Residential engineering quality data credibility evaluation method based on hierarchical Bayesian confidence model

Abstract

The invention discloses a residential engineering quality data credibility evaluation method based on a hierarchical Bayesian confidence model. The method comprises the following steps: forming a quality evaluation sample set by writing a process stage identifier and an acquisition equipment identifier into each quality index observation value; a non-parametric layered Bayesian confidence model is constructed in a process stage layer, and stage drift prior constraints are introduced, so that each quality index forms comparable posteriori prediction distribution under multi-peak, skewed and truncated observation conditions; the method comprises the following steps: extracting a credible interval based on posterior inference, generating a credibility score, further calculating stage drift sensitivity according to the credibility score and the credible interval, screening to-be-rechecked quality indexes according to a preset drift threshold, and outputting associated process stage identifiers and acquisition equipment identifiers, thereby providing a traceable pointing basis for quality rechecking.

Description

Technical Field

[0001] This invention relates to the technical field of data processing and reliability assessment for building engineering quality inspection and quality management, and in particular to a method for reliability assessment of residential engineering quality data based on a hierarchical Bayesian confidence model. Background Technology

[0002] With the development of smart construction sites, BIM collaboration, and digitalization of the construction process, residential engineering quality management is gradually shifting from post-construction sampling inspection to a process monitoring + data evaluation model. On-site testing equipment (rebound hammer, protective layer tester, moisture meter, leakage detector, etc.) and construction record systems continuously generate quality indicator observations. Management aims to use statistical inference to provide quantitative conclusions on the reliability of the data and allocate review resources to indicators with higher risks. However, residential engineering quality data exhibits significant multi-source heterogeneity and stage-specific characteristics: the physical state, material age, and environmental conditions of the same quality indicator differ at different process stages, and the observation distribution often shows multi-peak, skewed, or truncated characteristics introduced by equipment range and acceptance thresholds. At the same time, differences in the calibration status, sampling frequency, and on-site operation of different equipment can superimpose measurement biases. Existing methods often rely on fixed distribution assumptions or empirical threshold rules, which are prone to problems such as overly wide / narrow reliability intervals, scoring fluctuations, and unstable review directions under changes in distribution patterns and process stage drift. It is difficult to simultaneously take into account the generalization ability across indicators and the sensitivity to stage drift, thus affecting the accuracy and traceability of quality review.

[0003] CN113408927A discloses a method and system for evaluating the quality of prestressed construction based on big data. This scheme is more inclined towards model matching and sampling verification strategies. On the one hand, it does not provide a transferable statistical inference structure for the heterogeneity of the distribution of various quality indicators in residential engineering, making it difficult to provide a stable and reliable interval under multimodal, skewed, or truncated observation conditions. On the other hand, it does not include the process stage as a hierarchical variable in the inference link, and lacks a quantitative sensitivity calculation and threshold triggering verification and positioning mechanism for the distribution drift of adjacent process stages. Therefore, in scenarios with strong stage fluctuations, the verification direction and resource allocation are still prone to relying on experience adjustments.

[0004] CN111859301A discloses a data reliability evaluation method based on an improved Apriori algorithm and Bayesian network inference. This method focuses more on correlation and network inference. Its reliability output mainly relies on the prior network structure and conditional probability table construction, which makes it difficult to directly cover the hierarchical distribution differences of residential engineering quality indicators at different process stages. At the same time, it does not provide a path to extract the credibility interval based on the posterior prediction distribution and combine the coverage probability and interval width to form a score. It also lacks modeling of the truncated observations caused by the range limitation in conjunction with the credibility interval, making it difficult to further link the reliability conclusions to the traceable output structure of the indicator to be verified - process stage - acquisition equipment.

[0005] In summary, existing engineering quality assessment and data reliability evaluation technologies generally suffer from the following problems: insufficient adaptation to the heterogeneous distribution of multiple indicators in residential engineering, insufficient quantification of distribution drift at different process stages, and insufficient correlation between verification targets and data acquisition equipment. To address these issues, this invention proposes a method for assessing the credibility of residential engineering quality data based on a hierarchical Bayesian confidence model. This method involves creating a quality assessment sample set by writing process stage identifiers and data acquisition equipment identifiers into each quality indicator observation value; constructing a nonparametric hierarchical Bayesian confidence model at the process stage level and introducing prior constraints on stage drift, enabling each quality indicator to form a comparable posterior predicted distribution under multimodal, skewed, and truncated observation conditions; extracting confidence intervals based on posterior inference and generating confidence scores; further calculating stage drift sensitivity using confidence scores and confidence intervals; filtering quality indicators to be verified according to a preset drift threshold; and outputting associated process stage identifiers and data acquisition equipment identifiers, thereby providing traceable directional evidence for quality verification. Summary of the Invention

[0006] The purpose of this section is to outline some aspects of the embodiments of the present invention and to briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section, as well as in the abstract and title of the present application, to avoid obscuring the purpose of this section, the abstract and title of the invention. Such simplifications or omissions shall not be used to limit the scope of the present invention.

[0007] In view of the aforementioned existing problems, the present invention is proposed.

[0008] To solve the above-mentioned technical problems, the present invention provides the following technical solution: As a preferred embodiment of the residential engineering quality data credibility assessment method based on hierarchical Bayesian confidence model described in this invention, the method involves: collecting observation values ​​of residential engineering quality indicators, and writing process stage identifier and collection equipment identifier for each observation value to obtain a quality assessment sample set; The quality assessment sample set is stratified according to the process stage identifier, and a nonparametric hierarchical Bayesian confidence model is established for each quality indicator with stage drift prior constraints set. The quality assessment sample set is input into the nonparametric hierarchical Bayesian confidence model for posterior inference to obtain the confidence interval and confidence score corresponding to each quality indicator. Based on the credibility score and the credibility interval, the stage drift sensitivity is calculated. Quality indicators whose stage drift sensitivity is greater than the preset drift threshold are identified as quality indicators to be reviewed, and an evaluation result containing the quality indicators to be reviewed, their process stage identifiers, and the data acquisition device identifiers is output.

[0009] The beneficial effects of this invention are as follows: By collecting observed values ​​of residential engineering quality indicators and writing them into process stage identifiers and acquisition equipment identifiers, a traceable quality assessment sample set is formed. This makes the same indicator comparable under different construction stages and different equipment sources, facilitating the identification of data deviation sources and reducing misjudgments caused by data inconsistency. Furthermore, by stratifying the quality assessment sample set according to process stage identifiers and establishing a nonparametric hierarchical Bayesian confidence model with stage drift prior constraints, stable posteriors of quality indicators can be obtained without manual selection of distribution under multimodal, skewed, and truncated observation conditions, improving sensitivity to stage distribution drift and cross-indicator generalization ability. Confidence intervals and confidence scores are obtained through posterior inference, so that the assessment results simultaneously include uncertainty boundaries and confidence level quantification indicators, improving the interpretability of quality judgments. Furthermore, stage drift sensitivity is calculated based on confidence intervals and confidence scores, and quality indicators to be reviewed are screened. The associated process stage identifiers and acquisition equipment identifiers are output, allowing review resources to focus on high-drift, high-risk indicators, improving review efficiency and reducing the risk of missed detections. Attached Figure Description

[0010] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein: Figure 1 This is a flowchart illustrating the method for assessing the credibility of residential engineering quality data based on a hierarchical Bayesian confidence model, as shown in this invention. Detailed Implementation

[0011] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0012] Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without inventive effort should fall within the scope of protection of this invention.

[0013] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0014] According to an embodiment of the present invention, in combination Figure 1The flowchart shown illustrates a method for assessing the credibility of residential engineering quality data based on a hierarchical Bayesian confidence model, which specifically includes the following steps: S1. Collect the observed values ​​of quality indicators for residential engineering projects, and write the process stage identifier and data collection equipment identifier for each observed value to obtain a quality assessment sample set. Note that the following points should be noted in this step: S1.1 The quality data acquisition terminal acquires the quality index observation values ​​of the target residential project within a preset sampling period, and stores the quality index observation values ​​in association with the acquisition timestamp to obtain the original observation record set.

[0015] In a preferred embodiment, the quality data acquisition terminal is deployed at the quality inspection points of the target residential project. The target residential project is uniquely identified by the project code + building code + unit code, such as NJ-QA-2025-018#-B3-2U. The preset sampling period is set as a graded period based on the frequency of formation of quality indicators: 30 minutes for concrete rebound value, wall flatness, and steel reinforcement protective layer thickness; 4 hours for water tightness test leakage and waterproof layer thickness; and 24 hours for masonry mortar strength. At the beginning of each sampling period, the quality data acquisition terminal generates a sampling period number and writes the sampling period number into the same record line as the newly added original observation record within that sampling period.

[0016] It should be noted that in this embodiment, the acquisition of quality index observation values ​​adopts a dual-channel recording method of direct equipment sampling + manual verification and entry: For indicators that can be directly output by the measuring equipment (such as concrete rebound value, steel reinforcement protective layer thickness, and floor slab thickness), the quality data acquisition terminal reads the measurement result frame of the acquisition equipment through a wired interface or a short-range wireless interface, and parses the measurement result frame into numerical observation values; for indicators that need to be manually interpreted and entered (such as door and window opening size deviation and exterior wall finish hollow rate), the inspection personnel enter the observation values ​​in the indicator entry interface of the quality data acquisition terminal, and simultaneously enter the corresponding sampling location code and inspection batch code. The sampling location code and inspection batch code are stored in the same table as the quality index observation values ​​as additional fields of the original observation record set.

[0017] Furthermore, the acquisition timestamps are generated by the unified clock of the quality data acquisition terminal and synchronized with the on-site time server before the start of each day's work; examples of acquisition timestamps are "2025-12-05 09:30:00" and "2025-12-05 10:00:00". The acquisition timestamps are written to the original observation record set with second-level precision.

[0018] Preferably, through the above recording method, each record in the original observation record set has a unified time and a verifiable source chain; compared with the method of recording only a single measurement value in paper records, the above setting method enables the subsequent hierarchical modeling stage to organize samples under the constraints of the same sampling cycle number, the same inspection batch code, and the same sampling location code, thereby reducing the interval expansion and scoring offset caused by cross-batch and cross-location mixing.

[0019] S1.2 Read the equipment registration form of the quality data acquisition terminal, extract the acquisition equipment identifier and equipment calibration status identifier corresponding to the original observation record set, and write the acquisition equipment identifier and equipment calibration status identifier into the original observation record set to obtain the equipment-annotated observation record set.

[0020] In a preferred embodiment, the quality data acquisition terminal locally stores a device registration form, which includes at least the following fields: unique device number identifier field, device type identifier field, metrological calibration batch identifier field, measurement range identifier field, sampling frequency identifier field, most recent calibration date field, calibration validity period in days field, and device status field.

[0021] As an example, equipment type identifiers include rebound hammer, protective layer measuring instrument, ultrasonic thickness gauge and laser verticality meter; range identifiers are RNG-01: 0~100, RNG-02: 0~50, RNG-03: 0~300.

[0022] It should be noted that the extraction and writing in this embodiment are performed using a unique device identifier association matching method: the unique device identifier reported by the acquisition device is written into the original observation record set when it is generated; the quality data acquisition terminal uses the unique device identifier as the association key to retrieve the corresponding row in the device registration table, and reads the device type identifier, metrological calibration batch identifier, range interval identifier, and sampling frequency identifier as the acquisition device identifier and writes them into the corresponding fields of the original observation record set; the device calibration status identifier is calculated from the most recent calibration date field and the calibration validity period in days field: when the acquisition timestamp is earlier than the most recent calibration date + calibration validity period in days and the remaining validity period is not less than 30 days, the device calibration status identifier is assigned a value of valid; when the remaining validity period is less than 30 days but greater than 0 days, it is assigned a value of near expiration; when the remaining validity period is less than or equal to 0 days, it is assigned a value of expired; when the most recent calibration date is missing in the device registration table, it is assigned a value of not registered; subsequently, the device calibration status identifier is written into the original observation record set to obtain the device-labeled observation record set.

[0023] Preferably, by writing the range interval identifier and equipment calibration status identifier of the acquisition device into the sample record level, the subsequent posterior inference stage can provide consistent input for the range cutoff boundary and observation weight coefficient. Compared with the method of only saving calibration information in the equipment management ledger without binding it to the observation record, the above setting method enables the confidence interval calculation to distinguish the observation deviation of the same indicator under different equipment states, thereby reducing the impact of abnormal diffusion caused by calibration-expired equipment on the score.

[0024] S1.3. Based on the construction procedure plan, map the collection timestamps in the equipment annotation observation record set to process stage identifiers, and write the process stage identifiers into the equipment annotation observation record set to obtain the quality assessment sample set.

[0025] In a preferred embodiment, the construction procedure schedule consists of a process stage identifier field, a planned start time field, a planned end time field, a building code field, an inspection batch code field, and a responsible work team field; for example, the planned start time for the main structure construction stage is 2025-11-20 08:00:00, and the planned end time is 2025-12-20 18:00:00; the planned start time for the plastering construction stage is 2025-12-21 08:00:00, and the planned end time is 2026-01-10 18:00:00.

[0026] It should be noted that the mapping method in this embodiment is executed in the order of first inspection batch and then time window: For each record in the equipment annotation observation record set, firstly, its inspection batch code field is read, and the row set with the same inspection batch code is retrieved in the construction procedure plan table; when the retrieved row set corresponds to only one process stage identifier, the process stage identifier is written into the record; when the retrieved row set corresponds to multiple process stage identifiers, the collection timestamp of the record is read, and the inclusion determination is performed within the time window formed by the planned start time and planned end time of each candidate process stage, and the process stage identifier that is included is written; when multiple time windows are included, the preset process order in the construction procedure plan table is used as the priority; when none of the time windows are included, the record is marked as stage pending, and supplemented and corrected according to the responsible team field and construction log in the subsequent data review stage; after the mapping is completed, the process stage identifier is written into the equipment annotation observation record set to obtain the quality assessment sample set.

[0027] Preferably, this embodiment adopts a mapping method of inspection batch code constraint + time window inclusion judgment, so that the process stage identification does not depend on a single time threshold division, and can be compatible with process overlap and partial rework. Compared with the method of directly dividing the stages by date interval, the above mapping reduces the mixing of the same inspection batch across stages, thereby providing a clearer hierarchical input for the stage layer sample subset of S2.

[0028] As an example, the identification of the data acquisition device includes the device type identifier, the device unique number identifier, the metrological calibration batch identifier, the measurement range identifier, the sampling frequency identifier, and the device calibration status identifier.

[0029] As an example, the process stage markers include the foundation construction stage marker, the main structure construction stage marker, the masonry construction stage marker, the plastering construction stage marker, the waterproofing construction stage marker, the electromechanical installation stage marker, the decoration and finishing stage marker, and the final acceptance stage marker.

[0030] S2. Stratify the quality assessment sample set according to the process stage identifier, and establish a nonparametric hierarchical Bayesian confidence model for each quality indicator, setting stage drift prior constraints. Note that the following points should be noted in this step: S2.1 Read the process stage identifier from the quality assessment sample set, generate a stage-level sample subset according to the process stage identifier, and generate a stage-level index table corresponding to the stage-level sample subset.

[0031] Specifically, the quality assessment sample set is deduplicated according to the process stage identifier to obtain a process stage identifier set. This process stage identifier set is then sorted according to a preset process order to obtain a process stage sequence. Based on the process stage sequence, the quality assessment sample set is grouped and extracted to obtain a stage-level sample subset that corresponds one-to-one with each process stage identifier. A sample sequence number is written for each sample record. A stage-level index table is generated based on the sample sequence number. The stage-level index table includes a process stage identifier field, a sample sequence number field, and a data collection timestamp field, with the process stage identifier field having the same grouping key as the stage-level sample subset. The stage boundary time of adjacent process stage identifiers is calculated based on the data collection timestamp field in the stage-level index table, and the stage boundary time is written into the stage-level index table to obtain a stage-level index table with stage boundaries.

[0032] In a preferred embodiment, the preset process sequence is consistent with the quality acceptance organization sequence of residential engineering projects. For example, the sequence includes the foundation construction stage identifier, the main structure construction stage identifier, the masonry construction stage identifier, the plastering construction stage identifier, the waterproofing construction stage identifier, the electromechanical installation stage identifier, the decoration and finishing stage identifier, and the final acceptance stage identifier. The quality data acquisition terminal reads the process stage identifier field from the quality assessment sample set, removes duplicates from the process stage identifier field to obtain the process stage identifier set, and sorts the process stage identifier set according to the preset process sequence to obtain the process stage sequence.

[0033] It should be noted that the grouping extraction in this embodiment is performed using the process stage identifier field as the grouping key: records in the quality assessment sample set whose process stage identifier field is equal to the same stage identifier are extracted into a stage-level sample subset, and sorted in ascending order by the collection timestamp field in this stage-level sample subset; subsequently, a sample sequence number field is written for each record in this stage-level sample subset, and the sample sequence number increments from 1 according to the sorted sequence; a stage-level index table is generated based on the sample sequence number field, and the stage-level index table includes at least the process stage identifier field, the sample sequence number field, and the collection timestamp field, and the process stage identifier field is consistent with the grouping key of the stage-level sample subset.

[0034] It should be noted that the calculation of the stage boundary time in this embodiment is performed according to the rule of priority of the plan table and statistical supplementation: when there are planned end time and planned start time of two adjacent process stages in the construction process plan table, the two times are written into the stage layer index table as the stage boundary time of the two stages respectively; when the construction process plan table is missing any of the times, the midpoint between the last collection timestamp of the previous stage and the first collection timestamp of the next stage is written into the stage boundary time of the adjacent stage.

[0035] For example, if the last collection timestamp of the main structure construction stage identifier is 2025-12-20 17:40:00, and the first collection timestamp of the plastering construction stage identifier is 2025-12-21 08:20:00, then the stage boundary time is written as 2025-12-21 01:00:00.

[0036] Preferably, by explicitly writing the stage boundary moments into the stage layer index table, the stage sample range can be reproduced according to the same boundary when the index stage observation set is extracted in S3. Compared with the method of stratifying only by stage name without fixing the boundary moments, the above setting reduces sample drift caused by changes in stage division due to changes in personnel understanding.

[0037] S2.2 For each quality index in the stage-level sample subset, set the base distribution candidate set and concentration parameters of the nonparametric observation distribution family; wherein, the base distribution candidate set includes the base distribution type identifier field and the corresponding base distribution parameter prior field, and write the base distribution type identifier field and the base distribution parameter prior field into the prior parameter group to obtain the nonparametric prior of the quality index.

[0038] As an example, for the concrete compressive strength index, the basis distribution type identifier field is normal, log-normal, gamma, or Student's type; for the external wall cladding hollow rate index, the basis distribution type identifier field is beta, log-normal, or Student's type; the basis distribution parameter prior field stores the initial value range and update constraints of the parameter according to the basis distribution type; for example, for a normal basis distribution, the basis distribution parameter prior field stores the central value range of 20 to 60 and the discrete value range of 1 to 25; for a beta basis distribution, the basis distribution parameter prior field stores the shape parameter value range of 0.5 to 20.

[0039] Furthermore, in this embodiment, the concentration parameter is set as a control quantity based on the preference of non-parametric component quantity, such as setting the concrete compressive strength index to 20, the wall surface flatness index to 10, and the external wall finish hollow rate index to 15; then, the base distribution type identifier field and the base distribution parameter prior field are written into the prior parameter group, and the concentration parameter is written into the concentration field of the same prior parameter group to obtain the non-parametric prior of the quality index.

[0040] It should be noted that by setting an enumerable candidate set of base distributions and reproducible concentration parameter values, this step allows for the selection of observed distribution patterns within the candidate set during the subsequent posterior inference stage, avoiding strong assumptions about a single distribution pattern. Compared to the fixed modeling method using a single normal distribution, the above settings provide more stable interval outputs for heavy-tailed, skewed, and proportional indicators.

[0041] S2.3. Set a shared base distribution for the nonparametric priors corresponding to adjacent process stage identifiers, and set offset constraint terms on the base distribution parameters of the shared base distribution; wherein, the offset constraint terms include the difference terms of the corresponding base distribution parameters under adjacent process stage identifiers and the stage drift scale parameters, to obtain a set of shared prior parameters with offset constraints under adjacent process stage identifiers.

[0042] In a preferred embodiment, the same base distribution type is used to identify the field for the same quality index in two adjacent stages, and the prior fields of the base distribution parameters in the two stages share the same set of parameter value interval boundaries; then, an offset constraint term is set on the shared base distribution parameters, which consists of the parameter difference term of the adjacent stages and the stage drift scale parameter.

[0043] Furthermore, for the observed values ​​of this quality index in the sample subsets of the stage layer of two adjacent stages, the stage central statistic (such as the median) is calculated respectively. The difference between the two stage central statistics is used as the initial value of the difference term, and this initial value is written into the difference field of the offset constraint term. The generation of the stage drift scale parameter is determined by the method of historical stable period statistics + specification limit verification: read the difference sequence of the central statistics of the same quality index in adjacent stages in historical projects of similar projects, and take the upper quartile of its absolute value as the initial value of the scale. Then, compare the initial value of the scale with 1 / 3 of the allowable deviation in the corresponding quality acceptance specification, and take the smaller of the two as the stage drift scale parameter.

[0044] For example, if the allowable deviation of the wall verticality index is 4mm, then the dimensional parameter is taken as 1.3mm; if the interquartile range of the concrete compressive strength index between adjacent stages is 3.2MPa, and the standard check value is taken as 4.0MPa, then the dimensional parameter is taken as 3.2MPa.

[0045] By introducing a shared base distribution and a shift constraint term, the distribution pattern of similar indicators is maintained between adjacent stages, while allowing stage differences to occur within a controlled scale. Compared with the method of setting priors completely independently for each stage, the above constraints reduce the large fluctuations in interval output when the sample size is small, thus making the drift sensitivity calculation more focused on abnormal stage changes.

[0046] S2.4. Based on the stage layer index table, combine the nonparametric priors with the stage drift scale parameters to obtain the nonparametric hierarchical Bayesian confidence model corresponding to each quality index.

[0047] As an example, the quality indicators should include at least the concrete compressive strength, concrete rebound value, steel reinforcement protective layer thickness, floor slab thickness, masonry mortar strength, wall verticality, wall flatness, door and window opening size deviation, waterproof layer thickness, water tightness test leakage rate, exterior wall finish hollow rate, and floor moisture content.

[0048] In a preferred embodiment, a three-layer structure of observation layer—stage layer—shared and drift constraint layer is established for each quality index, and the drift scale parameters of adjacent stages are written into the offset constraint field of the shared structure of adjacent stages, so that the sample information within the stage and the constraint information of adjacent stages can be used simultaneously during posterior inference.

[0049] For example, the mathematical expression of a nonparametric hierarchical Bayesian confidence model is as follows: Observation layer: Phase layer: Shared and drift constraint layers: in, For the first The first process stage identifier The quality index observation values ​​corresponding to each sample record. For process stage indexing, This serves as the index for sample sequence numbers within a given phase. To compare with the observed values The corresponding observed distribution parameter vector, The probability density function of the observed distribution selected by the base distribution type identifier field. For in the interval The truncated distribution after truncating the observed distribution; For the first The first process stage identifier The lower cutoff boundary parameters corresponding to each sample record For the first The first process stage identifier The upper truncation boundary parameters corresponding to each sample record; For the first Stage-level random measure under each process stage identifier For Dirichlet process distribution, For concentration parameters, For the first Prior distribution of the basis distribution parameters under each process stage identifier This is a shared basis distribution prior distribution shared between adjacent process stages. For the prior of the shared basis distribution Apply offset to the parameter position The subsequent phase priors The offset operator; For the first Each process stage identifier corresponds to a stage offset parameter. It follows a normal distribution. For the stage drift scale parameters, This is the variance quantization form of the stage drift scale parameter.

[0050] S3. Input the quality assessment sample set into a nonparametric hierarchical Bayesian confidence model for posterior inference to obtain the confidence interval and confidence score corresponding to each quality indicator. Note that the following should be noted in this step: S3.1. Based on the stage-level index table, extract the observation values ​​of each quality indicator corresponding to each process stage identifier from the quality assessment sample set to form the indicator stage observation set.

[0051] In a preferred embodiment, the quality data acquisition terminal reads the stage-level index table and uses the process stage identifier field + sample sequence number field in the stage-level index table as the extraction positioning key. Specifically, for each process stage identifier, the corresponding sample records are traversed according to the sample sequence number field. The record row with the same process stage identifier and the same sample sequence number is retrieved in the quality assessment sample set, and the observation value field of each quality index is read from the record row to form the indicator stage observation set. The record structure of the indicator stage observation set includes at least the process stage identifier field, sample sequence number field, acquisition timestamp field, quality index identifier field, quality index observation value field, range interval identifier field, and equipment calibration status identifier field.

[0052] For example, 120 records of concrete rebound value were extracted from the main structure construction stage, and 260 records of wall flatness were extracted from the plastering construction stage.

[0053] S3.2 Read the range interval identifier and equipment calibration status identifier of each record in the observation set of the index stage, convert the range interval identifier into the truncation boundary parameter, and map the equipment calibration status identifier into the observation weight coefficient.

[0054] In a preferred embodiment, the range interval identifier is a fixed field in the equipment registration table, such as RNG-01: 0~100, RNG-02: 0~50, RNG-03: 0~300; further, the process of converting the range interval identifier into the cutoff boundary parameter is as follows: read the lower and upper bounds of the range corresponding to the range interval identifier, write the lower bound to the lower cutoff boundary field, and write the upper bound to the upper cutoff boundary field; when certain indicators have a physical lower bound allowed by the specification (such as leakage not less than 0), the larger of the physical lower bound and the lower bound of the range is taken as the lower cutoff boundary; for example, if the range interval identifier of the leakage index in the water tightness test is RNG-02: 0~50, then the lower cutoff boundary is 0 and the upper cutoff boundary is 50.

[0055] It should be noted that in this embodiment, the mapping of the equipment calibration status identifier to the observation weight coefficient is performed according to the discrete mapping table: when the equipment calibration status identifier is valid, the observation weight coefficient is 1; when it is close to expiration, it is 0.85; when it has expired, it is 0.60; when it is not registered, it is 0.70. Subsequently, the truncation boundary parameter and the observation weight coefficient are written into the corresponding field of the observation set of the index stage as input to S3.3.

[0056] S3.3 Input the observation set of the index stage, the truncation boundary parameters and the observation weight coefficients into the nonparametric hierarchical Bayesian confidence model, perform posterior iterative update, and obtain the posterior parameter set.

[0057] In a preferred embodiment, after inputting the indicator stage observation set, truncation boundary parameters, and observation weight coefficients into the nonparametric hierarchical Bayesian confidence model, the posterior iterative update is performed according to the following process: The observation distribution and parameter values ​​for each stage are initialized using the prior parameter set in S2.2, and the offset fields of adjacent stages are initialized using the offset constraint term in S2.3; secondly, the sample subset of each stage layer is traversed in record row order, and the component assignment probability of each observation record within the stage is updated based on its observation weight coefficient, thereby updating the stage layer accordingly. The component proportion of random measure; then, after completing one round of traversal, the parameter values ​​of the prior of the shared basis distribution are updated once, and the offset field of the adjacent stage is updated once simultaneously to make it consistent with the stage drift scale parameter; repeat the above cycle of traversal update, shared update, and offset update until the change in the confidence interval width is lower than the preset convergence threshold in several consecutive rounds; for example, the number of rounds is 800, the first 200 rounds are used as stabilization rounds and no intermediate results are output, and the last 600 rounds record the posterior parameter samples to form the posterior parameter set.

[0058] In a preferred embodiment, the preset convergence threshold is given in the form of a relative change threshold of the confidence interval width plus the number of consecutive satisfactions, and is used to determine whether the posterior iterative update has entered a stable state; for example, for the ... For a certain quality index identified by a process stage, the confidence interval obtained in the t-th iteration is denoted as . Its interval width is denoted as When the relative change in interval width for K consecutive iterations is less than the preset convergence threshold. At that time, it is determined that the posterior iterative update has converged.

[0059] For example, threshold The threshold is set to 0.5%, and the number of consecutive satisfying K is set to 30. The determination method is as follows: starting from the 201st iteration, after each iteration, the relative difference between the current interval width and the previous interval width is calculated and written into the convergence monitoring sequence. When the most recent 30 relative differences in the convergence monitoring sequence are all less than 0.5%, and the difference between the maximum and minimum confidence scores of the most recent 30 iterations is not greater than 0.01, the iteration is stopped and the posterior parameter set is output.

[0060] As an example, for the wall flatness index under the plastering construction stage marker, the interval width was 1.82mm in the 450th iteration and 1.81mm in the 451st iteration. The relative change is... The threshold was not met; during the 520th to 549th iterations, the interval width slowly fluctuated from 1.76mm to 1.75mm, with the relative change in each round being less than 0.5%, and the credibility score fluctuated between 0.83 and 0.84 during the same period, with a score difference of 0.01, thus meeting the stopping condition.

[0061] It should be noted that the reason for setting the above threshold in this embodiment is that the sample size and noise level of the residential engineering quality indicators vary greatly at different stages. Using a single round of change is easily affected by occasional anomalies and may cause accidental stopping. By introducing a combined threshold of relative change and number of consecutive satisfactions, the convergence determination can be made independent of the dimension and the premature termination caused by short-term fluctuations can be reduced, thereby making the output confidence interval and confidence score more stable and verifiable.

[0062] S3.4 Calculate the posterior prediction distribution of each quality indicator under each process stage identifier based on the posterior parameter set, and extract the upper and lower quantiles corresponding to the pre-set confidence level from the posterior prediction distribution as the confidence interval of the quality indicator.

[0063] In a preferred embodiment, the process of calculating the posterior prediction distribution based on the posterior parameter set is as follows: for each process stage, read its corresponding posterior parameter sample sequence, generate a prediction distribution curve for each parameter sample, and take the average of all prediction distribution curves to obtain the posterior prediction distribution of the stage; for example, the confidence level is preset to 0.90; the corresponding upper and lower quantiles are 0.05 and 0.95.

[0064] For example, the mathematical formula for calculating the confidence interval is: in, For the first The confidence interval under each process stage identifier For process stage indexing, For pre-set reliability; For the first Quantile function of the posterior prediction distribution under each process stage identifier; The quantile probability level, For the infimum operator, Let R be the set of real numbers and represent the values ​​of the variable on the posterior predictive distribution. For the first The cumulative distribution function of the posterior prediction distribution under each process stage identifier.

[0065] S3.5 Calculate the observation coverage probability based on the posterior predicted distribution, and calculate the uncertainty index based on the confidence interval width. Combine the observation coverage probability and the uncertainty index according to the preset scoring weight parameters to form a confidence score.

[0066] The credibility score is formed by combining the following steps: Based on the posterior predicted distribution and the credibility interval, the interval inclusion determination is performed on the observed values ​​of each quality index in the observation set of the indicator stage; the number of samples falling into the credibility interval is counted and divided by the total number of samples to obtain the observation coverage probability; the interval width is calculated for the credibility interval, and the interval width is normalized according to the range span corresponding to the range interval identifier to obtain the uncertainty index; the observation coverage probability (e.g., coverage weight is 0.7) and uncertainty (e.g., uncertainty weight is 0.3) index are weighted according to the preset scoring weight parameters to obtain the credibility score, wherein the credibility score increases with the increase of the observation coverage probability and decreases with the increase of the uncertainty index.

[0067] As an example, the mathematical formulas for calculating the observation coverage probability, uncertainty index, and confidence score are as follows: in, For the first Observational coverage probability under each process stage identifier For the first The number of observation records for this quality indicator under each process stage is identified. This is the index for the observation record sequence number within the period, with values ​​ranging from 1 to... ; This is an indicator function; the value is 1 if the condition inside the parentheses is true, and 0 otherwise. For the first The first process stage identifier The quality index observation values ​​corresponding to each observation record For the first Under each process stage identifier, the confidence level is... The confidence interval; For the first Uncertainty index under each process stage identifier This is the interval width operator, and its value is the upper bound of the interval minus the lower bound of the interval. For the first The range span under each process stage identifier For the first The upper limit of the confidence interval under each process stage identifier. For the first The lower limit of the confidence interval under each process stage identifier. This is the upper limit of the range obtained by resolving the range interval identifier. This is the lower bound of the measurement range obtained by resolving the range interval identifier. For the first Credibility ratings under each process stage identifier The scoring weight parameter for the observation coverage probability, The scoring weight parameter for the uncertainty index.

[0068] For example, for the wall flatness index marked in the plastering construction stage, the number of observation records is 260, and the number of records falling into the confidence interval is 221, so the coverage probability is 0.85; the lower limit of the confidence interval is 0.8mm, the upper limit is 2.6mm, and the range is 0 to 10mm, so the uncertainty index is 0.18; after substituting the weights, the confidence score is 0.84.

[0069] S4. Calculate the stage drift sensitivity based on the confidence score and confidence interval. Quality indicators with a stage drift sensitivity greater than a preset drift threshold are identified as quality indicators to be reviewed. Output the evaluation results, including the quality indicators to be reviewed, their process stage identifiers, and the data acquisition device identifiers. Note that the following should be noted in this step: S4.1. Based on the process stage sequence, extract the lower limit of the confidence interval, the upper limit of the confidence interval, and the confidence score corresponding to each process stage identifier for each quality indicator to obtain the set of stage evaluation items.

[0070] In a preferred embodiment, the set of stage evaluation items includes a quality indicator identifier field, a process stage identifier field, a confidence interval lower limit field, a confidence interval upper limit field, and a confidence score field.

[0071] As an example, the lower limit of the confidence interval for the concrete compressive strength index under the main structure construction stage is 32.5 MPa, the upper limit of the confidence interval is 39.8 MPa, and the confidence score is 0.81; under the completion and acceptance stage, the lower limit of the confidence interval is 30.1 MPa, the upper limit of the confidence interval is 36.0 MPa, and the confidence score is 0.73.

[0072] S4.2 For the two stage evaluation items corresponding to adjacent process stage identifiers, calculate the center difference value of the confidence interval and the overlap rate of the confidence interval. The center difference value of the confidence interval is the absolute value of the difference between the center values ​​of the two confidence intervals, and the overlap rate of the confidence interval is the ratio of the intersection length of the two confidence intervals to the larger of the width of the two confidence intervals.

[0073] S4.3. Divide the center difference of the confidence interval by the mean of the widths of the two confidence intervals to obtain the interval drift ratio, and subtract the overlap rate of the confidence interval from the result to obtain the interval separation.

[0074] S4.4 Calculate the confidence score difference value corresponding to the adjacent process stage identifiers, and sum the interval drift ratio, interval separation degree and confidence score difference value according to the preset sensitivity weight parameters to obtain the drift sensitivity of adjacent stages.

[0075] In a preferred embodiment, the confidence score difference value corresponding to the adjacent process stage identifier is calculated as the absolute value of the difference between the scores of the two adjacent stages; the preset sensitivity weight parameters are, for example, a drift ratio weight of 0.5, a separation weight of 0.3, and a score difference weight of 0.2.

[0076] As an example, the mathematical formula for calculating the drift sensitivity between adjacent stages is as follows: in, Identifying adjacent process stages and The sensitivity to adjacent stage drift, For process stage indexing, The next one in the process stage sequence The next stage index; The sensitivity weighting parameter for the interval drift ratio. The sensitivity weight parameter for interval separation. The sensitivity weighting parameter for the credibility score difference is... This represents the interval drift ratio. For interval separation, For the first Credibility rating under each process stage identifier; For the first Credibility ratings under each process stage identifier.

[0077] S4.5. Take the maximum value of the drift sensitivity of each adjacent stage in the process stage sequence for the same quality index to obtain the stage drift sensitivity.

[0078] Furthermore, the stage drift sensitivity corresponding to each quality indicator is read, and a review judgment identifier field is generated for each quality indicator; when the stage drift sensitivity is greater than the preset drift threshold (e.g., 0.65), the review judgment identifier field is assigned the first value, and when the stage drift sensitivity is less than or equal to the preset drift threshold, the review judgment identifier field is assigned the second value, thus obtaining a set of review judgment results. Select quality indicators whose review judgment identifier field is the first value from the review judgment result set as the quality indicators to be reviewed, and record the target process stage identifier pair corresponding to the stage drift sensitivity of the quality indicator to be reviewed when it reaches the maximum value. Based on the stage-level index table, the sample records of the quality indicators to be reviewed are retrieved from the corresponding stage-level sample subset for the target process stage identifier, the data acquisition device identifier is extracted and the frequency of occurrence is counted to obtain the data acquisition device ranking table. Write the quality indicators to be reviewed, the target process stage identifiers, and the identifier of the acquisition device that appears most frequently in the acquisition device ranking table into the evaluation results and output them.

[0079] It should be noted that the method for setting the preset drift threshold in this embodiment is as follows: read the stage drift sensitivity distribution of each quality indicator in historical projects of similar projects, take the upper quartile as the initial value, and adjust the initial value according to the project risk level; for example, lower the threshold of high-risk indicators (waterproof layer thickness index, water tightness test leakage index) by 0.05, and raise the threshold of low-risk indicators (wall flatness index) by 0.05.

[0080] For example, in this embodiment, the first value and the second value are 1 and 0 respectively. The review determination identifier field is 1 to indicate that it enters the set to be reviewed, and 0 to indicate that it does not enter the set to be reviewed.

[0081] Furthermore, the method for generating the target process stage identifier pair in this embodiment is as follows: locate the adjacent stage pair corresponding to the maximum value of all adjacent stage drift sensitivity records of the quality index, and write the adjacent stage pair as the target process stage identifier pair into the record to be reviewed; if the maximum adjacent stage drift sensitivity of a certain quality index appears in the plastering construction stage identifier - waterproofing construction stage identifier, then the target process stage identifier pair is recorded as an ordered pair of the two stage identifiers.

[0082] It should also be noted that the generation of the data acquisition device ranking table in this embodiment is based on the number of times the device appears in the target stage sample records. Specifically: according to the stage-level index table, locate the two stage-level sample subsets corresponding to the target process stage identifier pair, retrieve the sample records of the quality indicator to be reviewed in the two subsets, extract the data acquisition device identifier field, group and count according to the unique device number identifier, form a ranking table of the unique device number identifier field - the number of occurrence field - the device calibration status identifier field, and arrange them in descending order of the number of occurrences; when the number of occurrences is the same, the device with the device calibration status identifier indicating that it has expired is ranked last; then, the quality indicator to be reviewed, the target process stage identifier pair, and the data acquisition device identifier ranked first in the ranking table are written into the evaluation result and output.

[0083] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for assessing the credibility of residential engineering quality data based on a hierarchical Bayesian confidence model, characterized in that, include: Collect the quality index observation values ​​of residential engineering projects, and write the process stage identifier and the data collection equipment identifier for each observation value to obtain a quality assessment sample set; The quality assessment sample set is stratified according to the process stage identifier, and a nonparametric hierarchical Bayesian confidence model is established for each quality indicator with stage drift prior constraints set. The quality assessment sample set is input into the nonparametric hierarchical Bayesian confidence model for posterior inference to obtain the confidence interval and confidence score corresponding to each quality indicator. Based on the credibility score and the credibility interval, the stage drift sensitivity is calculated. Quality indicators whose stage drift sensitivity is greater than the preset drift threshold are identified as quality indicators to be reviewed, and an evaluation result containing the quality indicators to be reviewed, their process stage identifiers, and the data acquisition device identifiers is output.

2. The method for assessing the credibility of residential engineering quality data based on a hierarchical Bayesian confidence model according to claim 1, characterized in that, The quality assessment sample set is obtained by: The quality data acquisition terminal acquires the observed values ​​of the quality indicators of the target residential project within a preset sampling period, and stores the observed values ​​of the quality indicators in association with the acquisition timestamp to obtain the original set of observation records. Read the device registration table of the quality data acquisition terminal, extract the acquisition device identifier and device calibration status identifier corresponding to the original observation record set, and write the acquisition device identifier and device calibration status identifier into the original observation record set to obtain the device-annotated observation record set; Based on the construction procedure plan, the collection timestamps in the equipment annotation observation record set are mapped to process stage identifiers, and the process stage identifiers are written into the equipment annotation observation record set to obtain the quality assessment sample set.

3. The method for assessing the credibility of residential engineering quality data based on a hierarchical Bayesian confidence model according to claim 1 or 2, characterized in that, The data acquisition device identifier includes the device type identifier, the device unique number identifier, the metrological calibration batch identifier, the measurement range identifier, the sampling frequency identifier, and the device calibration status identifier. The process stage identifiers include the foundation construction stage identifier, the main structure construction stage identifier, the masonry construction stage identifier, the plastering construction stage identifier, the waterproofing construction stage identifier, the electromechanical installation stage identifier, the decoration and finishing stage identifier, and the completion and acceptance stage identifier.

4. The method for assessing the credibility of residential engineering quality data based on a hierarchical Bayesian confidence model according to claim 3, characterized in that, The establishment of a nonparametric hierarchical Bayesian confidence model for each quality indicator and the setting of stage drift prior constraints include: Read the process stage identifier from the quality assessment sample set, generate a stage-level sample subset according to the process stage identifier, and generate a stage-level index table corresponding to the stage-level sample subset; For each quality index in the sample subset of the stage layer, a candidate set of base distributions and a concentration parameter of the nonparametric observation distribution family are set; wherein, the candidate set of base distributions includes a base distribution type identifier field and a corresponding base distribution parameter prior field, and the base distribution type identifier field and the base distribution parameter prior field are written into the prior parameter group to obtain the nonparametric prior of the quality index; A shared base distribution is set for the nonparametric priors corresponding to adjacent process stage identifiers, and a offset constraint term is set on the base distribution parameters of the shared base distribution; wherein, the offset constraint term includes the difference term of the corresponding base distribution parameters under adjacent process stage identifiers and the stage drift scale parameter, to obtain a set of shared prior parameters with offset constraints under adjacent process stage identifiers. Based on the stage layer index table, the nonparametric priors are combined with the stage drift scale parameters to obtain the nonparametric hierarchical Bayesian confidence model corresponding to each quality index.

5. The method for assessing the credibility of residential engineering quality data based on a hierarchical Bayesian confidence model according to claim 4, characterized in that, Generating the stage-level index table includes: The quality assessment sample set is deduplicated according to the process stage identifier to obtain a process stage identifier set, and the process stage identifier set is sorted according to a preset process order to obtain a process stage sequence. Based on the process stage sequence, the quality assessment sample set is grouped and extracted to obtain a stage-level sample subset that corresponds one-to-one with each process stage identifier, and a sample sequence number is written for each sample record. A stage-level index table is generated based on the sample sequence number. The stage-level index table includes a process stage identifier field, a sample sequence number field, and a collection timestamp field. The process stage identifier field is consistent with the grouping key of the stage-level sample subset. The stage boundary time of the adjacent process stage identifier is calculated according to the acquisition timestamp field of the stage layer index table, and the stage boundary time is written into the stage layer index table to obtain a stage layer index table with stage boundaries.

6. The method for assessing the credibility of residential engineering quality data based on a hierarchical Bayesian confidence model according to claim 4, characterized in that, The quality indicators include at least the following: concrete compressive strength, concrete rebound value, steel reinforcement protective layer thickness, floor slab thickness, masonry mortar strength, wall verticality, wall flatness, door and window opening size deviation, waterproof layer thickness, water tightness test leakage rate, exterior wall finish hollow rate, and floor moisture content.

7. The method for assessing the credibility of residential engineering quality data based on a hierarchical Bayesian confidence model according to claim 5, characterized in that, The process of obtaining the confidence intervals and confidence scores corresponding to each quality indicator includes: Based on the stage-level index table, the observation values ​​of each quality indicator corresponding to each process stage identifier are extracted from the quality assessment sample set to form an indicator stage observation set. Read the range interval identifier and equipment calibration status identifier of each record in the observation set of the index stage, convert the range interval identifier into a cutoff boundary parameter, and map the equipment calibration status identifier into an observation weight coefficient; The observation set of the index stage, the cutoff boundary parameters, and the observation weight coefficients are input into the nonparametric hierarchical Bayesian confidence model, and a posterior iterative update is performed to obtain the posterior parameter set. Based on the set of posterior parameters, the posterior prediction distribution of each quality index under each process stage is calculated, and the upper and lower quantiles corresponding to the preset confidence level are extracted from the posterior prediction distribution as the confidence interval of the quality index. The observation coverage probability is calculated based on the posterior prediction distribution, and the uncertainty index is calculated based on the confidence interval width. The observation coverage probability and the uncertainty index are combined according to preset scoring weight parameters to form the confidence score.

8. The method for assessing the credibility of residential engineering quality data based on a hierarchical Bayesian confidence model according to claim 7, characterized in that, The credibility score is formed by combining the following components: Based on the posterior prediction distribution and the confidence interval, an interval inclusion determination is performed on the observation values ​​of each quality index in the indicator stage observation set. The number of samples falling into the confidence interval is counted and divided by the total number of samples to obtain the observation coverage probability. The width of the confidence interval is calculated, and the width of the interval is normalized according to the range span corresponding to the range interval identifier to obtain the uncertainty index. The observation coverage probability and the uncertainty index are weighted according to preset scoring weight parameters to obtain the credibility score, wherein the credibility score increases with the increase of the observation coverage probability and decreases with the increase of the uncertainty index.

9. The method for assessing the credibility of residential engineering quality data based on a hierarchical Bayesian confidence model according to claim 7, characterized in that, Calculating the stage drift sensitivity includes: Based on the process stage sequence, for each quality indicator, extract the lower limit of the confidence interval, the upper limit of the confidence interval, and the confidence score corresponding to each process stage identifier to obtain a set of stage evaluation items. For two stage evaluation entries corresponding to adjacent process stage identifiers, calculate the center difference value of the confidence interval and the overlap rate of the confidence interval, wherein the center difference value of the confidence interval is the absolute value of the difference between the center values ​​of the two confidence intervals, and the overlap rate of the confidence interval is the ratio of the intersection length of the two confidence intervals to the larger of the width of the two confidence intervals. The interval drift ratio is obtained by dividing the center difference of the confidence interval by the mean of the widths of the two confidence intervals, and the interval separation is obtained by subtracting the overlap rate of the confidence intervals from the center difference. Calculate the confidence score difference value corresponding to the adjacent process stage identifier, and then sum the interval drift ratio, the interval separation degree and the confidence score difference value according to the preset sensitivity weight parameter to obtain the adjacent stage drift sensitivity; The stage drift sensitivity is obtained by taking the maximum value of the drift sensitivity of each adjacent stage in the process stage sequence for the same quality index.

10. The method for assessing the credibility of residential engineering quality data based on a hierarchical Bayesian confidence model according to claim 9, characterized in that, The evaluation results are output, including: Read the stage drift sensitivity corresponding to each quality indicator and generate a review judgment identifier field for each quality indicator; wherein when the stage drift sensitivity is greater than the preset drift threshold, the review judgment identifier field is assigned a first value, and when the stage drift sensitivity is less than or equal to the preset drift threshold, the review judgment identifier field is assigned a second value, thereby obtaining a review judgment result set. Select quality indicators whose review judgment identifier field is the first value from the review judgment result set as quality indicators to be reviewed, and record the target process stage identifier pair corresponding to the stage drift sensitivity of the quality indicator to be reviewed taking the maximum value; According to the stage-level index table, the sample records of the quality indicators to be reviewed are retrieved from the corresponding stage-level sample subset of the target process stage identifier, the acquisition device identifier is extracted and the occurrence frequency is counted to obtain the acquisition device sorting table; The evaluation results are written into the quality indicators to be reviewed, the target process stage identifiers, and the identifier of the acquisition device that appears most frequently in the acquisition device sorting table, and then output.

Images