Uncertain model prediction and evaluation method based on construction quality big data

Abstract

The invention discloses an uncertainty model prediction and evaluation method based on construction quality big data, and the method comprises the steps: dividing construction data into construction quality influence factors and construction quality test results, and building a database; and constructing an uncertainty evaluation model between the construction quality influence factors and the construction quality test results under the single effect and the coupling effect, and predicting and evaluating the construction quality test results obtained under different construction quality influence factors except the database based on the model. The method not only can predict the accuracy of the construction quality in a project, but also can expand and verify the quality reliability of different construction parts, and verifies the rationality of detection data.

Description

Technical Field

[0001] The present invention belongs to the field of engineering quality control, and specifically relates to an uncertainty model prediction and evaluation method based on construction quality big data. Background Art

[0002] The quality of construction projects is related to the safety of people's lives and property, the future and inheritance of cities, and the level of development of new urbanization. With the continuous expansion of the scale and complexity of construction projects, the reliability and accuracy of construction quality have become important factors affecting the overall safety of the project. Traditional construction quality inspection methods usually rely on on-site inspections and empirical judgments. However, due to the complexity of the actual construction environment, the randomness of data collection, and the influence of human factors, the uncertainty of construction quality data has become increasingly prominent. This uncertainty not only affects the accuracy of quality inspection results, but may also lead to misjudgments in subsequent construction links and increased engineering risks.

[0003] Existing uncertainty analysis methods for construction quality data are mostly focused on empirical models and statistical methods, but these methods have great limitations when dealing with data under the influence of multiple factors. Specifically, it is difficult for traditional methods to systematically analyze and quantify the comprehensive impact of multiple factors on construction quality, especially when dealing with large-scale data under complex engineering conditions, the efficiency of traditional methods is insufficient. With the diversification of data collection methods and the increase in data volume, how to accurately evaluate the uncertainty of construction quality data under the background of big data has become an important challenge facing the current field of engineering quality control. At the same time, due to the lack of a comprehensive data comparison system, it is difficult to find differences and abnormalities in different projects, locations or time ranges based on existing engineering data, and the valuable information and laws hidden in the data are ignored, which has become another important challenge facing the current field of engineering quality control. In addition, in order to ensure the real-time comparison, hidden danger identification and defect tracing system of construction engineering construction quality, a sound system has not yet been established. The premise of establishing this system is to ensure the reliability and rationality of construction data. How to evaluate the uncertainty of construction quality data based on the needs of real-time comparison, hidden danger identification and defect tracing of construction engineering construction quality is also an urgent problem to be solved in the current field of engineering quality control.

[0004] Therefore, in the prediction of the uncertainty of quality data of a specific project, it is necessary to fully explore the patterns of construction data, establish an initial data source, and build a comprehensive prediction model to improve the accuracy of uncertainty analysis of construction quality data, and expand its application to the verification of quality reliability and rationality of inspection data under the influence of factors such as different construction locations, different construction levels, and different construction times.

[0005] In the current environment where construction quality prediction and uncertainty analysis are increasingly valued, the integration of actual engineering quality data and prediction models has received special attention, and it is urgent to improve the accuracy and reliability of uncertainty prediction of construction quality data. Especially in complex engineering projects, it is particularly important to adapt to specific engineering conditions to provide more scientific and systematic support for engineering quality control. Summary of the invention

[0006] Purpose of the invention: In view of the current situation in the field of construction where valuable information and rules hidden in construction data are ignored and it is difficult to systematically analyze and quantify the comprehensive impact of multiple factors on construction quality, the present invention combines uncertainty models such as polynomial chaos expansion to predict and evaluate construction quality data, thereby providing a basis for real-time comparison of construction quality, hidden danger identification and defect tracing of construction projects.

[0007] The technical solution adopted by the present invention is: an uncertainty model prediction and evaluation method based on construction quality big data, comprising the following steps:

[0008] (1) Establish an original database of factors affecting construction quality, including but not limited to construction location, design strength grade, and curing age, which may affect construction quality;

[0009] (2) Establishing an original database of construction quality test results, including but not limited to failure load, representative value of compressive strength, and other construction quality data obtained through measurement;

[0010] (3) constructing the relationship between the failure load value and the influencing factors, and establishing the first type of model between the construction quality influencing factors and the construction quality test results according to the above steps (1) and (2);

[0011] (4) Determine the reliability of the construction quality data, and predict and judge the influencing factors and test results outside the original database based on the model obtained in step (3);

[0012] (5) Construct the relationship between the representative value of compressive strength and the failure load. According to the above step (4), a second type of model between the construction quality test results is established.

[0013] Furthermore, in step (1), the classification method of the influencing factors is principal component analysis or cluster analysis.

[0014] Furthermore, in step (3), the first type of model adopts the control variable method to establish a single input and output model based on the construction location, design strength grade, and curing age factors, or couples all single factor models and adopts the polynomial chaos method to establish a multi-input, single output model.

[0015] Furthermore, in step (3), no less than 80% of the data in the original database used by the first type of model is used for model training, and no more than 20% of the data is used for model verification.

[0016] Furthermore, in step (4), when judging the reliability of the test results outside the original database, the corresponding influencing factors are brought into the first type of model and the second type of model. If the result calculated by the model is within the error range of the test result, the test result is considered to be reliable; otherwise, the test result is considered to be unreliable.

[0017] Furthermore, the influencing factors and test results in steps (1), (2) and (4) are all derived from the same specific project.

[0018] Beneficial effects of the present invention:

[0019] (1) By introducing principal component analysis, cluster analysis and polynomial chaos methods, the multi-input, single-output comprehensive prediction model can accurately capture the complex relationship between multiple key factors affecting construction quality, thereby significantly improving the reliability of uncertainty analysis of construction quality data;

[0020] (2) The present invention can more systematically predict the performance of construction quality by establishing a relationship model between the failure load and the representative value of compressive strength and coupling multiple single factor models. The coupling and comprehensive analysis of such models not only improves the reliability of the prediction results, but also reduces the errors caused by changes in a single factor;

[0021] (3) The present invention not only provides prediction and analysis of existing construction quality data, but can also be expanded to verify the quality reliability of different construction sites and the rationality of test data.

[0022] (4) The present invention performs uncertainty evaluation on construction quality data, thereby ensuring the rationality and reliability of the construction data and providing a basis for real-time comparison of construction quality, identification of hidden dangers and tracing of defects in building projects. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 It is the relationship between the mean of the maximum allowable standard deviation and the minimum allowable standard deviation and the age. DETAILED DESCRIPTION

[0024] The present invention will be further described below in conjunction with the accompanying drawings and specific embodiments.

[0025] The concrete strength of different locations, grades and ages of a certain project is shown in Table 1. Based on the above data, the quality of the data itself is evaluated and the future concrete strength is predicted.

[0026] Table 1 Concrete strength table of a certain project

[0027]

[0028] Step 1: Establish an original database of factors affecting construction quality. In this test data, the construction location and concrete design strength grade are fixed values, and the curing age is a variable. Therefore, the factor affecting construction quality is the curing age. The ages of the original database are 28, 40, 48, 60, and 61, respectively.

[0029] Step 2: Establish the original database of construction quality test results. In this test data, the original database of test results is the single value (strength) corresponding to the age of the original database.

[0030] Step 3: Establish a model between the maintenance age and single values.

[0031] (1) Data with ages of 28, 40, 48, and 60 are selected as the model training database, and data with age of 61 is selected as the verification database.

[0032] (2) Preliminary determination of the data reliability range. Using the data at age 28, calculate the standard deviation of the failure load data as the initial maximum allowable standard deviation, and the calculated result is 4.7. Then reduce the number that has the greatest impact on the standard deviation in turn, and recalculate the standard deviation, which are 3.2, 1.2, 1.1, and 0, respectively. Take the standard deviation before the standard deviation increases by more than twice except 0 as the minimum allowable standard deviation (if there is no standard deviation, take the first standard deviation), which is 1.2. Continue to calculate the maximum and minimum allowable standard deviations of 40, 48, and 60, as shown in Table 2.

[0033] Table 2 Concrete strength table of a certain project

[0034] Age 28 40 48 60 Minimum allowable standard deviation 1.2 1.7 1.0 0.85 Maximum allowable standard deviation 4.7 6.7 5.4 6.6 Single value mean 61.9 60.2 56.36 66.35

[0035] (3) Further determine the reliability range. Calculate the mean within the allowable range. Considering that the larger the age, the larger the mean, and the minimum maximum allowable error at age 28, correct the results at ages 40 and 48, as shown in the following table.

[0036] Age 28 40 48 60 Minimum allowable standard deviation 1.2 3.3 2.1 0.85 Maximum allowable standard deviation 4.7 6.7 5.4 6.6 Single value mean 61.9 62.3 63.9 66.35

[0037] (4) Using the mean and age, add the (0,0) point to establish a single value prediction model:

[0038] Y=5.69×lnx+42.28 (1)

[0039] Where Y is the predicted single value and x is the age.

[0040] Step 4: Determine the reliability of construction quality data. Figure 1As shown, the age 61 is substituted into formula (1) to calculate the predicted single value of 65.68Mpa. The mean of the maximum allowable standard deviation and the minimum allowable standard deviation is taken as the maximum allowable standard deviation (5.85) and the minimum allowable standard deviation (1.86) of the model. If the standard deviation of the measured value and the predicted value is less than the minimum allowable standard deviation, it is directly considered reliable; if the standard deviation of the measured value and the predicted value is greater than the minimum allowable standard deviation, it is directly considered unreliable; if the standard deviation of the measured value and the predicted value is greater than the minimum allowable standard deviation, it is directly considered unreliable; if it is between the two, it is weighted according to the distance between the mean square deviation and the maximum allowable standard deviation and the minimum allowable standard deviation. Obviously, the standard deviation of the single value 65.6MPa and 66.3MPa and the predicted value is less than the minimum allowable standard deviation, and can be directly judged as qualified. The mean square deviation corresponding to the single value 61.7 is 2.81, which is between the two.

[0041] Step 5: Construct the relationship between the single value of compressive strength and the failure load, establish the second type of model between the construction quality test results, and obtain the failure strength of the component. Since the single values ​​of 65.6MPa and 66.3MPa are directly qualified, and 61.7 is between qualified and unqualified, the reliability weighting of the above three sets of data is performed to obtain the component strength:

[0042] (65.6×1+66.3×1+61.7×(5.85-2.81) / (5.85-1.86)) / (1+1+(5.85-2.81) / (5.85-1.86))=64.78Mpa.

[0043] The above embodiment only expresses one implementation mode of the present invention, and its description is relatively specific and detailed, but it cannot be understood as limiting the scope of the patent of the present invention. It should be pointed out that for ordinary technicians in this field, several modifications and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be based on the attached claims.

Claims

1. An uncertainty model prediction and evaluation method based on construction quality big data, characterized by: The steps include: (1) Establish an original database of factors affecting construction quality, including but not limited to construction location, design strength grade, and curing age, which may affect construction quality; (2) Establishing an original database of construction quality test results, including but not limited to failure load, representative value of compressive strength, and other construction quality data obtained through measurement; (3) constructing the relationship between the failure load value and the influencing factors, and establishing the first type of model between the construction quality influencing factors and the construction quality test results according to the above steps (1) and (2); (4) Determine the reliability of the construction quality data, and predict and judge the influencing factors and test results outside the original database based on the model obtained in step (3); (5) Construct the relationship between the representative value of compressive strength and the failure load. According to the above step (4), a second type of model between the construction quality test results is established.

2. The uncertainty model prediction and evaluation method based on construction quality big data according to claim 1 is characterized by: In the step (1), the classification method of the influencing factors is principal component analysis or cluster analysis.

3. The uncertainty model prediction and evaluation method based on construction quality big data according to claim 2 is characterized by: In the step (3), the first type of model adopts the control variable method to establish a single input and output model based on the construction location, design strength grade, and curing age factors, or couples all single factor models and adopts the polynomial chaos method to establish a multi-input, single output model.

4. The uncertainty model prediction and evaluation method based on construction quality big data according to claim 3 is characterized by: In step (3), no less than 80% of the data in the original database used by the first type of model is used for model training, and no more than 20% of the data is used for model verification.

5. The uncertainty model prediction and evaluation method based on construction quality big data according to claim 4 is characterized by: In the step (4), when the reliability of the test results outside the original database is judged, the corresponding influencing factors are brought into the first type of model and the second type of model. If the result calculated by the model is within the error range of the test result, the test result is considered to be reliable; otherwise, the test result is considered to be unreliable.

6. The uncertainty model prediction and evaluation method based on construction quality big data according to claim 5 is characterized by: The influencing factors and test results in steps (1), (2) and (4) are all derived from the same specific project.

Images