CN-121835444-B - Hammering pile bearing capacity prediction method based on static sounding

CN121835444BCN 121835444 BCN121835444 BCN 121835444BCN-121835444-B

Abstract

The invention discloses a pile hammering bearing capacity prediction method based on static sounding, which comprises the following steps of collecting historical existing pile hammering construction related data and carrying out normalization processing, selecting a gradient lifting decision tree model as a prediction model, inputting the pile hammering construction data into the selected prediction model, training the prediction model in a field grouping cross validation mode to obtain a predicted bearing capacity, enabling the predicted bearing capacity to gradually approach to the actual measured bearing capacity, thus obtaining a trained machine learning model, obtaining actual pile hammering construction field construction data, inputting the actual pile hammering construction field construction data into the trained machine learning model, and outputting the predicted bearing capacity in real time by the trained machine learning model. The method solves the technical problems that mass historical data cannot be utilized in the prior art, accurate and rapid inversion prediction of the pile length of a new project is achieved, and the intellectualization, the precision and the high efficiency of pile foundation design are achieved.

Inventors

LIU XU
CHEN JIAN
Yu Beiyang
LUO HANBIN

Assignees

华中科技大学

Dates

Publication Date: 20260508
Application Date: 20260313

Claims (6)

1. A hammering pile bearing capacity prediction method based on static sounding is characterized by comprising the following steps: (1) Collecting historical related data of the prior hammering pile construction, wherein the related data of the hammering pile construction comprises static detection data of a site, pile parameters, hammering pile construction data and corresponding actual measurement bearing capacity; (2) Selecting a gradient lifting decision tree model as a prediction model, inputting the construction data of the hammering piles into the selected prediction model, training by adopting a site grouping cross verification mode, and performing iterative optimization on the prediction model so as to obtain a trained machine learning model; (3) Static detection data, pile parameters and pile construction data of an actual pile construction site are acquired, normalized and then input into the trained machine learning model, and the trained machine learning model outputs predicted bearing capacity in real time; in the step (2), lightGBM is selected as a prediction model to obtain a trained machine learning model, and the specific steps are as follows: (T1) feature vectors for the hammering pile construction related data Expressed as: , wherein, As the data of the static detection, Is a pile type, Is the diameter of the pile, the diameter of the pile is the diameter of the pile, For the length of the pile, For the energy of hammering E per meter, Pile verticality, i representing the ith sample; (T2) passing the feature vector in the step (T1) Calculating to obtain the predicted bearing capacity of the ith sample The specific calculation formula is as follows: ; Wherein, the Representing feature vectors Is the first of (2) A weak evaluator; Representing a histogram optimization classification regression tree set; representing a weak evaluator total; (T3) the predicted bearing capacity of the ith sample obtained according to step (T2) Construction LightGBM of an integral objective function of a model The following are provided: ; Wherein, the The loss term is used for quantifying the deviation between the predicted bearing capacity and the actually measured bearing capacity; The square loss function is expressed, and the following is satisfied: ; representing regularization terms for constraining complexity of the histogram-based classification regression tree, satisfying: ; Wherein T represents the number of leaf nodes, Output weights representing the j-th leaf node; penalty coefficients for the number of leaf nodes set; 、 all represent the regularization coefficients set; (T4) an overall objective function on the LightGBM model Adopting a greedy iteration strategy, and adding an optimal histogram classification regression tree in the t-th iteration Minimizing the objective function, the predicted bearing capacity of the t-th iteration Expressed as: ; Wherein, the Is t th Predicting bearing capacity of 1 iteration; (T5) predicting the bearing capacity of the T-th iteration Integral objective function with LightGBM model Obtaining the objective function of the t-th iteration An objective function for the t-th iteration After the second-order Taylor expansion and derivation, the integral objective function of the simplified LightGBM model is obtained : ; ; Wherein, the Representing an objective function Is the first derivative of (a); ; Representing an objective function Second derivative of (1), let , Then the optimal leaf node weight : ; Will be Substitution into objective function Obtaining the integral objective function of LightGBM model after t-th iteration optimization , The training machine learning model is completed; 。
2. The method for predicting pile-on-hammer bearing capacity based on static sounding according to claim 1, wherein the static sounding data is specific penetration resistance when a single-bridge probe is adopted, and the static sounding data is cone tip resistance and side wall friction resistance when a double-bridge probe is adopted.
3. The method for predicting the bearing capacity of a hammering pile based on static sounding according to claim 2, wherein when a double-bridge probe is adopted for testing, the static sounding data are two-dimensional data, and a classical unsupervised dimension reduction method is adopted for reducing dimensions of the static sounding data, so that one-dimensional data are obtained.
4. The method for predicting pile bearing capacity of hammering pile based on static sounding according to claim 3, wherein the pile parameters comprise pile type, pile diameter and pile length, and the hammering pile construction data comprises hammering energy and pile perpendicularity.
5. The method for predicting pile bearing capacity based on static sounding according to claim 4, wherein the specific calculation formula for normalization processing of the static sounding data is as follows: ; Wherein, the Every 10cm interval when the static detection depth is aligned with the depth of the pile The data of the static probe is obtained, Representing normalized statics data.
6. The method for predicting pile bearing capacity based on static sounding according to claim 5, wherein the normalization processing of pile construction data comprises converting the pile energy into pile energy consumed by pile penetration per meter, and the calculation formula of the pile energy consumed by pile penetration per meter is as follows: ; Wherein, the In order to obtain the number of hammers per meter, In order to set the hammering efficiency coefficient, For the mass of the hammer body, the weight of the hammer body is equal to that of the hammer body, The acceleration of the gravity is that, For each drop hammer height; the hammering energy consumed for each meter of pipe pile to enter soil is provided.

Description

Hammering pile bearing capacity prediction method based on static sounding Technical Field The invention belongs to the technical field of geotechnical engineering and engineering informatization, and particularly relates to a hammering pile bearing capacity prediction method based on static sounding. Background In the foundation treatment of construction engineering, precast piles (especially hammering or static pressure precast piles) are widely used due to the advantages of stable quality, high construction speed and the like. One of the core tasks of pile foundation design is to determine the pile length required to meet the design load bearing requirements. Because precast piles need to be prefabricated in a factory in advance, each length is fixed. If the pile length is calculated inaccurately, the pile is too short, so that the bearing capacity is insufficient, and hidden danger is brought to engineering safety. The pile is excessively long, so that the waste is caused, increasing engineering costs. At present, the following methods mainly exist for pile foundation design in engineering practice: (1) The in-situ test method is to determine the standard value of the vertical ultimate bearing capacity of a single precast pile of the concrete precast pile according to static sounding data of a single bridge and a double bridge probe, the given calculation formula is required to be judged and calculated in a classified mode according to soil layer types and pile lengths, meanwhile, parameters involved in a company are required to be obtained by taking an average value of relative depths, and some parameters are required to be checked in a table provided by a standard and the parameters are required to be checked in a picture to be obtained. Is more complex and has fewer practical applications. (2) And (3) estimating bearing capacities corresponding to different pile lengths according to stratum distribution and physical indexes (a pile side limit side friction resistance standard value and a limit pile end resistance standard value) of a drilling acquisition site based on an empirical formula in building pile foundation technical Specification (JGJ 94-2008). The method is complex in process, the empirical formula depends on regional experience, the range value is provided by the specification, and the range value is difficult to accurately acquire in engineering projects. (3) In the on-site pile testing method, a pile testing is firstly constructed and a static load test is carried out in the design stage, and the relation between the bearing capacity and the pile length is directly determined. The method is most reliable, but has high cost and long period, and cannot be widely applied in the early design stage. The common disadvantage of the above methods is the inability to efficiently mine and exploit the complex nonlinear relationship of "formation-pile-bearing capacity" contained in historical engineering data. Each completed engineering project, its static penetration test (CPT) data, actual construction pile length and final verified bearing capacity are all precious data samples. How to utilize the massive historical data to realize accurate and rapid inversion prediction of the pile length of a new project is a technical problem to be solved in the field. Disclosure of Invention Aiming at the defects or improvement demands of the prior art, the invention provides a hammering pile bearing capacity prediction method based on static sounding, thereby solving the technical problem that mass historical data cannot be utilized in the prior art, and realizing accurate and rapid inversion prediction of the pile length of a new project. According to the invention, a direct and rapid mapping model between the CPT characteristics of the site and the pile length required for meeting the target bearing capacity can be established by learning historical engineering data, so that the optimal pile length is intelligently recommended under the given design bearing capacity, and the intellectualization, the precision and the high efficiency of the pile foundation design are realized. In order to achieve the above object, according to one aspect of the present invention, there is provided a method for predicting bearing capacity of a pile hammer based on static sounding, comprising the steps of: (1) Collecting historical related data of the prior hammering pile construction, wherein the related data of the hammering pile construction comprises static detection data of a site, pile parameters, hammering pile construction data and corresponding actual measurement bearing capacity; (2) Selecting a gradient lifting decision tree model as a prediction model, inputting the construction data of the hammering piles into the selected prediction model, training the prediction model in a site grouping and cross-validation mode, performing iterative optimization on the prediction model to obtain a predicted bearing capacity,