Dredging construction parameter optimization method based on normality test

Abstract

The invention discloses a dredging construction parameter optimization method based on normality test. The method comprises the following steps of: grouping data, collecting construction process parameters, and forming a data set; data grouping: collecting construction process parameters to form a data set; carrying out self-adaptive normality detection based on a distribution pattern, and directly calculating a confidence interval if a data group accords with quasi-normal distribution; if not, a skewed compensation algorithm is adopted to process the data, and then secondary inspection is carried out; abnormal value detection and correction: dynamically detecting and eliminating abnormal values based on probability density; performing quadratic normality test and data completion, and if the test is passed, completing the data by using a confidence interval mean value; and carrying out weighted probability resampling on the non-normal data groups, and adopting a weighted probability resampling method for the data groups which still do not conform to normal distribution after the abnormal values are removed. According to the method, the complexity of multi-dimensional data in the construction process is comprehensively considered, the optimal process parameter combination is provided for ship construction, and the construction efficiency and the engineering safety are improved.

Description

Technical Field

[0001] The present invention relates to the field of data analysis, and in particular to a dredging construction parameter optimization method based on normality test. Background Art

[0002] In recent years, dredging technology has experienced rapid development as demands for efficiency and safety continue to rise in coastal port construction, waterway dredging, and inland waterway navigation. Modern dredging projects not only demand efficient construction capabilities but also place higher demands on data collection, analysis, and optimization during the construction process. During the actual operation process, dredging vessels generate a large amount of multi-dimensional data, including vessel numbers, dredging pump combinations, dredging depths, and mud concentrations, depending on the operating conditions. The accurate processing and in-depth analysis of this data is directly related to the optimal configuration of construction parameters and the improvement of construction efficiency, and is a critical foundation for precise construction management.

[0003] Traditional data processing methods primarily rely on empirical formulas, outlier detection with fixed thresholds, and single normality analysis methods. These methods can effectively process data under specific conditions, but are often limited by single assumptions about data distribution and struggle to fully address the complex and variable data distributions found in on-site construction. Furthermore, due to the presence of noise, outliers, and missing values ​​in dredging construction data, simple methods such as mean or median substitution can easily lead to information loss, hindering the accurate extraction of the optimal combination of construction process parameters, and thus reducing overall project efficiency and safety.

[0004] The rapid development of big data technology has provided new insights into dredging construction data processing. Using various statistical methods, such as normality tests, outlier detection, and resampling, adaptively cleaning and analyzing data from different construction conditions can fully reveal inherent patterns in the data, improving data quality and prediction accuracy. However, due to the unique characteristics of dredging construction data, existing data analysis methods cannot be directly applied. For example, in terms of normality tests, normal data do not strictly conform to the normal distribution, but rather exhibit a slight skewness from the standard normal distribution, which is considered a quasi-normal distribution in engineering. Directly using the Shapiro-Wilk normality test, due to its heavy reliance on a single statistic, can lead to misclassification of slightly skewed data. Outlier detection algorithms, such as the IQR (Independent Quantitative Regression) algorithm, require a fixed threshold, which is difficult to set due to the flexible and variable nature of dredging construction conditions. Furthermore, dredging construction data exhibits a multimodal distribution, which can lead to high misclassification rates. Traditional bootstrap resampling methods ignore the distributional characteristics of dredging data, resulting in poor accuracy, especially for slightly skewed data. Therefore, although some studies have explored the application of data processing and analysis methods in the construction process, most methods fail to fully integrate with the characteristics of construction data, consider the dynamic correlation between multiple process parameters, lack adaptive processing strategies for different data groupings, and cannot meet the actual needs of modern dredging projects for high-precision data analysis and optimal parameter configuration. Summary of the Invention

[0005] The present invention aims to provide a method for optimizing dredging construction parameters based on normality testing. This method comprehensively considers the quasi-normal distribution characteristics of dredging construction data and the complexity of multidimensional data in construction processes. Through techniques such as data group segmentation, adaptive normality testing based on distribution morphology, outlier detection based on probability density, and weighted probability resampling, it achieves precise data cleaning and efficient analysis. Its goal is to provide the optimal combination of process parameters for construction vessels, thereby significantly improving construction efficiency and project safety.

[0006] The technical solution to achieve the above purpose is:

[0007] A dredging construction parameter optimization method based on normality test includes:

[0008] Step S1, data grouping, collecting construction process parameters to form a data set;

[0009] Step S2, adaptive normality test based on distribution morphology, if the data set conforms to the quasi-normal distribution, directly calculate the confidence interval; if not, use the skewness compensation algorithm to process the data, and then perform a secondary test;

[0010] Step S3, outlier detection and correction, dynamically detecting and removing outliers based on probability density;

[0011] Step S4: quadratic normality test and data completion. If the test passes, the data is completed with the confidence interval mean.

[0012] Step S5, weighted probability resampling of non-normal data groups, using weighted probability resampling method for data groups that still do not conform to normal distribution after outliers are removed;

[0013] Step S6: extracting the optimal combination of construction process parameters.

[0014] Preferably, in step S1: the construction process parameters include dredging time, left flow velocity, right flow velocity, left drag head depth, right drag head depth, left density, and right density.

[0015] Preferably, the data set refers to a data set formed by calculation and classification with ship number, mud pump combination, dredging depth and mud concentration as dimensions.

[0016] Preferably, the step S2 includes:

[0017] Step S21, calculate the skewness S and kurtosis K of each grouped data:

[0018]

[0019] Among them, x i represents the i-th data point of the data set, is the mean value of the data set, n represents the number of data in the data set; if S>0, it means that the distribution of the data set is right-skewed, that is, the long tail is on the right; if S<0, it means that the distribution is left-skewed, that is, the long tail is on the left; K>0 means that the distribution of the data set is steeper than the normal distribution, and K<0 means that the distribution is flatter;

[0020] Step S22, calculate the distribution morphology index D shape , if D shape If it is less than the pre-set threshold α, it means that the data set satisfies the quasi-normal distribution, and the confidence interval is calculated directly; if it exceeds the threshold, the skewness compensation algorithm is triggered, and the right-skewed data is logarithmically transformed and the left-skewed data is squared.

[0021] Preferably, the step S22 includes:

[0022] S221, calculate the data group distribution morphology index D shape :

[0023] D shape =|S|+|K-3| (3)

[0024] Where S is the skewness calculated in step S21, which is used to measure the asymmetry of the distribution. The skewness of the standard normal distribution is 0. The kurtosis K is used to measure the thickness of the tail of the distribution. The kurtosis of the standard normal distribution is 3. The excess kurtosis calculated here is K-3.

[0025] S222, for the data set that passes the test, the confidence interval is directly calculated with a 95% confidence level:

[0026]

[0027] Among them, (a,b) represents the confidence interval, is the sample mean, Zα / 2 is the critical value of the standard normal distribution, for 95% confidence, Zα / 2 is approximately 1.96, σ is the population standard deviation, n is the size of the data set, is the standard error;

[0028] Step S223: The data set that fails the test is corrected using a skewness compensation algorithm, and then a second test is performed. The skewness compensation algorithm is as follows:

[0029]

[0030] For S>0, that is, right-skewed data, logarithmic transformation is performed; for S<0, that is, left-skewed data, square transformation is performed, ε represents a small disturbance term to avoid taking the logarithm of zero value, and then recalculate D for the transformed data shape If the test is passed, the confidence interval is calculated; if not, step S3 is executed.

[0031] Preferably, step S3 includes:

[0032] Step S31: construct the probability density function of each data group through kernel density estimation, regard each data point as a kernel function center, and superimpose all kernel functions to obtain the overall density distribution. The formula is as follows:

[0033]

[0034] Among them, x i represents the i-th data point of the data set, x is the location of the density to be estimated, n is the number of data in the data set, h is the bandwidth, which is used to control the smoothness of the density curve, and F is the kernel function;

[0035] Step S32, the rule for judging abnormal values ​​is: set the threshold value β through experience, if the density value f(x i ) satisfies f(x i )<β·max(f(x)), it is judged as an abnormal value;

[0036] Step S33: Set the dynamic bandwidth w, which is improved by combining the IQR and the data set standard deviation to obtain the dynamic bandwidth. The formula is as follows:

[0037] IQR=Q3-Q1 (7)

[0038] Among them, Q1 and Q3 represent the first quartile and the third quartile of the data. After sorting the data from small to large, the data at the 25th percentile is Q1, and the data at the 75th percentile is Q3. Data lower than Q1-1.5×IQR or higher than Q3+1.5×IQR are considered abnormal data.

[0039]

[0040] Where σ is the standard deviation of the data set, and the coefficient γ is dynamically adjusted according to the data sparsity. If the data is sparsely distributed, γ is increased to widen the bandwidth; if the data is densely distributed, γ is reduced to narrow the bandwidth.

[0041] Preferably, in step S4, after removing the outliers, the normality test is performed again on the remaining data; if the data passes the normality test, a 95% confidence interval is calculated, and the outliers and missing values ​​previously removed are replaced with the mean of the confidence interval.

[0042] Preferably, step S5 includes:

[0043] Step S51: Calculate the weight of each data point based on the kernel density estimation result; the formula is as follows:

[0044]

[0045] Among them, f(x i ) represents the data point x i The kernel density estimate of represents the sum of the density values ​​of all data points,

[0046] Step S52: group the data according to the weight ω i Perform sampling with replacement to generate B resampled sample sets For each resampling set Both contain n data points;

[0047] Step S53, weighted statistics calculation:

[0048]

[0049] in, represents the weighted mean of the b-th resampled set, ω i is the data point x i The weight of

[0050] Step S54, confidence interval estimation:

[0051] For B weighted means The ranking was performed and the 2.5% and 97.5% quantiles were taken as the 95% confidence intervals.

[0052] The beneficial effects of the present invention are:

[0053] The present invention has created a variety of statistical analysis techniques such as adaptive normality detection based on distribution morphology, outlier detection based on probability density, and weighted probability resampling to adapt to the characteristics of data distribution under different working conditions and realize adaptive cleaning of data. It can not only quickly calculate the confidence interval when the data presents a normal distribution, but also effectively exclude outliers and dynamically complete data when the data is non-normal, significantly improving the quality and reliability of the data. In addition, this method is particularly targeted at construction process data, and it performs group processing based on multiple dimensions such as ship number, mud pump combination, excavation depth and mud concentration, which can fully reveal the dynamic correlation between various process parameters during the construction process. Based on this method, the optimal combination of construction parameters can be recommended for different working conditions, providing a scientific basis for decision-making on construction ships. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 It is a flow chart of the dredging construction parameter optimization method based on normality test of the present invention;

[0055] Figure 2 It is a specific operation flow chart of the present invention;

[0056] Figure 3 It is a data partition result diagram in the present invention;

[0057] Figure 4 It is the original distribution diagram of the left high-pressure flushing pressure in the present invention;

[0058] Figure 5 This is a diagram showing the left high-pressure flushing pressure conversion in the present invention;

[0059] Figure 6 This is an effect diagram of outlier detection in the present invention;

[0060] Figure 7 is the weighted probability resampling graph of the present invention;

[0061] Figure 8 Schematic diagram of the confidence interval calculation using weighted probability resampling in the present invention;

[0062] Figure 9 It is a statistical diagram of the confidence interval of the left high-pressure flushing pressure in the present invention. DETAILED DESCRIPTION

[0063] The present invention will be further described below with reference to the accompanying drawings.

[0064] See also Figure 1 and Figure 2 The present invention discloses a method for optimizing dredging construction parameters based on normality test, and the specific steps are as follows:

[0065] Step S1, Data Grouping: Collect construction process parameters, including dredging time, left and right flow rates, left and right drag head depths, left and right drag head depths, and left and right density. Through calculation and classification, a data set is formed based on vessel number, dredge pump combination, dredging depth, and mud concentration.

[0066] Dredging time: A dredging trip refers to the entire construction cycle of a construction vessel from an empty to a fully loaded state. During this cycle, dredging time is the core indicator, accumulated from the moment the vessel begins dredging until the vessel reaches a fully loaded state, at which point the accumulation of dredging time ceases and marks the end of the dredging trip.

[0067] Left / Right Flow Rate: The left / right flow rate refers to the fluid delivery rate of the corresponding left and right pumps during operation. It generally reveals the working condition of each pump and its contribution to mud delivery. If the left flow rate is 0 and the right flow rate is not 0, the dredge pump combination is judged to be the right pump only; if the left flow rate is not 0 and the right flow rate is 0, the dredge pump combination is judged to be the left pump only; if both the left and right flow rates are not 0, the dredge pump combination is judged to be the left pump only.

[0068] Left / Right Drag Head Depth: Drag head depth represents the vertical distance of the drag head relative to a reference surface (such as the seabed or hull baseline). Since dredging is primarily about extracting mud from the seabed or sediment, the depth of the drag head represents the actual excavation depth. For calculations, the average of the left and right drag head depths can be used as the excavation depth.

[0069] Left / right density: During the construction process, the mud density is measured in real time by the density sensors of the left and right mud conveying systems, which directly reflects the content of solid matter in the mud per unit volume. Therefore, in the calculation, the average value of the left / right density is taken as the mud concentration.

[0070] Following the above steps, we can obtain a dataset of ship number, mud pump combination, dredging depth, and mud concentration. Due to the high similarity of each dataset, subsequent processing is performed on only one set of data.

[0071] For example, in one embodiment, the dredging and filling project of berths CT10-CT13 in Port Klang, Malaysia is used as the background. The construction vessel is Xinhailong, the soil is medium-coarse sand, and the transportation distance is 8.9 to 18.3 nautical miles from the sand source area to the construction area. The waves at the construction site are mainly in the southwest direction, with a maximum wave height of 0.7m, an average of 0.2m, an average flow rate of 0.4 to 0.5m / s, and a maximum flow rate of 0.8m / s to 1.4m / s. Construction data is collected once a minute. This method and experimental content take the data from 00:00:01 on February 13, 2025 to 23:59:59 on February 13, 2025 as an example. The division of each data is as follows Figure 3 shown.

[0072] Step S2, adaptive normality testing based on distribution morphology: Considering that conventional construction data in most cases does not perfectly conform to the standard normal distribution, but rather exhibits a slightly skewed distribution, it is considered normal. Therefore, the present invention proposes the concept of a quasi-normal distribution. The threshold α for the normality test is set using empirical values. If the data set conforms to the quasi-normal distribution, the confidence interval can be directly calculated. If not, the data is processed using a skewness compensation algorithm, followed by a secondary test.

[0073] Step S21, calculate the skewness S and kurtosis K of each grouped data:

[0074]

[0075] Among them, x i represents the i-th data point of the data set, is the mean of the data set, and n is the number of data points in the data set. If S > 0, the distribution is right-skewed, with a long tail on the right; if S < 0, the distribution is left-skewed, with a long tail on the left. K > 0 indicates that the distribution is steeper than a normal distribution, while K < 0 indicates a flatter distribution.

[0076] Step S22, calculate the distribution morphology index D shape , if D shape If the value is less than a pre-set threshold α, it indicates that the data set exhibits a slightly normal distribution, that is, it satisfies a quasi-normal distribution, and the confidence interval can be directly calculated. If it exceeds the threshold, the skewness compensation algorithm is triggered, performing a logarithmic transformation on right-skewed data and a square transformation on left-skewed data. Note that a small perturbation term is required to avoid taking the logarithm of zero values.

[0077] Step S221, calculate the data group distribution morphology index D shape :

[0078] D shape =|S|+|K-3| (3)

[0079] Where S is the skewness calculated in step S21, which is used to measure the asymmetry of the distribution. The skewness of the standard normal distribution is 0. The kurtosis K is used to measure the thickness of the tail of the distribution. The kurtosis of the standard normal distribution is 3. In the present invention, the excess kurtosis is calculated, that is, K-3. In this way, when calculating the distribution morphology index of the data set, the absolute values ​​of the skewness and excess kurtosis are added together to comprehensively quantify the degree of deviation of the data from the normal distribution, that is, D shape If it is close to 0 or less than the threshold, it means that the skewness and kurtosis are close to the ideal value of the standard normal distribution. shape If the value is large, it means that the data set deviates significantly from the normal distribution in terms of symmetry or tail shape.

[0080] Step S222: For the data set that passes the test, the confidence interval is directly calculated with a confidence level of 95%:

[0081]

[0082] Among them, (a,b) represents the confidence interval, is the sample mean, Z α / 2 is the critical value of the standard normal distribution, for 95% confidence level, Z α / 2 is approximately 1.96, σ is the population standard deviation, n is the size of the data set, is the standard error.

[0083] Step S223: The data set that fails the test is corrected using a skewness compensation algorithm, and then a second test is performed. The skewness compensation algorithm is as follows:

[0084]

[0085] For S>0, that is, right-skewed data, logarithmic transformation is performed; for S<0, that is, left-skewed data, square transformation is performed, ε represents a small disturbance term to avoid taking the logarithm of zero value. Then recalculate D for the transformed data shape If the test is passed, the confidence interval is calculated; if not, step S3 is executed.

[0086] The adaptive normality test based on distribution morphology is a detection method designed by the present invention. Traditional normality tests (such as Shapiro-Wilk) rely on a single statistic and are not sensitive enough to slight skewness or tail anomalies, which can easily lead to misjudgment. The improved dynamic morphology index quantifies the degree of distribution deviation by combining the absolute values ​​of skewness and excess kurtosis, and combines it with adaptive threshold judgment to effectively reduce the misjudgment rate.

[0087] Adaptive normality detection based on distribution morphology: Different normality test algorithms are selected according to the amount of data in each group. In this invention, the adaptive normality detection method based on distribution morphology is used to perform normality test on the grouped data. Taking the left high-pressure flushing pressure as an example, in the original data, the distribution of the left high-pressure flushing pressure is as follows: Figure 4 As shown, it does not conform to the normal distribution.

[0088] After data transformation, normality test was performed again, such as Figure 5 , which is the left high-pressure flushing pressure transformation diagram.

[0089] Step S3, outlier detection and correction:

[0090] The present invention proposes to dynamically detect outliers based on probability density, construct the probability density function of the data group through kernel density estimation, and identify data points in low-density areas as outliers.

[0091] Step S31: construct the probability density function of each data group through kernel density estimation, regard each data point as a kernel function center, and superimpose all kernel functions to obtain the overall density distribution. The formula is as follows:

[0092]

[0093] Among them, x i represents the i-th data point in the data set, x is the location of the density to be estimated, n is the number of data in the data set, h is the bandwidth, which is used to control the smoothness of the density curve, and F is the kernel function.

[0094] Step S32: The rule for determining abnormal values ​​is to set a threshold value β through experience. If the density value f(x i ) satisfies f(x i )<β·max(f(x)), it is judged as an abnormal value.

[0095] Step S33: Set the dynamic bandwidth w. In the traditional method, the bandwidth is a fixed value, which may cause over-smoothing or under-smoothing. Therefore, the present invention improves the dynamic bandwidth by combining the IQR and the standard deviation of the data set. The formula is as follows:

[0096] IQR=Q3-Q1 (7)

[0097] Q1 and Q3 represent the first and third quartiles of the data. After sorting the data from smallest to largest, the data at the 25th percentile is Q1, and the data at the 75th percentile is Q3. Data below Q1 - 1.5 × IQR or above Q3 + 1.5 × IQR are considered outliers.

[0098]

[0099] Where σ is the standard deviation of the data set. The coefficient γ can be dynamically adjusted according to the data sparsity. If the data distribution is sparse (such as multimodal), γ can be increased to widen the bandwidth; if the data distribution is dense (such as unimodal normal), γ can be reduced to narrow the bandwidth.

[0100] The outlier detection method designed by the present invention is based on probability density. Traditional outlier detection (such as IQR / Z-Score) uses a fixed threshold, which is difficult to adapt to the multi-peak or asymmetric distribution characteristics of dredging construction data. However, by dynamically identifying low-density areas through kernel density estimation, it can adapt to a variety of data distributions, effectively improve the outlier detection rate, and reduce the false positive rate.

[0101] When the test results in step S2 do not conform to the normal distribution, it is considered whether the overall data is biased due to the interference of outliers. Therefore, a dynamic detection method based on probability density is used to detect and eliminate outliers. Outlier detection is as follows: Figure 6 As shown, the location of the outliers can be clearly identified and corrected (or directly eliminated).

[0102] Step S4, secondary normality test and data completion: After removing outliers, the remaining data are retested for normality. If the data passes the normality test, a 95% confidence interval is calculated, and the confidence interval mean is used to replace the previously removed outliers and missing values, thereby completing the data and restoring continuity.

[0103] Step S5: Weighted probability resampling of non-normal data sets. For data sets that still do not conform to a normal distribution after outlier removal, a weighted probability resampling method is used. First, based on the kernel density estimation results, the weight of each data point is calculated. Then, sampling with replacement is performed based on the weights to form multiple resampled sets, and finally, confidence intervals are calculated.

[0104] Step S51: Calculate the weight of each data point based on the kernel density estimation result. The significance is to assign higher weights to low-density areas and lower weights to high-density areas. This can effectively reduce the oversampling problem in dense areas. The formula is as follows:

[0105]

[0106] Among them, f(x i ) represents the data point x i The kernel density estimate is used to reflect the local probability density. Represents the sum of the density values ​​of all data points and is used to normalize the weights.

[0107] Step S52: group the data according to the weight ω i Perform sampling with replacement to generate B resampled sample sets For each resampling set Both contain n data points.

[0108] Step S53, weighted statistics calculation:

[0109]

[0110] in, represents the weighted mean of the b-th resampled set, ω i is the data point x i The weight of .

[0111] Step S54, confidence interval estimation:

[0112] For B weighted means The ranking was performed and the 2.5% and 97.5% quantiles were taken as the 95% confidence intervals.

[0113] The weighted probability resampling method, devised by this invention, uses density-based weighting to balance sampling bias between sparse and dense regions, while traditional bootstrap uniform sampling ignores data distribution characteristics. This method effectively reduces errors in confidence interval estimation in skewed data. Through dynamic adaptive design and multi-dimensional optimization, this overall solution significantly improves analytical accuracy and robustness for complex construction data.

[0114] For the data group that still does not conform to the normal distribution after the outliers are removed, the weighted probability resampling method is used. The sampling data is as follows: Figure 7 For the resampled data, the 2.5% and 97.5% quantiles are taken as the 95% confidence interval, as shown in Figure 8 As shown:

[0115] Step S6, extracting the optimal combination of construction process parameters. Based on the processed data in each group, extract the optimal value of each construction process parameter, such as Figure 9 As shown in the figure, the statistics of each group data are recorded, and the optimal value results are fed back to the construction system, providing a decision-making basis for adjusting the ship construction plan, and ultimately achieving higher construction efficiency.

[0116] The above embodiments are only used to illustrate the present invention, rather than to limit the present invention. Those skilled in the art may make various changes or modifications without departing from the spirit and scope of the present invention. Therefore, all equivalent technical solutions should also fall within the scope of the present invention and should be defined by the claims.

Claims

1. A dredging construction parameter optimization method based on normality test, characterized in that: include: Step S1, data grouping, collecting construction process parameters to form a data set; Step S2, adaptive normality test based on distribution morphology, if the data set conforms to the quasi-normal distribution, directly calculate the confidence interval; if not, use the skewness compensation algorithm to process the data, and then perform a secondary test; Step S3, outlier detection and correction, dynamically detecting and removing outliers based on probability density; Step S4: quadratic normality test and data completion. If the test passes, the data is completed with the confidence interval mean. Step S5, weighted probability resampling of non-normal data groups, using weighted probability resampling method for data groups that still do not conform to normal distribution after outliers are removed; Step S6: extracting the optimal combination of construction process parameters.

2. The method for optimizing dredging construction parameters based on normality test according to claim 1, characterized in that: In step S1, the construction process parameters include dredging time, left flow velocity, right flow velocity, left drag head depth, right drag head depth, left density, and right density.

3. The method for optimizing dredging construction parameters based on normality test according to claim 1, characterized in that: The data set refers to a data set formed by calculation and classification with ship number, mud pump combination, dredging depth, and mud concentration as dimensions.

4. The method for optimizing dredging construction parameters based on normality test according to claim 1, characterized in that: The step S2 comprises: Step S21, calculate the skewness S and kurtosis K of each grouped data: Among them, x i represents the i-th data point of the data set, is the mean value of the data set, n represents the number of data in the data set; if S>0, it means that the distribution of the data set is right-skewed, that is, the long tail is on the right; if S<0, it means that the distribution is left-skewed, that is, the long tail is on the left; K>0 means that the distribution of the data set is steeper than the normal distribution, and K<0 means that the distribution is flatter; Step S22, calculate the distribution morphology index D shape , if D shape If it is less than the pre-set threshold α, it means that the data set satisfies the quasi-normal distribution, and the confidence interval is calculated directly; if it exceeds the threshold, the skewness compensation algorithm is triggered, and the right-skewed data is logarithmically transformed and the left-skewed data is squared.

5. The method for optimizing dredging construction parameters based on normality test according to claim 4, characterized in that: The step S22 includes: S221, calculate the data group distribution morphology index D shape : D shape =|S|+|K-3| (3) Where S is the skewness calculated in step S21, which is used to measure the asymmetry of the distribution. The skewness of the standard normal distribution is 0. The kurtosis K is used to measure the thickness of the tail of the distribution. The kurtosis of the standard normal distribution is 3. The excess kurtosis calculated here is K-3. S222, for the data set that passes the test, the confidence interval is directly calculated with a 95% confidence level: Among them, (a,b) represents the confidence interval, is the sample mean, Z α / 2 is the critical value of the standard normal distribution, for 95% confidence, Zα / 2 is approximately 1.96, σ is the population standard deviation, n is the size of the data set, is the standard error; Step S223: The data set that fails the test is corrected using a skewness compensation algorithm, and then a second test is performed. The skewness compensation algorithm is as follows: For S>0, that is, right-skewed data, logarithmic transformation is performed; for S<0, that is, left-skewed data, square transformation is performed, ε represents a small disturbance term to avoid taking the logarithm of zero value, and then recalculate D for the transformed data shape If the test is passed, the confidence interval is calculated; if not, step S3 is executed.

6. The method for optimizing dredging construction parameters based on normality test according to claim 1, characterized in that: The step S3 comprises: Step S31: construct the probability density function of each data group through kernel density estimation, regard each data point as a kernel function center, and superimpose all kernel functions to obtain the overall density distribution. The formula is as follows: Among them, x i represents the i-th data point of the data set, x is the location of the density to be estimated, n is the number of data in the data set, h is the bandwidth, which is used to control the smoothness of the density curve, and F is the kernel function; Step S32, the rule for judging abnormal values ​​is: set the threshold value β through experience, if the density value f(x i ) satisfies f(x i )<β·max(f(x)), it is judged as an abnormal value; Step S33: Set the dynamic bandwidth w, which is improved by combining the IQR and the standard deviation of the data set. The formula is as follows: IQR=Q3-Q1 (7) Among them, Q1 and Q3 represent the first quartile and the third quartile of the data. After sorting the data from small to large, the data at the 25th percentile is Q1, and the data at the 75th percentile is Q3. Data lower than Q1-1.5×IQR or higher than Q3+1.5×IQR are considered abnormal data. Where σ is the standard deviation of the data set, and the coefficient γ is dynamically adjusted according to the data sparsity. If the data is sparsely distributed, γ is increased to widen the bandwidth; if the data is densely distributed, γ is reduced to narrow the bandwidth.

7. The method for optimizing dredging construction parameters based on normality test according to claim 1, characterized in that: In step S4, after removing the outliers, the normality test is performed again on the remaining data; if the data passes the normality test, the 95% confidence interval is calculated, and the outliers and missing values ​​previously removed are replaced with the mean of the confidence interval.

8. The method for optimizing dredging construction parameters based on normality test according to claim 1, characterized in that: The step S5 comprises: Step S51: Calculate the weight of each data point based on the kernel density estimation result; the formula is as follows: Among them, f(x i ) represents the data point x i The kernel density estimate of represents the sum of the density values ​​of all data points, Step S52: group the data according to the weight ω i Perform sampling with replacement to generate B resampled sample sets For each resampling set Both contain n data points; Step S53, weighted statistics calculation: in, represents the weighted mean of the b-th resampled set, ω i is the data point x i The weight of Step S54, confidence interval estimation: For B weighted means The ranking was performed and the 2.5% and 97.5% quantiles were taken as the 95% confidence intervals.

Images