pressure and geological survey parameters in the pile sinking process, and uniformly maps multiple source construction data to a pile end depth coordinate to form a standardized depth sequence; a pile sinking state vector containing historical response and depth lag characteristics is constructed based on the sequence, and a candidate dynamics function library facing the pile-soil interaction mechanism is established; sparse dynamics identification is carried out in different depth sections, equivalent pile sinking resistance is inverted and is decomposed into side friction resistance, end resistance and abnormal resistance contribution; further, the stratum mutation position and type are identified according to the structural change of adjacent section sparse equations, and construction parameter correction suggestions are output. The application can be used for bearing stratum identification, abnormal resistance diagnosis and pile sinking construction optimization control.

Description

Technical Field

[0001] This invention relates to the field of geotechnical engineering technology, specifically to a method for pile driving resistance inversion and ground abrupt change identification based on sparse dynamics identification. Background Technology

[0002] Pile driving is a crucial step in pile foundation engineering. During the penetration process, the pile is affected by multiple factors, including pile side friction, pile end resistance, pore water pressure, soil disturbance recovery, construction pause effects, and changes in hammer energy or pile driving force. For large-diameter piles, long piles, or pile driving in complex strata, when the pile tip enters a hard interlayer, weak interlayer, bearing layer, or abnormal resistance zone, the construction response will manifest as abrupt changes in penetration depth, increased hammer blows, changes in pile driving force, abnormal pile top rebound, or abnormal pore pressure response.

[0003] Existing pile driving analysis technologies mainly focus on the following aspects: First, monitoring construction and assisting in stopping hammering by measuring penetration depth, hammer blow count, pile driving force, or final penetration depth; second, calculating pile side friction, pile end resistance, and bearing capacity parameters based on CPT, CPTU, or static cone penetration test parameters; third, realizing data acquisition, real-time display, and construction control of the pile driving process through digital twins or construction monitoring platforms; and fourth, evaluating the quality or feasibility of pile driving construction using data-driven models such as neural networks.

[0004] The above technologies can achieve pile driving process monitoring, penetration identification, and bearing parameter calculation to a certain extent, but they still have the following shortcomings: First, existing methods mostly rely on penetration depth, number of hammer blows, pile driving force, or resistance threshold as the basis for judgment, which makes it difficult to reveal the changes in the dominant mechanism of pile-soil interaction during pile driving. Second, existing methods mostly rely on geological parameters or empirical formulas to calculate pile side friction and pile end resistance, making it difficult to use construction response data to invert the dynamic changes of actual pile driving resistance along depth. Third, existing methods for judging geological changes often rely on a single threshold or empirical rules, which are easily affected by fluctuations in construction energy, equipment status, pile top rebound, increased resistance during shutdowns, and changes in pore pressure. Fourth, while general machine learning or general dynamic identification methods can fit construction data, if they lack a candidate function library and engineering constraints for pile driving resistance mechanisms, they are prone to black box predictions and cannot explain the contributions of side friction, end resistance and abnormal resistance.

[0005] Therefore, it is necessary to propose a sparse dynamic identification method for the pile driving process, which combines the pile driving construction response, pile-soil interaction mechanism and ground abrupt change identification. The method identifies ground abrupt changes by changing the equation structure rather than by changing a single threshold, and outputs resistance inversion results and parameter correction suggestions that can be used for construction control. Summary of the Invention

[0006] Based on the above description, this invention provides a method for pile driving resistance inversion and ground abrupt change identification based on sparse dynamics identification, in order to solve the problems that existing pile driving monitoring methods mainly rely on penetration depth, hammer blows or pile driving force threshold, which are difficult to explain the source of pile driving resistance, difficult to identify changes in the dominant mechanism of pile-soil interaction, and difficult to accurately determine the type of ground abrupt change.

[0007] The technical solution of the present invention to solve the above-mentioned technical problems is as follows: The method for pile driving resistance inversion and ground abrupt change identification based on sparse dynamics identification includes the following steps: S1. Obtain multi-source construction response data during the pile driving construction process. The multi-source construction response data includes at least the pile driving force or hammer energy, penetration per blow, cumulative penetration depth, penetration speed, pile top rebound, pile strain, pore water pressure, and geological survey parameters. All data are appended with construction event numbers, timestamps, and pile end depth indexes. S2. Perform anomaly removal, missing data repair, and depth alignment processing on multi-source construction response data, and convert time series data and hammer impact event data into standardized depth series data with pile tip depth as the independent variable, so that all variables are expressed under a unified depth coordinate system. S3. Construct a pile driving state vector based on the standardized depth sequence, which includes the current construction response, historical depth window response and depth lag response, to characterize the historical dependence and resistance accumulation characteristics of pile-soil action. S4. Construct a candidate dynamic function library for pile driving resistance oriented towards the pile-soil interaction mechanism. The function library includes construction input items, penetration response items, side friction accumulation items, end resistance mutation items, pore pressure response items, and stratum segmentation indicator items. S5. Divide the pile into multiple depth sections according to the pile tip depth, and use the sparse regression method to identify the dynamics in each depth section to obtain the sparse expression equation of the pile driving resistance in each section. S6. Based on the sparse expression equation, the equivalent pile driving resistance is inverted and decomposed into side friction contribution, end resistance contribution and abnormal resistance contribution. S7. Identify abrupt changes in formation locations based on sparse equation structure changes in adjacent depth sections; S8. Output construction parameter correction strategies based on the type of geological mutation to optimize construction control.

[0008] Based on the above technical solution, the present invention can be further improved as follows.

[0009] Furthermore, the data cleaning and depth alignment in step S2 includes: outlier detection and removal for pile driving force, hammer impact energy, penetration depth, pore water pressure, and pile strain; repair of missing data using local window interpolation or neighborhood reconstruction; and construction of a recursive relationship for pile tip depth based on the construction event sequence. Simultaneously, time series data, hammer impact event data, and geological exploration layered data are uniformly mapped to a depth grid, aligning data from different sampling frequencies within the same depth coordinate system. This alignment process is achieved by combining penetration depth accumulation or pile driving displacement integration, transforming the construction response data from discrete records into a continuous depth function expression to ensure the physical consistency and numerical stability of subsequent resistance inversion.

[0010] Furthermore, the pile driving state vector in step S3 is composed of current response characteristics, historical response characteristics, and depth hysteresis characteristics. The current response includes at least pile driving force, hammer energy, penetration velocity, penetration acceleration, rebound amount, and pore water pressure. The historical response is composed of multiple adjacent depth windows to describe the resistance evolution path. The depth hysteresis response is used to characterize the pile side friction accumulation and pore pressure dissipation hysteresis effect, so that the pile driving state vector can simultaneously express the instantaneous construction state, historical loading influence, and long-term stratum action mechanism.

[0011] Furthermore, the candidate dynamic function library in step S4 is a set of constraint functions for pile-soil interaction mechanisms, including construction input function terms, unit penetration energy consumption function terms, penetration velocity decay function terms, side friction accumulation function terms, end resistance mutation function terms, pore pressure change function terms, and quiescent resistance increase function terms. Each function term corresponds to the construction input effect, pile side friction growth mechanism, pile end resistance mutation mechanism, and pore pressure influence mechanism, respectively, and is used to replace the traditional unconstrained feature space, so that the dynamic identification results have clear engineering physical meaning.

[0012] Furthermore, the side friction accumulation function is constructed by weighted accumulation of pile-soil contact response within historical depth intervals, and is used to describe the memory effect of the gradual increase of pile side friction with soil penetration depth; the end resistance mutation function is constructed based on depth gradient change and stratum prior information, and is used to characterize the characteristics of sudden increase in resistance when the pile tip enters hard interlayer or bearing layer; the pore pressure change function is used to describe the generation, accumulation and dissipation process of pore water pressure in saturated soil, and is used to correct the pile tip resistance inversion deviation.

[0013] Furthermore, in step S5, a sparse regression method is used to identify the dynamics of each depth segment. The sparse regression method includes L1 regularized regression, sequential threshold least squares or elastic network method, and low confidence data is suppressed by introducing a weight matrix, so as to reduce the impact of abnormal data or missing data on the model coefficients, thereby improving the stability and robustness of the sparse equation under complex construction conditions.

[0014] Furthermore, the depth segmentation in step S5 adopts a fixed depth window, an adaptive resistance variation window, or a method based on geological stratification boundaries. When there is a sudden change in penetration resistance or unit penetration energy consumption, the window length is automatically reduced to enhance the ability to distinguish areas of abrupt changes in strata and improve the identification accuracy in the boundary areas of soft soil, hard interlayers, and bearing layers.

[0015] Furthermore, in step S6, the equivalent pile driving resistance is obtained by weighted reconstruction of the selected dominant function term and its corresponding coefficient in the sparse expression equation, and is divided into side friction contribution, end resistance contribution and abnormal resistance contribution according to the physical properties of the function term. The abnormal resistance contribution includes at least pore pressure anomaly, rebound anomaly and pause recovery effect, thereby realizing the interpretable decomposition of the source of pile driving resistance.

[0016] Furthermore, the formation abrupt change identification in step S7 is based on the joint determination of the degree of change of the set of dominant function terms of the sparse equation between adjacent depth segments, the magnitude of the change of the sparse coefficient, the change of the resistance gradient, and the change of the residual. In this way, the changes in the dynamic mechanism of different depth segments are quantified by the structural difference index, avoiding the reliance on a single threshold of penetration or pile driving force, thereby improving the identification stability under complex formation conditions.

[0017] Furthermore, the construction parameter correction strategies output in step S8 include final pressure correction, final penetration correction, pile driving speed control, hammer energy adjustment, hammer stop standard correction, and re-pressure and re-driving construction strategy optimization. Based on the identified stratum type, differentiated construction control suggestions are output for hard interlayers, soft interlayers, bearing layer entry, and abnormal resistance zones, thereby achieving closed-loop optimization control and dynamic adjustment of the pile driving construction process.

[0018] Compared with the prior art, the technical solution of this application has the following beneficial technical effects: 1. This invention maps multi-source data such as pile driving force, hammering energy, penetration depth, penetration speed, pile top rebound, pile strain, pore water pressure and geological exploration parameters to the pile end depth coordinates, thereby realizing a unified expression of pile driving construction response data and improving the matching accuracy between data from different sources and the reliability of subsequent resistance inversion. 2. This invention constructs a candidate dynamic function library for pile driving resistance oriented towards the pile-soil interaction mechanism, and incorporates engineering features such as unit penetration energy consumption, side friction accumulation, end resistance abrupt change, pore pressure dissipation, intermittent resistance increase and stratum segmentation indication into the identification process, thus avoiding the problem of lack of engineering interpretability in general algorithm models; 3. The present invention adopts a layered sparse dynamic identification method to screen the dominant function terms that control the change of pile driving resistance in different depth sections. This can reflect the differences in the pile driving resistance mechanism in different soil layers and avoid the averaging of local stratum change information caused by uniform fitting across the entire depth. 4. This invention can invert the equivalent pile driving resistance based on the sparse dynamic equation and decompose it into equivalent side friction contribution, equivalent end resistance contribution and abnormal resistance contribution, thereby determining whether the difficulty in pile driving mainly comes from increased pile side friction, enhanced pile end resistance or construction abnormalities, thus improving the interpretability of the pile driving process analysis. 5. This invention identifies the location of abrupt changes in strata by combining changes in sparse equation structure, sparse coefficient, resistance gradient, and inversion residual, avoiding reliance on a single threshold such as penetration depth, hammer blow count, or pile driving force for judgment, thereby improving the accuracy and stability of identifying hard interlayers, weak interlayers, bearing layer entry, and abnormal resistance zones. 6. This invention can output final pressure correction, final penetration correction, hammer stop standard adjustment, pile driving speed adjustment, hammer impact energy adjustment, re-pressing and re-driving or supplementary geological exploration suggestions based on the type of abrupt change in strata and the resistance contribution results. It realizes closed-loop control from construction data analysis to construction parameter correction, and improves the quality of pile driving construction and the efficiency of on-site decision-making. Attached Figure Description

[0019] Figure 1 A flowchart of the pile driving resistance inversion and formation change identification method based on sparse dynamics identification provided in this embodiment of the invention; Figure 2 This is a flowchart of multi-source response data acquisition for pile driving construction provided in an embodiment of the present invention; Figure 3 This is a flowchart illustrating the multi-source construction response data cleaning, synchronization, and pile tip depth alignment provided in this embodiment of the invention. Figure 4 A flowchart illustrating the construction of pile driving state vectors and the generation of candidate dynamic function libraries provided in this embodiment of the invention; Figure 5 A closed-loop flowchart for layered sparse dynamics identification, resistance inversion, and formation abrupt change identification provided in this embodiment of the invention; Figure 6 This is a schematic diagram illustrating an example of data acquisition for static pressure pile driving and hammer pile driving construction response provided in an embodiment of the present invention; Figure 7 A schematic diagram illustrating the unified mapping of time-series data, hammer impact event data, and geological exploration stratification data to pile tip depth coordinates for embodiments of the present invention; Figure 8 A schematic diagram of the pile driving state vector structure, including current response, historical response, deep hysteresis response, and local change characteristics, provided for embodiments of the present invention; Figure 9 This is a schematic diagram of the interface for entering project and pile foundation information and displaying pile-soil layer profiles provided in an embodiment of the present invention. Figure 10This is a schematic diagram of the interface for importing multi-source pile driving construction response data, evaluating quality, and displaying raw data provided in an embodiment of the present invention. Figure 11 The data curves showing the changes in pile driving force, hammer energy, penetration depth, and pile top rebound as a function of pile end depth during the pile driving process are provided for embodiments of the present invention. Figure 12 The inversion results of equivalent pile driving resistance, side friction contribution, end resistance contribution, and abnormal resistance contribution as a function of pile tip depth are provided in the embodiments of the present invention. Figure 13 Statistical charts of changes in sparse dominant function terms, sparse coefficients, and structural difference indices in adjacent depth segments provided in embodiments of the present invention; Figure 14 Statistical charts showing the comprehensive index of formation mutation, mutation confidence, and mutation type determination results provided in embodiments of the present invention; Figure 15 This is a computational power analysis diagram showing the number of partitions at different depths, the size of the candidate function library, the time consumed by sparse identification, and the computational resource consumption for embodiments of the present invention. Detailed Implementation

[0020] To facilitate understanding of this application, a more complete description will be provided below with reference to the accompanying drawings, which illustrate embodiments of the present application. However, the present application can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided so that the disclosure of this application will be more thorough and complete.

[0021] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the application.

[0022] Example 1: Reference Figures 1-15 A method for pile driving resistance inversion and ground abrupt change identification based on sparse dynamics identification includes the following steps: S1. Obtain pile driving construction response data

[0023] Multi-source construction response data is acquired during pile driving construction, and the correspondence between construction events, time, and pile tip depth is established. The construction response data includes construction equipment input data, pile penetration response data, pile-soil interaction response data, and prior ground data. The construction equipment input data includes the driving force, final pressure, stabilization time, number of re-driving times, and driving speed during static pressure pile driving, or the hammer energy, number of blows, hammer frequency, hammer weight, drop distance, and energy per blow during hammer driving. The pile penetration response data includes the cumulative penetration depth, penetration per blow, penetration per unit time, final penetration, pile top rebound, pile acceleration, and pile inclination angle. The pile-soil interaction response data includes pile strain, pore water pressure around the pile, soil disturbance response around the pile, rest time, and resistance recovery response after re-driving or re-driving. The stratum prior data includes the soil layer name, soil layer thickness, groundwater level, standard penetration test blow count, static cone tip resistance, side friction resistance, pore pressure parameters, and design bearing layer depth from the engineering geological survey report.

[0024] For static pressure pile driving, time-sampling sequences are used as the original recording units to collect continuous data such as pile driving force, penetration depth, and pile driving speed. For hammer-driven pile driving, a single hammer event is used as the original recording unit to collect data such as hammer energy per blow, penetration per blow, pile top rebound, and cumulative number of blows. To ensure that data from different sources can be used for subsequent unified modeling, each data record is appended with a timestamp, construction event number, pile number, pile type, pile diameter, pile length, and construction equipment number during the data collection phase, so that the construction process data can correspond one-to-one with the specific pile body and specific depth location.

[0025] The collected raw dataset can be represented as: In the formula, For pile driving force, For the first Energy from each hammer blow For the first Penetration per hammer blow The depth of the pile tip. For penetration speed, For penetration acceleration, This refers to the rebound amount at the pile top. For the strain of the pile body, Pore ​​water pressure, This refers to the interval between rest or construction periods. These are geological parameters that vary with depth.

[0026] For hammer-driven piles, the pile tip depth can be obtained by summing the penetration depth per blow: ;in, This is the initial pile tip depth. For the first The penetration depth corresponding to each hammer blow. For static pressure pile driving, the penetration velocity and penetration acceleration can be calculated from the depth-time series: This provides energy characteristics for subsequent pile driving resistance inversion, and further calculates the unit penetration energy consumption. ;in, For the first Energy consumption per unit penetration per hammer blow To prevent instability in calculations due to excessively small penetration depth, a small positive number is used. A larger unit penetration energy consumption indicates that the construction energy required per unit depth of penetration is higher, which usually corresponds to situations such as increased pile end resistance, entry into hard interlayers, entry into dense layers, or approach to the bearing layer.

[0027] Through the above data acquisition methods, this step not only collects single penetration depth or single pile driving force data, but also forms a multi-source pile driving construction dataset that includes construction input, pile body response, pore pressure response, rebound response and stratum prior information, providing a complete data foundation for subsequent depth alignment, state vector construction, pile driving resistance candidate dynamic function library construction and hierarchical sparse dynamic identification.

[0028] S2. Perform construction data cleaning and depth alignment.

[0029] The raw pile driving response data obtained in step S1 is cleaned, synchronized, and depth aligned to unify data with different sampling frequencies, recording formats, and physical meanings into a standardized sequence indexed by pile tip depth. Since data such as pile driving force, penetration depth, pore water pressure, and pile strain are typically collected continuously over time, while data such as hammer energy, penetration per blow, and pile top rebound are typically recorded discretely based on hammer events, and geological parameters are given according to soil depth intervals, it is necessary to first eliminate the differences in data recording scales before forming depth sequence data that can be used for sparse dynamics identification.

[0030] First, the original construction response data is screened for validity. For variables such as pile driving force, hammer energy, penetration depth, pile top rebound, pore water pressure, and pile strain, data points that are due to sensor disconnection, equipment startup impact, short-term invalid records, duplicate records, or data points that significantly exceed the physical reasonable range are removed. For data segments with short missing periods, neighborhood interpolation can be used to fill in the gaps; for data segments with long missing periods or continuous anomalies, they are marked as low-confidence segments and do not directly participate in subsequent sparsity identification. Outliers can be identified by combining local median and local fluctuation scale, and the judgment criteria can be expressed as: In the formula, For the data points to be judged, For its neighborhood window, The window value is the median. The absolute median difference within the window. This is the anomaly detection coefficient. Data points that meet the above conditions are marked as outliers, and the appropriate handling method is selected based on the variable type: removal, interpolation, or retention with reduced weight.

[0031] Secondly, a unified pile tip depth grid is established. Based on the actual pile driving depth range and data resolution, the pile tip depth is divided into depth nodes with equal or adaptive spacing: ;in, A standard depth coordinate sequence, This represents the number of depth nodes. The depth step size is used. For sections with relatively gentle soil layer changes, a larger depth step size can be used; for sections with abrupt penetration response, sudden increase in hammer energy consumption, or close to the design bearing layer, a smaller depth step size can be used to preserve information on abrupt changes in the strata.

[0032] Then, data from different sources are mapped to a unified depth node. For time-continuous data, it is converted into a depth function based on the time-depth relationship; for event-discrete data, it is assigned to the corresponding depth node based on the pile tip depth corresponding to the hammer impact event; for geological stratification data, soil layer type, SPT blow count, cone tip resistance, and side skin resistance are assigned to the corresponding depth interval based on the upper and lower boundary depths of the soil layer. For any construction response variable y, its aligned value at the standard depth node can be expressed as: In the formula, For variables At deep nodes Alignment results at the location, This represents the original record coordinates, which can be time, hammer strike number, or original depth. This represents interpolation, neighbor merging, or interval assignment operators.

[0033] Furthermore, the aligned depth sequence is smoothed and labeled with confidence level. For continuous responses such as pile driving force, pore pressure, and strain, a sliding window smoothing method can be used to reduce short-term noise; for event responses such as penetration per blow and rebound, abrupt changes are preserved to avoid over-smoothing that could weaken information about hard interlayers or bearing layers. Simultaneously, a data confidence level label is assigned to each depth node, with confidence level determined by sensor effectiveness, missing data ratio, outlier ratio, and consistency of multi-source data. Depth Node The reliability of the data at this location can be expressed as: In the formula, As an indicator of sensor effectiveness, For missing data integrity indicators, As a consistency indicator for multi-source data, , , These are the weighting coefficients.

[0034] After the above processing, the pile driving force, hammer impact energy, penetration response, rebound response, pore pressure response, strain response, resting information, and geological parameters are all uniformly expressed in the pile tip depth coordinates, forming standardized depth sequence data. The purpose of this step is not simply to organize construction records, but to transform the raw monitoring data into a unified input that can be used for pile driving state vector construction and layered sparse dynamic identification. This ensures that subsequent resistance inversion results accurately correspond to specific depth locations and provides a reliable data foundation for identifying abrupt changes in strata.

[0035] S3. Construct the pile driving state vector

[0036] After completing the cleaning, synchronization, and depth alignment of the construction data, a pile driving state vector for pile driving resistance inversion is constructed using standard depth nodes as the basic modeling unit. This state vector is not only composed of the construction response at the current depth point, but also incorporates the current response, historical response, depth lag response, and local variation characteristics to characterize the path dependence, resistance accumulation, and sensitivity to abrupt changes in the soil layer during pile driving.

[0037] In the At each standard depth node, first construct the current response feature vector: In the formula, , , , , , , , , , These represent the pile driving force, hammer energy, penetration depth, penetration velocity, penetration acceleration, pile top rebound, pile strain, pore water pressure, dwell time, and geological parameters after depth alignment and standardization. For data not collected under a certain construction method, null value marking, confidence reduction weighting, or supplementation with parameters from similar projects can be used without changing the overall structure of the state vector.

[0038] To reflect phenomena such as the gradual accumulation of pile side friction, the dissipation of pore pressure after hysteresis, and the recovery of soil strength after a period of inactivity, historical response characteristics are introduced based on the current response characteristics. Historical response characteristics are composed of the responses of several depth nodes preceding the current depth node, and can be expressed as: ;in, The historical window length can be determined based on the pile diameter, soil layer thickness, data sampling interval, and construction response change rate. When the soil layer is relatively uniform and the construction response is stable, a shorter historical window can be used; when the strata change frequently or the pile driving resistance fluctuates greatly, the historical window can be appropriately increased to enhance the model's ability to express the resistance evolution process.

[0039] Meanwhile, to highlight the impact of abrupt changes in formation on the construction response, local variation characteristics are further constructed. These local variation characteristics include increments in pile driving force, unit penetration energy consumption, penetration velocity, rebound, pore pressure, and geological parameters, used to describe the degree of abrupt changes in response between adjacent depth nodes. They can be expressed as: ;when When the pile tip continuously increases within a local depth range, it usually indicates a change in the pile driving resistance mechanism, such as the pile tip approaching a hard interlayer, entering a dense layer, rapid accumulation of pore pressure, or abnormal energy transfer during construction. This change characteristic can subsequently be used in conjunction with changes in the sparse equation structure for the identification of abrupt changes in the formation.

[0040] Finally, the current response characteristics, historical response characteristics, local change characteristics, and data reliability are combined to form the pile driving state vector: ;in, For the first The pile driving state vector corresponding to each depth node. This serves as the data credibility label for that depth node. (This is achieved by introducing...) In the subsequent sparse identification process, the influence of low-reliability data on the equation coefficients can be reduced, and the distortion of resistance inversion results caused by sensor anomalies, data loss or construction record errors can be avoided.

[0041] The pile driving state vector constructed in the above manner can simultaneously express construction input, pile motion response, pile-soil interaction response, prior information of the strata, and local variation trends. This step provides a unified variable basis for the subsequent construction of the candidate dynamic function library, enabling sparse identification to move beyond a single penetration depth or a single pile driving force change, and instead to screen out the dominant factors that truly affect pile driving resistance from multi-dimensional construction states, thereby improving the interpretability of pile driving resistance inversion and strata abrupt change identification.

[0042] S4. Construct a candidate dynamic function library for pile driving resistance.

[0043] Based on the pile driving state vector obtained in step S3, a candidate dynamic function library for pile driving resistance is constructed, oriented towards the pile-soil interaction mechanism. This function library does not simply employ general polynomials, trigonometric functions, or black-box feature combinations. Instead, it is built around engineering mechanisms such as construction inputs, penetration response, pile side friction accumulation, abrupt changes in pile end resistance, pore pressure changes, intermittent resistance increases, and stratum segmentation during pile driving. This ensures that the equations obtained through subsequent sparse identification have clear physical meaning and construction interpretation value.

[0044] In the At each depth node, the candidate dynamics function library is represented as: In the formula, For depth nodes Candidate function libraries at the location, This is the pile driving state vector. For the first 1 candidate function item This represents the total number of candidate function terms. Each category of candidate function terms corresponds to an engineering mechanism that may affect the change in pile driving resistance. Subsequently, the dominant term is automatically selected through sparse identification.

[0045] First, construct the construction input function terms. These functions characterize the driving effect of external construction equipment on the pile penetration process, including pile driving force response, hammer impact energy response, hammer impact frequency response, pressure stabilization time response, and repeated driving and pressure application response. For static pressure pile driving, focus on retaining terms related to pile driving force, pile driving speed, and pressure stabilization time; for hammer-driven pile driving, focus on retaining terms related to hammer impact energy, number of hammer blows, rebound amount, and final penetration depth. These functions distinguish between resistance changes caused by variations in construction input and resistance changes caused by variations in the strata themselves.

[0046] Secondly, a penetration response function term is constructed. This function term describes the relationship between the pile's motion state and the pile driving resistance, including a penetration velocity attenuation term, a penetration acceleration change term, a penetration difficulty per unit depth term, and a local penetration response fluctuation term. When the pile tip enters a dense layer, hard interlayer, or bearing layer, it typically exhibits a decrease in penetration velocity, an increase in penetration difficulty per unit depth, and enhanced local variation characteristics. This function term provides a basis for subsequently identifying sections with increased resistance.

[0047] Next, we construct a class function for the accumulation of pile side friction. Pile side friction typically accumulates gradually with increasing penetration depth and is influenced by soil type, pile perimeter disturbance, and historical penetration process. To express this accumulation effect, a side friction memory term of the following form can be constructed: In the formula, For depth The side friction cumulative function term at the location, For the first Local response functions related to side friction at depth nodes This is a weight that decreases with depth. This term reflects the cumulative contribution of the portion of the pile already embedded in the soil to the current pile driving resistance.

[0048] Then, a function term representing abrupt changes in pile end resistance is constructed. Pile end resistance is highly sensitive to abrupt changes in formation; when the pile tip enters a hard interlayer, dense sand layer, gravel layer, or the design bearing layer, the end resistance contribution typically increases rapidly. To avoid misclassifying ordinary noise as abrupt changes in formation, the end resistance abrupt change term is not directly constructed using a single-point threshold, but rather by combining local resistance surrogate, depth gradient, and formation priors. For example, an end resistance sensitivity term can be defined: In the formula, The local resistance proxy formed by construction input and penetration response, Its rate of change along the depth direction, This is a priori indication of whether the layer is hard, dense, or bearing. This item is used to highlight structural changes when the pile tip enters a high-resistance stratum.

[0049] Furthermore, functions related to pore pressure and cessation effects are constructed. For saturated soft soil or cohesive soil, pile driving causes the accumulation of excess pore water pressure, which dissipates over time. After construction pauses, re-pile driving, or re-driving, the recovery of the soil structure may also lead to increased resistance. Therefore, the function library introduces pore pressure response, pore pressure change, cessation resistance increase, and disturbance recovery terms to identify resistance changes caused by soil time effects, avoiding misinterpreting increased resistance after a pause as the entry of the bearing layer.

[0050] Finally, stratigraphic segmentation and coupling interaction function terms are constructed. The stratigraphic segmentation function term is used to incorporate initial stratigraphic information obtained from geological survey reports, CPT, or CPTU tests; the coupling interaction term is used to express the combined effects between construction inputs and formation conditions, pore pressure response and penetration rate, and rebound response and hammer impact energy. The interaction function term can be represented as: In the formula, and These are two candidate function terms with different physical meanings. The interaction term can be used to describe the complex pile-soil interaction process, which is difficult to express with a single variable.

[0051] After the above construction, the candidate dynamic function library for pile driving resistance forms a dedicated set of functions consisting of construction input terms, penetration response terms, side friction accumulation terms, end resistance mutation terms, pore pressure dissipation terms, paused resistance increase terms, stratum indicator terms, and coupling interaction terms. The core function of this step is to transform the pile driving engineering mechanism into identifiable mathematical function terms, enabling subsequent sparse dynamic identification to filter out the truly dominant resistance mechanism in different depth sections from numerous candidate options, thus distinguishing it from ordinary construction monitoring, ordinary penetration identification, and general algorithm fitting methods.

[0052] S5. Divide the depth segments and perform hierarchical sparse identification.

[0053] Based on the candidate dynamic function library for pile driving resistance constructed in step S4, the entire pile driving process is divided into zones according to the pile tip depth, and sparse dynamic identification is carried out separately in different depth zones to obtain the sparse equations for pile driving resistance corresponding to each zone. The purpose of this step is to avoid the averaging of stratigraphic differences caused by using a uniform model across the entire depth, and to enable the dominant resistance mechanisms in soft soil layers, hard interlayers, dense layers, bearing layers, and abnormal resistance zones to be identified separately.

[0054] First, determine the method for dividing depth zones. Depth zones can be divided according to a fixed depth window, or adaptively based on initial geological stratification, changes in penetration response, locations of sudden increases in unit penetration energy consumption, or changes in data reliability. For zones with relatively homogeneous strata and stable construction response, a longer window can be used; for zones close to the design bearing layer, with rapidly decreasing penetration, abnormally increasing rebound, or significant changes in pore pressure, the window length should be shortened to improve the ability to distinguish local abrupt changes. Each depth segment can be represented as: In the formula, For the first Each depth segment and These represent the start and end depths of the section.

[0055] Secondly, local identification samples are constructed within each depth segment. The pile driving state vector and candidate function terms within that segment are used as inputs, and the equivalent resistance surrogate quantity, resistance increment, or resistance change rate are used as identification targets. For statically driven piles, the identification target can be determined comprehensively by the pile driving force, penetration velocity, and pile response; for hammer-driven piles, the identification target can be determined comprehensively by the hammer energy, penetration per blow, and pile top rebound response. The identification target is not directly equivalent to a single sensor value, but rather serves as a comprehensive response quantity reflecting the difficulty of pile driving and the trend of resistance change in the current depth segment.

[0056] Then, sparsity identification is performed for each depth segment. By selecting a small number of function terms that contribute the most to the change in resistance from the candidate function library, a local pile driving resistance sparse equation is formed. Its solution objective can be expressed as: In the formula, For the first Target identification in each depth segment This is the candidate function library matrix for this section. Let be the sparse coefficients to be solved. The weight matrix is ​​determined by the data reliability. is the sparsity constraint coefficient. This is achieved by introducing... This can reduce the impact of abnormal data, missing data, or low-confidence segments on the identification results.

[0057] Furthermore, the sparse identification results are screened using engineering constraints. Combinations of function terms that do not conform to the pile driving mechanism are eliminated or weighted less. For example, in soft soil sections where the pile end resistance is not significantly enhanced, end resistance abrupt change terms should not be selected solely due to local noise; in sections with long rest periods, rest-increased resistance terms should be retained to participate in the competition; in saturated clay layers, pore pressure response terms should be allowed to enter the sparse equation. Through engineering constraints, the sparse identification results are ensured to meet both the data fitting accuracy requirements and the fundamental laws governing pile-soil interaction.

[0058] Finally, the stability of the sparse equations for adjacent depth sections is checked. If the dominant function terms of adjacent sections are basically consistent, the coefficients change little, and the residual levels are stable, they are considered to belong to the same drag mechanism control zone; if the dominant function terms are replaced, key coefficients increase suddenly, or the residuals shift continuously, then this location is used as a candidate boundary for subsequent stratigraphic abrupt change identification. The differences in equations between adjacent sections can be described by the following indicators: In the formula, For the first The equation difference index between each segment and the previous segment. and These are the sets of dominant function terms selected for adjacent segments. This is a structural difference measurement function. For structural difference weights.

[0059] Through the above-described layered sparsity identification, a set of dominant function terms and sparsity coefficients with engineering significance can be obtained for each depth segment. This step provides an equational basis for subsequent equivalent pile driving resistance inversion, decomposition of side friction and end resistance contributions, determination of abrupt changes in strata depth, and determination of abrupt change type, avoiding the simple mixing of pile driving responses in different strata into a single empirical relationship.

[0060] S6. Invert the equivalent pile driving resistance and decompose the resistance contribution.

[0061] Based on the sparse equations of pile driving resistance for each depth section obtained in step S5, the equivalent pile driving resistance in the depth direction of the pile tip is inverted and further decomposed into side friction contribution, end resistance contribution, and abnormal resistance contribution. This step does not directly use pile driving force, hammer blows, or penetration depth as the final judgment result, but uses the selected dominant function terms and sparse coefficients to convert the multi-source construction response into a resistance inversion curve with engineering interpretation significance.

[0062] First, within each depth segment, the local equivalent pile driving resistance is reconstructed based on the dominant function term obtained from sparse identification. For the ... Depth nodes within a depth segment The equivalent resistance can be expressed as: In the formula, For the first Each depth segment is at depth The inversion equivalent pile driving resistance at the location, This is the set of dominant function terms selected for this segment. For the corresponding sparsity coefficient, This represents the candidate function term. This expression indicates that the inversion drag is determined by a small number of dominant mechanisms, rather than by a black-box fitting of all input variables.

[0063] For statically driven piles, the equivalent resistance is mainly derived from the changes in pile driving force, penetration velocity, steady-state response, pile strain, and pore pressure. When the penetration velocity decreases, the pile driving force continuously increases during the steady-state phase, and the pile strain increases synchronously, the resistance level at the current depth can be considered to be enhanced. If the pore pressure term is selected in the sparse equation, it indicates that the resistance change includes soil disturbance and pore pressure effects. For hammer-driven piles, the distribution of hammer energy among pile top rebound, pile elastic deformation, and soil resistance consumption needs to be considered, and the effective penetration energy can be constructed as follows: In the formula, For the first Effective penetration energy of a single hammer blow To input hammering energy, This refers to the rebound amount at the pile top. This is the rebound energy consumption correction factor. This is a correction factor for equipment and transmission losses. Lower effective penetration energy and smaller penetration volume generally indicate a significant increase in pile-soil resistance or energy loss.

[0064] Secondly, the equivalent resistance obtained from the inversion is attributed to specific factors. Based on the physical meaning of the candidate function terms, the selected dominant terms are divided into side friction, end resistance, and abnormal resistance categories. Side friction mainly includes penetration depth-related terms, historical accumulation terms, and pile side friction memory terms; end resistance mainly includes end resistance mutation terms, resistance gradient terms, bearing layer indicator terms, and unit penetration difficulty enhancement terms; abnormal resistance mainly includes pile top rebound anomaly terms, pore pressure anomaly terms, intermittent resistance increase terms, residual sudden increase terms, and construction disturbance recovery terms. The contributions of each category can be expressed as: In the formula, For the first Resistance-like contribution, , , These represent the contributions of side friction, end resistance, and abnormal drag, respectively. This is the set of function items corresponding to the category.

[0065] Next, calculate the proportion of different resistance contributions in the total equivalent resistance to determine the main sources of pile driving difficulties:

[0066] In the formula, For the first The contribution ratio of resistance-like structures To prevent the denominator from being too small for the stability term. When the contribution of side friction is high and gradually increases with depth, it indicates that the pile driving resistance mainly comes from the accumulation of friction after the pile enters the soil; when the contribution of end resistance increases rapidly at local depths, it indicates that the pile tip may have entered a hard interlayer, dense layer or bearing layer; when the contribution of abnormal resistance increases, it is necessary to combine pore pressure, rebound, rest time and pile posture to determine whether there is construction abnormality, soil disturbance recovery or geological exploration deviation.

[0067] Finally, the local inversion results of each depth segment are stitched together, and the continuity of the resistance curves at the segment boundaries is checked. For adjacent segments with the same dominant mechanism and gradual changes in the inversion results, a smooth connection can be made; for segments with significant changes in the dominant mechanism and synchronous changes in the resistance contribution ratio, the boundary abrupt change characteristics should be retained as an important basis for subsequent identification of stratigraphic abrupt changes. Through this step, the equivalent pile driving resistance curve, side friction contribution curve, end resistance contribution curve, and abnormal resistance contribution curve along the pile tip depth can be obtained, providing a quantitative basis for determining the bearing layer entry, identifying abnormal resistance segments, and adjusting construction parameters.

[0068] S7. Identifying Stratigraphic Abrupt Changes Based on Sparse Equation Structure Changes

[0069] Based on the sparse equations for each depth segment obtained in step S5 and the resistance contribution decomposition results obtained in step S6, the locations of abrupt changes in strata during pile driving are identified. This step's judgment is not based on a single change in penetration depth, single pile driving force, or single hammer blow count, but rather on whether the "equation structure controlling pile driving resistance" changes within adjacent depth segments. When the pile tip is within the same stable stratum, the selected dominant function term, corresponding coefficients, residual distribution, and resistance change trend in the sparse equations usually remain relatively stable; when the pile tip enters a hard interlayer, weak interlayer, dense layer, bearing layer, or abnormal resistance zone, the dominant function term and its contribution ratio will undergo structural changes.

[0070] First, determine the set of dominant function terms within each depth segment. For the... For each depth segment, the function term whose absolute value of the sparse coefficients is greater than a set threshold is defined as the dominant term of that segment: In the formula, For the first The set of dominant function terms for each depth segment For the first in this section The sparse coefficients of each function term, Thresholds are selected based on the primary criteria. This is done by comparing adjacent segments. Changes can indicate whether the pile driving resistance control mechanism has been replaced. For example, if the cumulative side friction term is replaced by the abrupt change in end resistance term, it usually means that the pile end resistance has begun to become the dominant factor.

[0071] Secondly, the structural differences in the equations between adjacent depth segments are calculated. These structural differences can be characterized by the proportion of change in the dominant function term: In the formula, For the first The difference in the dominant term structure between each segment and the previous segment It is a stable term. When The larger the value, the more significant the difference in the dominant function terms selected by sparse identification in adjacent sections, and the higher the possibility of changes in the strata or pile-soil interaction mechanism.

[0072] Then, further comparisons are made of the changes in the amplitude of the sparse coefficients, the changes in the residuals, and the changes in the drag gradient. If the dominant function term is not completely replaced, but the coefficients of the key terms show a sudden increase or decrease, it may also indicate that the pile tip is transitioning from one stratum to another. By combining structural differences, coefficient changes, residual shifts, and drag gradient changes, a comprehensive index of stratum abrupt change can be constructed. In the formula, For the first Comprehensive mutation index for each depth segment This is an indicator of sparsity coefficient variation. As an index for inversion residual changes, As an indicator of resistance gradient change, , , , This is the weighting coefficient. This comprehensive index avoids misjudgments caused by single indicators being affected by fluctuations in construction energy, sensor noise, or short-term equipment disturbances.

[0073] Furthermore, the comprehensive abrupt change index is compared with the local adaptive discrimination criterion. When the comprehensive stratigraphic abrupt change index satisfies the following formula, the corresponding segment boundary is determined as the candidate stratigraphic abrupt change location: In the formula, The length of the current segment ahead is The mean of comprehensive mutation index within the local window, This represents the standard deviation of the comprehensive mutation index within this window. The sensitivity coefficient is used. By employing a local adaptive discrimination criterion, this method can adapt to the response differences under different construction equipment, pile types, and geological conditions.

[0074] Finally, the candidate stratigraphic abrupt change locations are continuously verified. If, near a certain depth boundary, not only does the equation structure change, but the proportions of end resistance, side friction, or anomalous resistance contributions also change synchronously, and this change persists across several adjacent depth nodes, then the location is confirmed as a stratigraphic abrupt change location. If the abrupt change index appears only briefly at a single depth node, and the residual and resistance contributions do not change synchronously, then it is marked as a construction disturbance or data anomaly. Through this step, the depth, intensity, and reliability of the stratigraphic abrupt change can be obtained, providing a basis for subsequent identification of hard interlayers, weak interlayers, bearing layer entry, anomalous resistance zones, or deviation zones in geological survey profiles.

[0075] S8. Determine the type of formation abrupt change and output suggestions for correcting construction parameters.

[0076] Based on the location of the formation abrupt change determined in step S7, and combined with the lateral friction contribution, end resistance contribution, and abnormal resistance contribution obtained in step S6, the type of formation abrupt change is determined, and corresponding construction parameter correction suggestions are output. The purpose of this step is to transform the resistance inversion results and equation structure change results into engineering decision-making information that can directly guide on-site construction, rather than simply outputting the abrupt change depth or curve.

[0077] First, a local discrimination window is established near the depth of each candidate mutation. Let the candidate mutation location be... Then its neighborhood analysis range can be expressed as: In the formula, This serves as a local discrimination window for candidate mutation sites. and These represent the analysis lengths before and after the abrupt change. By comparing the changes in the dominant function term, resistance contribution ratio, and construction response within the window before and after the abrupt change, it is determined whether the abrupt change was caused by normal geological changes, construction anomalies, equipment fluctuations, or geological exploration deviations.

[0078] Secondly, a rule for identifying stratigraphic abrupt change types is constructed. If the end resistance contribution continuously increases after the abrupt change location, the unit penetration difficulty increases, and the end resistance abrupt change term or resistance gradient term becomes the dominant term, it is determined to be a hard interlayer or the entry of a bearing layer; if the equivalent pile driving resistance decreases after the abrupt change location, the penetration velocity recovers, or the penetration volume increases, and the pore pressure term, disturbance recovery term, or weak layer indicator term increases, it is determined to be a weak interlayer; if the resistance increases significantly after a pause, and the pause resistance increase term is selected as the dominant term, it is determined to be a pause resistance increase zone; if there is a continuous inconsistency between the inverted resistance, geological parameters, and construction response, it is determined to be a geological profile deviation zone; if the abnormal resistance contribution suddenly increases and is accompanied by rebound, tilting, or abnormal residual difference, it is determined to be an abnormal resistance zone.

[0079] To improve the stability of mutation type identification, a comprehensive score can be set for each mutation type. Taking the h-th type mutation at the c-th mutation position as an example, its score can be expressed as: In the formula, For the first Mutation type discrimination score, For the first related to this type One discriminant indicator, The corresponding weights are used. Discrimination indicators may include changes in end resistance contribution, side friction contribution, anomalous resistance contribution, dominant function term, residual changes, pore pressure changes, and geological stratification consistency. The type with the highest score that meets the minimum confidence condition is used as the determination result for the current abrupt change location.

[0080] Furthermore, the reliability of the stratigraphic abrupt change determination is calculated. Reliability can be determined jointly by data reliability, abrupt change persistence, and consistency of multiple indicators: In the formula, For the first The reliability of determining the mutation location This represents the mean confidence level of the data within a local window. This is an indicator of the persistence of mutation characteristics in the depth direction. As a consistency indicator among multi-source indicators, , , For weighting coefficients. When At lower levels, the system does not directly provide definitive stratigraphic conclusions, but instead outputs suggestions for verification or supplementary testing.

[0081] Finally, based on the type and reliability of the mutation, suggestions for correcting construction parameters are provided. If the penetration is determined to be into the bearing stratum, suggestions for correcting the final pressure, final penetration depth, hammer stop criteria, stabilization time, or hammer stop conditions are provided. If the penetration is determined to be into a hard interlayer, suggestions for appropriately adjusting hammering energy, controlling pile driving speed, verifying pile stress, and preventing pile damage are provided. If the penetration is determined to be into a weak interlayer, suggestions for extending the stabilization time, re-pressing and re-driving, supplementing bearing capacity verification, or verifying pile length are provided. If the penetration is determined to be into an abnormal resistance zone, suggestions for checking pile verticality, verifying pile location, investigating underground obstacles, adjusting construction energy, or suspending construction for verification are provided. If the penetration is determined to be into a geological profile deviation zone, suggestions for supplementing CPT, CPTU, borehole verification, or correcting the stratigraphic boundary depth are provided.

[0082] The output can be organized as follows: In the formula, For the first The output results for each mutation location, The mutation depth, It is a mutation type. For credibility, , , These represent the contributions of side friction, end resistance, and abnormal resistance near that location, respectively. This step provides suggestions for correcting construction parameters. Through this step, the system can transform the sparse dynamics identification results into executable control information during the pile driving process, realizing a closed-loop analysis from data acquisition, resistance inversion, stratum identification to construction adjustments.

[0083] The working principle of this application is as follows:

[0084] Based on the multi-source construction response data generated throughout the entire pile driving construction process, the equipment input, pile penetration response, pile-soil interaction response and geological exploration prior information during the construction process are unified under the pile tip depth coordinate. By identifying and extracting the dominant factors controlling the change of pile driving resistance in different depth sections through sparse dynamics, the variation law of pile driving resistance along the depth is inverted, and the location and type of abrupt changes in the strata are further identified, ultimately forming suggestions for correcting construction parameters. Specifically, during pile driving, the pile penetrates the soil downwards under the action of driving force or hammer energy. As the pile depth increases, changes occur in pile side friction, pile end resistance, pore water pressure, soil disturbance recovery, pile top rebound, and the effects of construction pauses. These changes are reflected in construction response data such as driving force, hammer energy, penetration depth, penetration velocity, pile top rebound, pile strain, pore water pressure, and final penetration depth. Therefore, this invention first collects the above data and uses it as the basic information for judging pile driving resistance and soil changes. Because different data sources use different sampling methods—for example, pile driving force and pore pressure are usually collected continuously over time, hammer energy and penetration per blow are usually recorded according to hammer events, and geological parameters are given in depth layers—this invention maps all data uniformly to the pile tip depth coordinate system. In this way, each depth location corresponds to a complete set of construction response and prior geological information, so that subsequent analysis no longer treats time or a single hammer blow as an isolated object, but rather uses the actual penetration depth of the pile tip as the unified analysis object. After completing depth alignment, this invention constructs a pile driving state vector. This state vector includes not only the pile driving force, hammer energy, penetration velocity, rebound, pore pressure, strain, and geological parameters at the current depth, but also historical response and depth hysteresis response to describe phenomena such as pile side friction accumulation, pore pressure hysteresis dissipation, resistance recovery after pause, and the continuous impact of construction disturbances. In this way, the pile driving process is expressed as a dynamic process that evolves with depth, rather than a simple process of monitoring penetration or pile driving force. This invention further constructs a dedicated candidate dynamic function library for pile driving resistance inversion. This function library is not a simple combination of mathematical functions, but rather is structured around the pile-soil interaction mechanism, including construction input terms, unit penetration energy consumption terms, penetration velocity attenuation terms, pile top rebound energy consumption terms, side friction accumulation terms, end resistance abrupt change terms, pore pressure accumulation and dissipation terms, pause-induced resistance increase terms, construction disturbance recovery terms, and stratum segmentation indicator terms. Through these candidate terms, the engineering mechanism in the pile driving process can be transformed into identifiable mathematical expressions, enabling subsequent models to determine which mechanism primarily controls the current resistance change. Since pile driving typically traverses multiple soil layers, and the resistance composition and dominant mechanisms differ across these layers, this invention does not employ a uniform fitting method across the entire depth. Instead, it divides the soil into zones based on pile tip depth and performs sparse dynamic identification within each zone. The purpose of sparse identification is to filter out a few truly dominant function terms from a large number of candidate function terms, forming sparse equations for pile driving resistance in the corresponding depth zones. This allows us to derive the dominant resistance mechanisms within different depth zones. For example, one zone may be dominated by accumulated side friction, another by abrupt changes in end resistance, and yet another may be significantly affected by pore pressure or intermittent resistance increases. After obtaining the sparse dynamic equations for each depth range, this invention inverts the equivalent pile driving resistance based on the selected dominant function terms and corresponding coefficients, and further decomposes the inversion results into equivalent side friction contribution, equivalent end resistance contribution, and abnormal resistance contribution. Side friction contribution mainly reflects the accumulation of friction due to increased pile penetration depth; end resistance contribution mainly reflects the increased resistance generated when the pile tip enters a hard interlayer, dense layer, or bearing layer; abnormal resistance contribution mainly reflects abnormal factors such as abnormal pile top rebound, abnormal pore pressure, intermittent resistance increase, construction disturbance, or geological exploration deviations. Through resistance decomposition, the main sources of pile driving difficulties can be identified, rather than simply relying on a decrease in penetration depth or an increase in pile driving force. The core of this invention for identifying abrupt changes in formation does not rely on a single threshold, but rather on changes in the structure of the sparse equation. Within the same stable formation, the dominant function term, sparsity coefficient, resistance contribution ratio, and inversion residual controlling pile driving resistance are typically relatively stable. However, when the pile tip enters a hard interlayer, weak interlayer, bearing layer, or abnormal resistance zone, the selected dominant term in the sparse equation changes, and the resistance contribution ratio also changes accordingly. For example, when entering a hard interlayer or bearing layer, the end resistance abrupt change term and the unit penetration energy consumption term will increase; when entering a weak interlayer, the penetration velocity, pore pressure, or disturbance recovery related terms may change significantly; when construction anomalies occur, the rebound anomaly term, residual term, or abnormal resistance contribution will increase. By combining this judgment of changes in equation structure and resistance contribution, the location of abrupt changes in formation can be identified more accurately. After determining the location of abrupt changes in the formation, this invention further combines end resistance contribution, side friction contribution, abnormal resistance contribution, geological survey profile changes, and construction response changes to determine the type of abrupt change. Abrupt change types may include hard interlayers, weak interlayers, bearing layer entry, abnormal resistance zones, intermittent resistance-increasing zones, and geological survey profile deviation zones, etc. Subsequently, the system outputs construction parameter adjustment suggestions based on different abrupt change types, such as correcting the final pressure, adjusting the final penetration depth or hammer stop criteria, adjusting hammering energy, controlling pile driving speed, extending the pressure stabilization time, performing re-pressure and re-driving, checking pile verticality, verifying pile location, or suggesting supplementary penetration tests and geological surveys for verification. In summary, the working principle of this invention can be summarized as follows: pile driving construction data is transformed into a dynamic response sequence under depth coordinates; the pile-soil interaction mechanism is transformed into a dedicated candidate dynamic function library; the dominant resistance mechanism in different depth sections is extracted through layered sparse identification; and then the pile driving resistance is inverted from the dominant mechanism to identify abrupt changes in the strata. This method achieves a closed-loop analysis from construction data acquisition, depth alignment, state representation, resistance inversion, identification of abrupt changes in the strata, to correction of construction parameters, which can improve the accuracy and interpretability of pile driving resistance interpretation, bearing layer entry judgment, and construction control in complex strata.

[0085] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for pile driving resistance inversion and ground abrupt change identification based on sparse dynamics identification, characterized in that, Includes the following steps: S1. Obtain multi-source construction response data during the pile driving construction process. The multi-source construction response data includes at least the pile driving force or hammer energy, penetration per blow, cumulative penetration depth, penetration speed, pile top rebound, pile strain, pore water pressure, and geological survey parameters. All data are appended with construction event numbers, timestamps, and pile end depth indexes. S2. Perform anomaly removal, missing data repair, and depth alignment processing on multi-source construction response data, and convert time series data and hammer impact event data into standardized depth series data with pile tip depth as the independent variable, so that all variables are expressed under a unified depth coordinate system. S3. Construct a pile driving state vector based on the standardized depth sequence, which includes the current construction response, historical depth window response and depth lag response, to characterize the historical dependence and resistance accumulation characteristics of pile-soil action. S4. Construct a candidate dynamic function library for pile driving resistance oriented towards the pile-soil interaction mechanism. The function library includes construction input items, penetration response items, side friction accumulation items, end resistance mutation items, pore pressure response items, and stratum segmentation indicator items. S5. Divide the pile into multiple depth sections according to the pile tip depth, and use the sparse regression method to identify the dynamics in each depth section to obtain the sparse expression equation of the pile driving resistance in each section. S6. Based on the sparse expression equation, the equivalent pile driving resistance is inverted and decomposed into side friction contribution, end resistance contribution and abnormal resistance contribution. S7. Identify abrupt changes in formation locations based on sparse equation structure changes in adjacent depth sections; S8. Output construction parameter correction strategies based on the type of geological mutation to optimize construction control.

2. The method for pile driving resistance inversion and ground abrupt change identification based on sparse dynamics identification according to claim 1, characterized in that: The data cleaning and depth alignment in step S2 includes: outlier detection and removal for pile driving force, hammer impact energy, penetration depth, pore water pressure, and pile strain; repair of missing data using local window interpolation or neighborhood reconstruction; and construction of a recursive relationship for pile tip depth based on the construction event sequence. Simultaneously, time series data, hammer impact event data, and geological exploration layered data are uniformly mapped to a depth grid, aligning data from different sampling frequencies within the same depth coordinate system. This alignment process is achieved by combining penetration depth accumulation or pile driving displacement integration, transforming the construction response data from discrete records into a continuous depth function expression to ensure the physical consistency and numerical stability of subsequent resistance inversion.

3. The method for pile driving resistance inversion and ground abrupt change identification based on sparse dynamics identification according to claim 1, characterized in that: The pile driving state vector in step S3 consists of current response characteristics, historical response characteristics, and depth hysteresis characteristics. The current response includes at least pile driving force, hammer energy, penetration velocity, penetration acceleration, rebound amount, and pore water pressure. The historical response consists of multiple adjacent depth windows to describe the resistance evolution path. The depth hysteresis response is used to characterize the accumulation of pile side friction and the hysteresis effect of pore pressure dissipation. Thus, the pile driving state vector can simultaneously express the instantaneous construction state, the influence of historical loading, and the long-term action mechanism of the strata.

4. The method for pile driving resistance inversion and ground abrupt change identification based on sparse dynamics identification according to claim 1, characterized in that: The candidate dynamic function library in step S4 is a set of constraint functions for pile-soil interaction mechanisms, including construction input function terms, unit penetration energy consumption function terms, penetration velocity decay function terms, side friction accumulation function terms, end resistance mutation function terms, pore pressure change function terms, and quiescent resistance increase function terms. Each function term corresponds to the construction input effect, pile side friction growth mechanism, pile end resistance mutation mechanism, and pore pressure influence mechanism, respectively. It is used to replace the traditional unconstrained feature space, so that the dynamic identification results have clear engineering physical meaning.

5. The method for pile driving resistance inversion and ground abrupt change identification based on sparse dynamics identification according to claim 4, characterized in that: The side friction accumulation function is constructed by weighting and accumulating the pile-soil contact response within the historical depth range, and is used to describe the memory effect of the gradual increase of pile side friction with the depth of soil penetration; the end resistance mutation function is constructed based on the depth gradient change and the prior information of the strata, and is used to characterize the characteristics of the sudden increase in resistance when the pile end enters the hard interlayer or bearing layer; the pore pressure variation function is used to describe the generation, accumulation and dissipation process of pore water pressure in saturated soil, and is used to correct the inversion bias of pile end resistance.

6. The method for pile driving resistance inversion and ground abrupt change identification based on sparse dynamics identification according to claim 1, characterized in that: In step S5, a sparse regression method is used to identify the dynamics of each depth segment. The sparse regression method includes L1 regularized regression, sequential threshold least squares or elastic network method. By introducing a weight matrix, low confidence data is suppressed, so that the influence of abnormal data or missing data on the model coefficients is reduced, thereby improving the stability and robustness of the sparse equation under complex construction conditions.

7. The method for pile driving resistance inversion and ground abrupt change identification based on sparse dynamics identification according to claim 1, characterized in that: The depth segmentation in step S5 adopts a fixed depth window, an adaptive resistance variation window, or a method based on geological stratification boundaries. When there is a sudden change in penetration resistance or unit penetration energy consumption, the window length is automatically reduced to enhance the ability to distinguish areas of abrupt changes in strata and improve the identification accuracy in the boundary areas of soft soil, hard interlayers, and bearing layers.

8. The method for pile driving resistance inversion and ground abrupt change identification based on sparse dynamics identification according to claim 1, characterized in that: In step S6, the equivalent pile driving resistance is obtained by weighted reconstruction of the selected dominant function term and its corresponding coefficient in the sparse expression equation. Based on the physical properties of the function term, it is divided into side friction contribution, end resistance contribution and abnormal resistance contribution. The abnormal resistance contribution includes at least pore pressure anomaly, rebound anomaly and pause recovery effect, thereby realizing the interpretable decomposition of the source of pile driving resistance.

9. The method for pile driving resistance inversion and ground abrupt change identification based on sparse dynamics identification according to claim 1, characterized in that: The formation abrupt change identification in step S7 is based on the joint determination of the degree of change of the set of dominant function terms of the sparse equation between adjacent depth segments, the magnitude of the change of the sparse coefficient, the change of the resistance gradient, and the change of the residual. In this way, the change of the dynamic mechanism in different depth segments is quantified by the structural difference index, avoiding the reliance on a single threshold of penetration or pile driving force, thereby improving the identification stability under complex formation conditions.

10. The method for pile driving resistance inversion and formation abrupt change identification based on sparse dynamics identification according to claim 1, characterized in that: The construction parameter correction strategy output in step S8 includes final pressure correction, final penetration correction, pile driving speed control, hammer energy adjustment, hammer stop standard correction, and re-pressure and re-driving construction strategy optimization. Based on the identified stratum type, it outputs differentiated construction control suggestions for hard interlayers, soft interlayers, bearing layer entry, and abnormal resistance zones, thereby achieving closed-loop optimization control and dynamic adjustment of the pile driving construction process.

Images