Engineering quality supervision method and system with risk early warning function

Abstract

This invention discloses a method and system for engineering quality supervision with risk warning function. The method includes data acquisition, designing an engineering loss function, constructing an engineering quality supervision model, and engineering quality supervision. This invention belongs to the field of quality supervision, specifically referring to a method and system for engineering quality supervision with risk warning function. This solution defines an engineering loss function to clearly identify invalid detection samples, suppress equipment errors, normal detection samples, and suppress label errors; introduces equipment error suppression strength to reduce the sensitivity of detection equipment reading fluctuations; eliminates interference through an invalid detection sample threshold; introduces a time-history coherence penalty term into the objective function to constrain prediction differences between adjacent construction stages; and decomposes the engineering loss function, designs heuristic adjustment factors based on detection sample deviations, corrects residuals in different scenarios, avoids iterative bias, and improves adaptability in complex detection environments; thereby improving the effectiveness of engineering quality supervision.

Description

Technical Field

[0001] This invention relates to the field of quality supervision technology, specifically to an engineering quality supervision method and system with risk early warning function. Background Technology

[0002] Engineering quality supervision methods involve collecting testing data throughout the construction process and employing technical means or management measures to control the quality status of key stages. However, general engineering quality supervision methods suffer from several drawbacks: they fail to distinguish between different types of deviations in engineering testing, are susceptible to irrelevant interference, and lack specific suppression mechanisms for engineering testing scenarios, leading to poor reliability in engineering quality supervision. Furthermore, these methods often fail to consider the temporal characteristics of engineering construction, are easily affected by accidental errors from extreme values ​​of single test samples, and are prone to deviation accumulation during iterative processes, resulting in poor overall engineering quality supervision effectiveness. Summary of the Invention

[0003] To address the above issues and overcome the shortcomings of existing technologies, this invention provides an engineering quality supervision method and system with risk warning functionality. Addressing the problems of general engineering quality supervision methods failing to distinguish between different types of deviations in engineering testing, being susceptible to irrelevant interference, and lacking specific suppression mechanisms for engineering testing scenarios, thus leading to poor reliability in engineering quality supervision, this solution defines an engineering loss function to clearly identify corresponding invalid test samples, suppress equipment errors, normal test samples, and suppress label errors; introduces equipment error suppression strength to reduce the sensitivity of testing equipment reading fluctuations, including rebound hammers; eliminates interference through invalid test sample thresholds; and sets a tolerance for label misjudgment. A threshold is used to control the upper limit of loss for human-incorrectly labeled test samples, avoiding the influence of noise and thus improving the reliability of engineering quality supervision. Addressing the issues of general engineering quality supervision methods failing to consider the temporal characteristics of construction, being susceptible to random errors from extreme values ​​of single test samples, and experiencing accumulated biases during iteration, leading to poor engineering quality supervision results, this solution introduces a temporal continuity penalty term into the objective function to constrain prediction differences between adjacent construction stages. Furthermore, it decomposes the engineering loss function, designs heuristic adjustment factors based on test sample bias, corrects residuals in different scenarios, avoids iteration bias, and improves adaptability in complex testing environments, thereby enhancing the effectiveness of engineering quality supervision.

[0004] The technical solution adopted by this invention is as follows: The engineering quality supervision method with risk early warning function provided by this invention includes the following steps:

[0005] Step S1: Data Acquisition;

[0006] Step S2: Design the engineering loss function;

[0007] Step S3: Construct an engineering quality supervision model;

[0008] Step S4: Project quality supervision.

[0009] Furthermore, in step S1, the data acquisition involves obtaining historical engineering inspection data, including structural data, process data, and material data; standardizing the acquired data; and labeling the quality level to obtain an engineering inspection dataset.

[0010] Furthermore, in step S2, the engineering loss function is defined as the prediction bias as the reverse index of the model's prediction credibility for the detected samples, and the engineering loss function is defined by four segments.

[0011] Furthermore, in step S3, the construction of the engineering quality supervision model is based on GBDT, and an engineering inspection dataset and engineering loss function are used to establish the engineering quality supervision model;

[0012] Specifically, the following steps are included:

[0013] Step S31: Model optimization objective design; GBDT generates M decision trees through iteration. In the m-th iteration, the new tree minimizes the engineering loss function and regularization term, and introduces a time history coherence penalty term to constrain the prediction differences of the same monitoring object in adjacent construction stages.

[0014] Step S32: Engineering quality level expansion; adopting a one-to-many strategy, a total of K GBDT sub-models are trained, and the final quality level is determined by the level corresponding to the maximum predicted value;

[0015] Step S33: Solve the optimization algorithm; specifically including:

[0016] Step S331: Decompose the engineering loss function; decompose the engineering loss function into an outer curvature function and an inner curvature function;

[0017] Step S332: Residual Correction; In the m-th iteration, the residuals are corrected using the heuristic adjustment factor of the inner curvature function on the current predicted value;

[0018] Step S333: Accelerate the solution; solve the weight of the leaf nodes of each tree using the dual coordinate descent algorithm, while pruning the weight range.

[0019] Furthermore, in step S4, the engineering quality supervision involves inputting the pre-processed new test samples into the engineering quality supervision model, and issuing supervision and early warning based on the quality level output by the model.

[0020] The engineering quality supervision system with risk early warning function provided by the present invention includes a data acquisition module, an engineering loss function design module, an engineering quality supervision model construction module, and an engineering quality supervision module;

[0021] The data acquisition module obtains historical engineering inspection data and constructs an engineering inspection dataset.

[0022] The engineering loss function design module introduces equipment error suppression strength and invalid detection sample threshold; and defines the engineering loss function.

[0023] The engineering quality supervision model construction module is based on GBDT, combines engineering inspection datasets and engineering loss functions, and constructs an engineering quality supervision model by designing an optimization objective that includes a time-history continuity penalty term.

[0024] The engineering quality supervision module monitors engineering quality based on real-time engineering inspection data using an engineering quality supervision model.

[0025] The beneficial effects achieved by the present invention using the above solution are as follows:

[0026] (1) In view of the problem that general engineering quality supervision methods do not distinguish the types of deviations in engineering testing, are affected by irrelevant interference, and lack specific suppression mechanisms for engineering testing scenarios, thus leading to poor reliability of engineering quality supervision, this solution defines an engineering loss function to clarify the corresponding invalid test samples, suppress equipment errors, normal test samples, and suppress label errors; introduces equipment error suppression strength to reduce the sensitivity of test equipment reading fluctuations, including rebound hammers; eliminates interference through invalid test sample thresholds; sets a label misjudgment tolerance threshold to control the upper limit of loss of test samples mislabeled by humans and avoid noise impact; thereby improving the reliability of engineering quality supervision.

[0027] (2) In view of the problems that general engineering quality supervision methods do not take into account the time characteristics of engineering construction, are easily affected by the random error of extreme values ​​of single test samples, and are prone to the accumulation of deviations in the iteration process, thus leading to poor engineering quality supervision effect, this scheme introduces a time continuity penalty term into the objective function to constrain the prediction difference between adjacent construction stages; and splits the engineering loss function, designs a heuristic adjustment factor based on the deviation of test samples, corrects the residuals in different scenarios, avoids iteration deviation, and improves adaptability in complex test environments; thereby improving the engineering quality supervision effect. Attached Figure Description

[0028] Figure 1 A flowchart illustrating the engineering quality supervision method with risk warning function provided by the present invention;

[0029] Figure 2 This is a schematic diagram of an engineering quality supervision system with risk warning function provided by the present invention.

[0030] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof. Detailed Implementation

[0031] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0032] In the description of this invention, it should be understood that the terms "upper", "lower", "front", "rear", "left", "right", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the system or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0033] Example 1, see Figure 1 The present invention provides an engineering quality supervision method with risk early warning function, which includes the following steps:

[0034] Step S1: Data Acquisition; Obtain historical engineering inspection data and construct an engineering inspection dataset;

[0035] Step S2: Design the engineering loss function; introduce equipment error suppression strength and invalid detection sample threshold; define the engineering loss function;

[0036] Step S3: Construct an engineering quality supervision model; Based on GBDT, and combining the engineering inspection dataset and engineering loss function, construct an engineering quality supervision model by designing an optimization objective that includes a time-history coherence penalty term.

[0037] Step S4: Engineering quality supervision; Engineering quality supervision is carried out based on real-time engineering inspection data using the engineering quality supervision model.

[0038] Example 2, see Figure 1 This embodiment is based on the above embodiment. In step S1, data acquisition involves obtaining historical engineering inspection data, including structural data, process data, and material data. The structural data includes the compressive strength of concrete cubes, the yield strength of steel bars, and the verticality of components. The process data includes the thickness of the steel bar protective layer, the welding length, and the gap of the formwork splice. The material data includes the weather resistance grade of waterproof materials, the mud content of sand and gravel, and the dosage of admixtures. The collected data is standardized and labeled with quality grades, including qualified, minor defects, serious defects, and rework. The resulting engineering inspection dataset is then obtained.

[0039] Example 3, see Figure 1 This embodiment is based on the above embodiment. In step S2, the engineering loss function is designed by defining the prediction bias as a reverse index of the model's prediction confidence for the detected samples. The reduction of dual interference is achieved through a four-segment segmentation definition, expressed as follows: The four segments correspond to invalid test samples, suppression of equipment errors, normal test samples, and suppression of label errors, respectively; among them, It is the strength of equipment error suppression. ,when At that time, the sensitivity to fluctuations in the rebound hammer readings decreases; It is the tolerance threshold for label misjudgment. ,when At the same time, the upper limit of the loss for human-labeled detection samples is controlled within 0.2 to avoid the model being biased; This is the threshold for invalid detection samples. ,when When the prediction bias is less than -0.3, the test samples are considered invalid; e is the prediction bias. , It represents the model's predicted quality level for the i-th detected sample.

[0040] By performing the above operations, this solution addresses the problems of poor reliability in engineering quality supervision caused by the general engineering quality supervision methods' failure to distinguish the types of deviations in engineering testing, susceptibility to irrelevant interference, and lack of specific suppression mechanisms for engineering testing scenarios. Specifically, it defines an engineering loss function to clearly identify invalid test samples, suppress equipment errors, normal test samples, and suppress label errors; introduces equipment error suppression strength to reduce the sensitivity of testing equipment reading fluctuations, including rebound hammers; eliminates interference through invalid test sample thresholds; and sets a label misjudgment tolerance threshold to control the upper limit of loss from human-caused mislabeled test samples and avoid noise impact; thereby improving the reliability of engineering quality supervision.

[0041] Example 4, see Figure 1 This embodiment is based on the above embodiment. In step S3, the engineering quality supervision model is built on GBDT, based on the engineering inspection dataset and the engineering loss function.

[0042] Specifically, it includes:

[0043] Step S31: Model optimization objective design; GBDT generates M decision trees through iteration. In the m-th iteration, the new tree minimizes the engineering loss function and regularization term to avoid overfitting to random errors in the engineering (extreme detection values ​​of a single detection sample). A temporal continuity penalty term is introduced to constrain the prediction differences of the same monitoring object in adjacent construction stages, utilizing temporal patterns to reduce the impact of random errors, expressed as: ;in, is the parameter set for generating the decision tree in the m-th iteration; l is the total number of detection samples, is the size of the engineering detection dataset used for training, and i is the index of the detection sample; It is the ensemble prediction result (quality grade prediction value) of the first m-1 decision trees. and These are the i-th and (i-1)-th detection samples, respectively; and These are the prediction results of the (m-1)th and mth trees for the detected samples, respectively; It is the depth of the tree; This is the depth penalty coefficient, with a value ranging from 0.01 to 1.0; It is the time-history continuity penalty coefficient, with a value ranging from 0.1 to 1.0;

[0044] Step S32: Engineering quality level expansion; adopting a one-to-many strategy, a total of K GBDT sub-models are trained, and the ensemble prediction of the q-th sub-model is... Represented as: The final quality grade is determined by the grade corresponding to the highest predicted value, expressed as: Where M is the total number of decision trees contained in each sub-model, and t is the decision tree index; is the t-th decision tree of the q-th sub-model; x is the input detection sample data; These are the parameters of the decision tree; This is the final quality rating determination;

[0045] Step S33: Solve the optimization algorithm; the segmentation of engineering loss will cause deviations in the residual calculation of GBDT, so an optimization algorithm is used to solve the problem; specifically including:

[0046] Step S331: Decomposition of Engineering Loss; Decompose the engineering loss function into an outer curvature function and an inner curvature function, as follows: ; ; ;in, and These are the outer curvature function and the inner curvature function obtained by splitting; The three segments, which offset the non-outer curved portion, respectively correspond to the excess loss due to label misjudgment, the lack of offset within the normal loss range, and the excess loss due to equipment error.

[0047] Step S332: Residual Correction; In the m-th iteration, the inner curvature function is approximated by the heuristic adjustment factor of the inner curvature function in the current predicted value to correct the residuals and avoid iteration bias, expressed as: ; ; ; The three segments correspond to the following samples: misjudged label detection samples (residual correction to 1), normal detection samples (residual calculated normally), and equipment error detection samples (residual correction to -1). It is a heuristic adjustment factor that is dynamically adjusted based on the bias of the test samples; It is the corrected residual of the i-th detection sample in the m-th iteration; It is the gradient of the external curvature function; It is the prediction bias of the i-th detected sample in the (m-1)-th iteration;

[0048] Step S333: Accelerate the solution; assign weights to the leaf nodes of each tree. The solution is obtained through a dual coordinate descent algorithm, while simultaneously cropping the weight range to [0, W]. max ], represented as: ;in, It is the optimal weight of the j-th leaf node; It is a weighted clipping operation; It is the set of all samples contained in the j-th leaf node;

[0049] The model parameters are optimized using a Bayesian optimization algorithm to ensure that the performance of the completed engineering quality supervision model meets the standards.

[0050] By performing the above operations, this solution addresses the problems of general engineering quality supervision methods, such as failing to consider the temporal characteristics of engineering construction, being susceptible to random errors from extreme values ​​of single test samples, and the tendency for deviations to accumulate during the iterative process, leading to poor engineering quality supervision results. This solution introduces a temporal continuity penalty term into the objective function to constrain the prediction differences between adjacent construction stages; it also decomposes the engineering loss function, designs heuristic adjustment factors based on test sample deviations, corrects residuals in different scenarios, avoids iterative deviations, and improves adaptability in complex testing environments; thereby improving the effectiveness of engineering quality supervision.

[0051] Example 5, see Figure 1 This embodiment is based on the above embodiment. In step S4, the engineering quality supervision is to input the new test sample into the engineering quality supervision model after preprocessing, and to conduct supervision and early warning based on the quality level output by the model; if the output quality level is serious defect or rework, an early warning is issued to the management personnel.

[0052] Example 6, see Figure 2 Based on the above embodiments, the engineering quality supervision system with risk warning function provided by the present invention includes a data acquisition module, an engineering loss function design module, an engineering quality supervision model construction module, and an engineering quality supervision module.

[0053] The data acquisition module obtains historical engineering inspection data and constructs an engineering inspection dataset.

[0054] The engineering loss function design module introduces equipment error suppression strength and invalid detection sample threshold; and defines the engineering loss function.

[0055] The engineering quality supervision model construction module is based on GBDT, combines engineering inspection datasets and engineering loss functions, and constructs an engineering quality supervision model by designing an optimization objective that includes a time-history continuity penalty term.

[0056] The engineering quality supervision module monitors engineering quality based on real-time engineering inspection data using an engineering quality supervision model.

[0057] It should be noted that, in this document, the terms “comprising,” “including,” or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0058] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention.

[0059] The present invention and its embodiments have been described above. This description is not restrictive, and the accompanying drawings are only one embodiment of the present invention; the actual structure is not limited thereto. In conclusion, if those skilled in the art are inspired by this description and design similar structures and embodiments without departing from the spirit of the invention, such designs should fall within the protection scope of the present invention.

Claims

1. A method for supervising engineering quality with risk early warning function, characterized in that: The method includes the following steps: Step S1: Data Acquisition; Obtain historical engineering inspection data and construct an engineering inspection dataset; Step S2: Design the engineering loss function; define the engineering loss function by introducing the equipment error suppression strength and the invalid detection sample threshold; Step S3: Construct an engineering quality supervision model; Based on GBDT, and combining the engineering inspection dataset and engineering loss function, construct an engineering quality supervision model by designing an optimization objective that includes a time-history coherence penalty term. Step S4: Engineering quality supervision; Engineering quality supervision is carried out based on real-time engineering inspection data using the engineering quality supervision model.

2. The engineering quality supervision method with risk early warning function according to claim 1, characterized in that: In step S2, the engineering loss function is defined as the prediction bias as the reverse index of the model's prediction confidence of the test samples, and the engineering loss function is defined by four segments.

3. The engineering quality supervision method with risk early warning function according to claim 2, characterized in that: In step S3, the construction of the engineering quality supervision model is based on GBDT, and the engineering inspection dataset and engineering loss function are used to establish the engineering quality supervision model. Specifically, the following steps are included: Step S31: Model optimization objective design; GBDT generates M decision trees through iteration. In the m-th iteration, the new tree minimizes the engineering loss function and regularization term, and introduces a time history coherence penalty term to constrain the prediction differences of the same monitoring object in adjacent construction stages. Step S32: Engineering quality level expansion; adopting a one-to-many strategy, a total of K GBDT sub-models are trained, and the final quality level is determined by the level corresponding to the maximum predicted value; Step S33: Optimize the algorithm to solve.

4. The engineering quality supervision method with risk early warning function according to claim 3, characterized in that: In step S3, the optimization algorithm solution specifically includes: Step S331: Decompose the engineering loss function; decompose the engineering loss function into an outer curvature function and an inner curvature function; Step S332: Residual Correction; In the m-th iteration, the residuals are corrected using the heuristic adjustment factor of the inner curvature function on the current predicted value; Step S333: Accelerate the solution; solve the weight of the leaf nodes of each tree using the dual coordinate descent algorithm, while pruning the weight range.

5. The engineering quality supervision method with risk early warning function according to claim 4, characterized in that: In step S1, the data acquisition involves obtaining historical engineering inspection data, including structural data, process data, and material data; and standardizing the acquired data. The quality levels are then labeled to obtain the engineering inspection dataset.

6. The engineering quality supervision method with risk early warning function according to claim 5, characterized in that: In step S4, the engineering quality supervision involves inputting the pre-processed new test samples into the engineering quality supervision model, and issuing supervision and early warning based on the quality level output by the model.

7. An engineering quality supervision system with risk early warning function, used to implement the engineering quality supervision method with risk early warning function as described in any one of claims 1-6, characterized in that: It includes a data acquisition module, an engineering loss function design module, an engineering quality supervision model construction module, and an engineering quality supervision module; The data acquisition module obtains historical engineering inspection data and constructs an engineering inspection dataset. The engineering loss function design module introduces equipment error suppression strength and invalid detection sample threshold; and defines the engineering loss function. The engineering quality supervision model construction module is based on GBDT, combines engineering inspection datasets and engineering loss functions, and constructs an engineering quality supervision model by designing an optimization objective that includes a time-history continuity penalty term. The engineering quality supervision module monitors engineering quality based on real-time engineering inspection data using an engineering quality supervision model.

Images