Dispersion estimation method for overflow flow and overflow density of trailing suction hopper dredger
By applying the cascading observer of the CD-FPF-UKF combination algorithm on the rake suction dredger, the problem of difficult measurement of overflow flow and density is solved, and higher estimation accuracy and shorter estimation time are achieved, cabin loading performance is optimized and crew operation burden is reduced.
Patent Information
- Application Number
- CN202510063389.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-01-15
- Publication Date
- 2025-05-16
AI Technical Summary
During the cabin loading operation, the rake suction dredger has difficulty in accurately measuring the overflow flow and overflow density due to the presence of air in the overflow weir, which increases the difficulty and workload of the crew.
The cascading observer using the CD-FPF-UKF combination algorithm collects real-time data and performs denoising processing, combines the mud tank deposition model and the Wiener process model to disperse the overflow flow and overflow density.
It improves the accuracy and accuracy of overflow loss estimation, shortens the estimation time, optimizes the cabin loading performance, reduces the operating burden of crew members, and supports intelligent decision-making.
Smart Images
Figure CN120012395A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of tank loading optimization in a dredging process of a trailing suction hopper dredger, and in particular to a method for estimating the overflow flow and overflow density dispersion of a trailing suction hopper dredger. Background Art
[0002] As a type of hydraulic dredger, the trailing suction dredger uses the drag head to dig the bottom mud and sand during navigation and transports it to the mud tank for sedimentation through pipelines to periodically complete the dredging and loading operations. Compared with other dredgers, it has the unique advantages of self-propelled, self-dig, self-loading, self-unloading, deep digging, and flexible maneuverability. It has been widely used in the construction and maintenance of ports, waterways, beaches, silt removal and flood control, land reclamation and many other engineering fields.
[0003] After the dredger arrives at a specific construction area, as the loading and overflowing operation proceeds, the mud and sand excavated by the drag head gradually settle in the cabin to form a sand bed, while the low-density mud-water mixture will flow out of the ship through the overflow weir, resulting in a decrease in the loading quality and an extension of the loading time, which in turn affects the output and efficiency of the dredger during the dredging operation. Timely and accurate acquisition of overflow flow and overflow density helps the crew to grasp the sedimentation of mud and sand in the mud cabin in real time, thereby facilitating the adjustment of dredging strategies and correction of equipment parameters. However, due to the presence of air in the overflow weir and other reasons, the overflow flow and overflow density cannot be measured by instruments and equipment. The crew can only rely on their own ability and experience to judge the size of the overflow loss, which undoubtedly increases the difficulty and workload of the crew to operate the equipment. Although the overflow loss can be estimated by establishing a mud cabin deposition model and combining it with actual ship data, the model contains difficult-to-handle time-varying soil parameters, which makes it difficult to obtain the overflow flow and density. Therefore, it is necessary to explore practical and feasible methods to make a practical and effective estimate of the flow and density of the mixture in the overflow weir, and get rid of the dependence on the crew's construction ability and experience. This is of great significance for the optimization of the loading performance of the bucket suction dredger and the creation of an auxiliary intelligent decision support system. Summary of the invention
[0004] In view of this, the present invention provides a method for estimating the overflow flow and overflow density dispersion of a trailing suction hopper dredger, which improves the accuracy and precision of overflow loss estimation and shortens the estimation time, thereby optimizing the loading performance.
[0005] The present invention achieves the above technical objectives through the following technical means.
[0006] A method for estimating the overflow flow and overflow density dispersion of a trailing suction hopper dredger:
[0007] Collect real-time data of the trailing suction hopper dredger during normal dredging and loading operations, and perform denoising processing;
[0008] According to the volume and mass balance equations of the mud tank deposition model, the Wiener process model equations of the mud-water mixture overflow flow and overflow density, the state equations and observation equations of the overflow subsystem 1 and the state equations and observation equations of the overflow subsystem 2 are determined respectively;
[0009] A cascade observer of the CD-FPF-UKF combined algorithm is constructed to perform decentralized estimation of the overflow flow and overflow density: Observer 1 based on the CD-FPF algorithm uses the state equation and observation equation of the overflow subsystem 1 to estimate the overflow flow; Observer 2 based on the UKF algorithm uses the state equation and observation equation of the overflow subsystem 2 and the overflow flow estimation result of observer 1 to estimate the overflow density.
[0010] Furthermore, the real-time data of the trailing suction hopper dredger during normal dredging and loading operations include the total mass and volume of the mud-water mixture, the flow rate and density of the mud-water mixture entering the tank, and the height of the overflow weir.
[0011] Furthermore, the balance equation of the volume and mass of the mud tank deposition model is:
[0012]
[0013] Where V t is the total volume of the mud-water mixture in the mud tank, m t is the total mass of the mud-water mixture in the mud tank, Q i is the flow rate of mud-water mixture into the tank, ρ i is the density of the mud-water mixture entering the cabin, Q o is the overflow flow of mud-water mixture, ρ o It is the overflow density of mud-water mixture.
[0014] Furthermore, the Wiener process model equation of the overflow flow and overflow density of the mud-water mixture is:
[0015]
[0016] In the formula, Q o,t+1 is the overflow flow of the mud-water mixture at time t+1, ρ o,t+1 is the overflow density of the mud-water mixture at time t+1, ω q,t is the process noise of the overflow flow of the mud-water mixture at time t, ω ρ,t is the process noise of the mud-water mixture overflow density at time t.
[0017] Furthermore, the state vector x of overflow subsystem 1 is 1 , input variable u 1 and the observed variable z 1 Defined as: 1 =[V t Qo ] T ,u 1 =Q i , z 1 =V t .
[0018] Furthermore, the state equation and observation equation of the overflow subsystem 1 are:
[0019]
[0020] In the formula, represents the first state variable of overflow subsystem 1 at time t, represents the input variable of overflow subsystem 1 at time t, represents the second state variable of overflow subsystem 1 at time t, represents the first component of the process noise of overflow subsystem 1 at time t, represents the second component of the process noise of overflow subsystem 1 at time t, represents the observed variable of overflow subsystem 1 at time t, is the measurement noise of overflow subsystem 1 at time t, T s It is the sampling interval of the shipboard data acquisition system.
[0021] Furthermore, the state vector x of overflow subsystem 2 is 2 , input vector u 2 and the observed variable z 2 Defined as: 2 =[m t ρ o ] T ,u 2 =[Q i ρ i Q o ] T , z 2 =m t .
[0022] Furthermore, the state equation and observation equation of the overflow subsystem 2 are:
[0023]
[0024] In the formula, represents the first state variable of overflow subsystem 2 at time t, represents the second state variable of overflow subsystem 2 at time t, represents the first component of the process noise of overflow subsystem 2 at time t, represents the second component of the process noise of overflow subsystem 2 at time t, represents the observed variable of overflow subsystem 2 at time t, represents the first input variable of overflow subsystem 2 at time t, represents the second input variable of overflow subsystem 2 at time t, represents the third input variable of overflow subsystem 2 at time t, is the measurement noise of overflow subsystem 2 at time t, T s It is the sampling interval of the shipboard data acquisition system.
[0025] Furthermore, the denoising process is specifically performed by using a Savitzky-Golay filtering method in the time domain.
[0026] Furthermore, the method also includes using normalized mean square error and absolute error as evaluation functions for evaluating the performance of the cascade observer.
[0027] Compared with the prior art, the present invention has the following advantages and beneficial effects:
[0028] (1) The present invention realizes the decentralized estimation of overflow flow and overflow density by constructing the cascade observer of the CD-FPF-UKF combined algorithm of the trailing suction hopper dredger, and effectively solves the problem that the overflow loss of the dredger is difficult to accurately measure and effectively estimate. Moreover, the crew can grasp the actual situation of mud and sand deposition in the mud tank of the trailing suction hopper dredger in real time according to the instantaneous estimated values of the overflow flow and overflow density, and then adjust the dredging strategy in time, optimize the loading performance, and improve the dredging output.
[0029] (2) Since the cascade observer structure designed by the present invention includes the CD-FPF algorithm that introduces the concept of pseudo time, it can solve the problem of significant decrease in observer accuracy and efficiency under larger sampling intervals. BRIEF DESCRIPTION OF THE DRAWINGS
[0030] Figure 1 A flow chart of an implementation method of the present invention;
[0031] Figure 2 is a schematic diagram of a cascade observer of the present invention;
[0032] FIG3( a ) is a graph showing the absolute error of the first ship overflow flow dispersion estimation in the Yangtze River Estuary in the present invention;
[0033] FIG3( b ) is a graph showing the absolute error of the first ship overflow density dispersion estimation in the Yangtze River Estuary in the present invention. DETAILED DESCRIPTION
[0034] The present invention is further described below in conjunction with the accompanying drawings and specific embodiments, but the protection scope of the present invention is not limited thereto.
[0035] The present invention utilizes a cascade observer of a CD-FPF-UKF (Continuous-Discrete Time Feedback Particle Filter-Unscented Kalman Filter) combined algorithm to estimate the overflow flow and overflow density of a trailing suction hopper dredger during the loading operation, and obtains the instantaneous overflow loss (the product of the overflow flow and the density), so as to further grasp the sedimentation of mud and sand in the mud tank in real time.
[0036] like Figure 1 As shown, a method for estimating the overflow flow and overflow density dispersion of a trailing suction dredger based on a CD-FPF-UKF combined algorithm of the present invention comprises the following steps:
[0037] S1: Collect real-time data of the trailing suction hopper dredger during normal dredging and loading operations.
[0038] S1.1, using different types of instruments and equipment such as ship draft sensors, liquid level radar sensors, electromagnetic flowmeters, gamma-ray densitometers, and position sensors, the total mass and volume of the mud-water mixture in the mud tank of the trailing suction dredger, the mud-water mixture inlet flow and density, and the overflow weir height are measured as real-time data sets.
[0039] S2: Use appropriate and effective methods to eliminate noise interference in the data.
[0040] S2.1, the Savitzky-Golay filtering method in the time domain is used to denoise the real-time data set. This method is based on the polynomial least squares method and uses a moving window to smooth the data noise. The expression of the smoothing process is:
[0041]
[0042] Where: and are the data sets to be smoothed and smoothed, a0 and a i is the smoothing coefficient, m and n are the width of the sliding window and the number of data contained therein, respectively, and n=2m+1.
[0043] S3: According to the balance equations of the volume and mass of the mud tank sedimentation model and the Wiener process model equations of the overflow flow and overflow density, the state and observation equations of the overflow subsystems 1 and 2 are determined respectively.
[0044] S3.1, the volume and mass balance equation of the mud tank sediment model is:
[0045]
[0046] Where: V t and m t are the total volume and total mass of the mud-water mixture in the mud tank, Q i and ρ i are the flow rate and density of mud-water mixture entering the cabin, Q o and ρ o are the overflow flow rate and overflow density of mud-water mixture respectively.
[0047] Using the Euler method to discretize the volume and mass balance equation (2), we can obtain:
[0048]
[0049] Where: V t,t 、V t,t+1 、m t,t and m t,t+1 are the volume and mass of the mud-water mixture at time t and t+1 respectively, Q i,t and ρ i,t are the flow rate and density of the mud-water mixture entering the tank at time t, Q o,t and ρ o,t are the overflow flow and overflow density of mud-water mixture at time t, T s It is the sampling interval of the shipboard data acquisition system.
[0050] S3.2, in order to simplify the calculation of overflow flow and overflow density, it can be obtained using the Wiener process model, that is:
[0051]
[0052] Where: Q o,t+1 and ρ o,t+1 are the overflow flow and overflow density of the mud-water mixture at time t+1, ω q,t and ω ρ,t are the process noise of the mud-water mixture overflow flow and overflow density at time t respectively.
[0053] S3.3, Cascade observers design a separate observer for each subsystem, and the observers of each subsystem are interrelated and dependent on each other, which will significantly reduce the complexity of the estimation problem and the cost of calculation. Therefore, according to the discrete overflow loss estimation model (3)-(4), the state equation and observation equation of overflow subsystem 1 and overflow subsystem 2 are established respectively.
[0054] S3.3.1, the state vector x of overflow subsystem 1 1 , input variable u1 and the observed variable z 1 The definition is as follows:
[0055] x 1 =[V t Q o ] T , u 1 =Q i , z 1 =V t (5)
[0056] The state equation and observation equation of overflow subsystem 1 are:
[0057]
[0058] Where: is the measurement noise of overflow subsystem 1 at time t, represents the first state variable of overflow subsystem 1 at time t, represents the input variable of overflow subsystem 1 at time t, represents the second state variable of overflow subsystem 1 at time t, represents the first component of the process noise of overflow subsystem 1 at time t, represents the second component of the process noise of overflow subsystem 1 at time t, represents the observed variable of overflow subsystem 1 at time t; the measurement noise and process noise are independent zero-mean additive Gaussian noise sequences;
[0059] S3.3.2, the state vector x of overflow subsystem 2 2 , input vector u 2 and the observed variable z 2 The definition is as follows:
[0060] x 2 =[m t ρ o ] T ,u 2 =[Q i ρ i Q o ] T , z 2 =m t (7)
[0061] The state equation and observation equation of overflow subsystem 2 are:
[0062]
[0063] Where: is the measurement noise of overflow subsystem 2 at time t, represents the first state variable of overflow subsystem 2 at time t, represents the second state variable of overflow subsystem 2 at time t, represents the first component of the process noise of overflow subsystem 2 at time t, represents the second component of the process noise of overflow subsystem 2 at time t, represents the observed variable of overflow subsystem 2 at time t, represents the first input variable of overflow subsystem 2 at time t, represents the second input variable of overflow subsystem 2 at time t, represents the third input variable of overflow subsystem 2 at time t; the measurement noise and process noise therein are independent zero-mean additive Gaussian noise sequences.
[0064] S4: Figure 2 As shown, a cascade observer of the CD-FPF-UKF combined algorithm is constructed to perform decentralized estimation of the overflow flow and overflow density.
[0065] S4.1, observer 1 based on CD-FPF algorithm uses the state and observation equations (6) of overflow subsystem 1 to estimate the overflow flow rate, and uses the obtained results as the input variable of observer 2. The specific implementation process of CD-FPF algorithm is as follows. S4.1.1, for the estimation problem of continuous discrete time, the state equation and observation equation of the system are:
[0066]
[0067] In the formula, and are the state vector and input vector of overflow subsystem 1 at continuous time t, At discrete time t k The observation vector of overflow subsystem 1 at time ; f 1 (·)=(f1,…,f d ) and h 1 (·)=(h1,…,h m ) are the state transfer function and observation function of overflow subsystem 1, both of which belong to C 1 function; and are the independent zero-mean Gaussian process noise and measurement noise of overflow subsystem 1 with covariance matrices Q and R respectively.
[0068] S4.1.2 For continuous time t∈[t k-1 ,tk ), the dynamic change process of the i-th particle can be expressed by the following formula:
[0069]
[0070] In the formula, and are the i-th particle state and process noise of overflow subsystem 1 at time t, respectively, where N is the number of particles.
[0071] At the starting time t=0, all particles are sampled from the prior distribution p*(x,0), The right limit of is defined as:
[0072]
[0073] S4.1.3, when the latest measurement vector Z of overflow subsystem 1 k At discrete time t = t k When the measurement is obtained, a particle flow with the differential equation form is introduced
[0074]
[0075] Where, parameter λ is pseudo time; the initial condition of the i-th particle When is known, after iterative calculation, at time t=t k When , the state of the i-th particle is in and are the constant feedback gain function and the update term based on the innovation error, respectively, which can be calculated by equations (13) and (14).
[0076]
[0077] In the formula,
[0078] S4.2, observer 2 based on UKF algorithm uses the state and observation equation (8) of overflow subsystem 2 and the estimated result of overflow flow of observer 1 to estimate the overflow density. The specific implementation process of UKF algorithm is as follows.
[0079] S4.2.1, for the discrete-time estimation problem, the state equation and observation equation of the system are:
[0080]
[0081] In the formula, and are the state vector and input vector of overflow subsystem 2 at discrete time t, is the observation vector of overflow subsystem 2 at discrete time t; f 2 (·)=(f1,...,f d ) and h 2 (·)=(h1,...,h m ) are the nonlinear state transfer function and observation function of overflow subsystem 2 respectively; and The overflow subsystems 2 are independent zero-mean and their covariance matrices are Q 2 and R 2 Gaussian process noise and measurement noise.
[0082] S4.2.2, calculate the Sigma point set and the corresponding weights.
[0083]
[0084] In the formula, and are the sample mean and covariance of the Sigma point set of overflow subsystem 2 at time t, and are the state of the i-th Sigma point of overflow subsystem 2 at time t and the i-th column of the matrix square root, λ and n are the dimensions of the parameters to be determined and the system state variables, respectively. Parameter a is usually a small positive number, and parameter κ is a non-negative number.
[0085]
[0086] In the formula, and are the states of the i-th Sigma point of overflow subsystem 2 and covariance The corresponding weight, parameter β is a non-negative weight coefficient.
[0087] S4.2.3, time update.
[0088]
[0089] In the formula, and are the sample mean, covariance and observed mean of the one-step prediction of the Sigma point set of overflow subsystem 2, and are the state and observation values predicted by the i-th Sigma point of overflow subsystem 2 in one step, and is the process noise and observation noise of the i-th Sigma point of overflow subsystem 2.
[0090] S4.2.4, Measurement Update.
[0091]
[0092] In the formula, and They are the one-step prediction of the observation autocovariance matrix and the one-step prediction of the state observation cross-covariance matrix of overflow subsystem 2, and are the sample mean and covariance of the Sigma point set of overflow subsystem 2 at time t+1, is the observation vector of overflow subsystem 2 at discrete time t+1, K t+1 is the Kalman gain matrix at time t+1.
[0093] S5: nMSE (normalized mean square error) and AE (absolute error) are used as evaluation functions to evaluate the performance of the cascade observer.
[0094] S5.1, Estimates The nMSE measures the estimated value within a certain period of time. The difference between the true value x(t) and the real value x(t), which is defined in discrete time as:
[0095]
[0096] In the formula, and are the estimated value and true value at time t respectively, and M is the total number of samples.
[0097] S5.2, AE measures the estimated value at time t. With the true value The absolute value of the error between , which is defined as:
[0098]
[0099] The smaller the values of nMSE and AE, the better the performance of the cascade observer.
[0100] The data of three ship trips of the "Xinhaihu 8" trailing suction dredger during normal dredging and loading operations at Xiamen Port and the Yangtze River Estuary are selected as the sample data set (the data sampling interval is 30s). From the results in Table 2, it can be observed that under a larger sampling interval, the cascade observer of the CD-FPF-CG-UKF combination algorithm has a smaller cascade observer than the other four combination algorithms. and It shows that the cascade observer has a higher overflow flow Q o and overflow density ρ oIn addition, by comparing the centralized estimation results of the six centralized observers in Table 1, it can be found that the cascade observer of the CD-FPF-CG-UKF combined algorithm has a smaller and This shows that the proposed observer is superior to other centralized observers in terms of the accuracy and precision of overflow flow and overflow density estimation.
[0101] Table 1 Six centralized observers and result
[0102]
[0103] Table 2 Five cascade observers and result
[0104]
[0105]
[0106] It can be seen from Figure 3(a) and (b) that the absolute error of the cascade observer of the CD-FPF-UKF combination algorithm is smaller than that of the cascade observers of the other four combination algorithms, indicating that the accuracy and precision of the cascade observer are better than those of other cascade observers, and it can better realize the decentralized estimation of overflow flow and overflow density.
[0107] The present invention avoids the uncertain parameters and complex functional relationships in the traditional overflow loss model. Compared with the centralized observer based on a single estimation method, the modular design of the cascade observer makes it easy to debug, reduces the complexity of the estimation problem, has higher estimation accuracy and shorter estimation time, and is more suitable for online estimation of the overflow flow and overflow density of a trailing suction dredger at a larger sampling time interval.
[0108] The embodiments are preferred implementations of the present invention, but the present invention is not limited to the above-mentioned implementations. Any obvious improvements, substitutions or modifications that can be made by those skilled in the art without departing from the essential content of the present invention belong to the protection scope of the present invention.
Claims
1. A method for estimating the overflow flow and overflow density dispersion of a trailing suction hopper dredger, characterized by: Collect real-time data of the trailing suction hopper dredger during normal dredging and loading operations, and perform denoising processing; According to the volume and mass balance equations of the mud tank deposition model, the Wiener process model equations of the mud-water mixture overflow flow and overflow density, the state equations and observation equations of the overflow subsystem 1 and the state equations and observation equations of the overflow subsystem 2 are determined respectively; A cascade observer of the CD-FPF-UKF combined algorithm is constructed to perform decentralized estimation of the overflow flow and overflow density: Observer 1 based on the CD-FPF algorithm uses the state equation and observation equation of the overflow subsystem 1 to estimate the overflow flow; Observer 2 based on the UKF algorithm uses the state equation and observation equation of the overflow subsystem 2 and the overflow flow estimation result of observer 1 to estimate the overflow density.
2. The method for estimating overflow flow and overflow density dispersion of a trailing suction hopper dredger according to claim 1, characterized in that: The real-time data of the trailing suction hopper dredger during normal dredging and loading operations include the total mass and volume of the mud-water mixture, the flow rate and density of the mud-water mixture entering the tank, and the height of the overflow weir.
3. The method for estimating overflow flow and overflow density dispersion of a trailing suction hopper dredger according to claim 1, characterized in that: The volume and mass balance equation of the mud tank deposition model is: Where V t is the total volume of the mud-water mixture in the mud tank, m t is the total mass of the mud-water mixture in the mud tank, Q i is the flow rate of mud-water mixture into the tank, ρ i is the density of the mud-water mixture entering the cabin, Q o is the overflow flow of mud-water mixture, ρ o It is the overflow density of mud-water mixture.
4. The method for estimating overflow flow and overflow density dispersion of a trailing suction hopper dredger according to claim 3, characterized in that: The Wiener process model equations of the overflow flow and overflow density of the mud-water mixture are: In the formula, Q o,t+1 is the overflow flow of the mud-water mixture at time t+1, ρ o,t+1 is the overflow density of the mud-water mixture at time t+1, ω q,t is the process noise of the overflow flow of the mud-water mixture at time t, ω ρ,t is the process noise of the mud-water mixture overflow density at time t.
5. The method for estimating overflow flow and overflow density dispersion of a trailing suction hopper dredger according to claim 3, characterized in that: The state vector x of overflow subsystem 1 1 , input variable u 1 and the observed variable z 1 Defined as: 1 =[V t Q o ] T ,u 1 =Q i , z 1 =V t .
6. The method for estimating overflow flow and overflow density dispersion of a trailing suction hopper dredger according to claim 5, characterized in that: The state equation and observation equation of the overflow subsystem 1 are: In the formula, represents the first state variable of overflow subsystem 1 at time t, represents the input variable of overflow subsystem 1 at time t, represents the second state variable of overflow subsystem 1 at time t, represents the first component of the process noise of overflow subsystem 1 at time t, represents the second component of the process noise of overflow subsystem 1 at time t, represents the observed variable of overflow subsystem 1 at time t, is the measurement noise of overflow subsystem 1 at time t, T s It is the sampling interval of the shipboard data acquisition system.
7. The method for estimating overflow flow and overflow density dispersion of a trailing suction hopper dredger according to claim 3, characterized in that: The state vector x of overflow subsystem 2 2 , input vector u 2 and the observed variable z 2 Defined as: 2 =[m t ρ o ] T ,u 2 =[Q i ρ i Q o ] T , z 2 =m t .
8. The method for estimating overflow flow and overflow density dispersion of a trailing suction hopper dredger according to claim 7, characterized in that: The state equation and observation equation of the overflow subsystem 2 are: In the formula, represents the first state variable of overflow subsystem 2 at time t, represents the second state variable of overflow subsystem 2 at time t, represents the first component of the process noise of overflow subsystem 2 at time t, represents the second component of the process noise of overflow subsystem 2 at time t, represents the observed variable of overflow subsystem 2 at time t, represents the first input variable of overflow subsystem 2 at time t, represents the second input variable of overflow subsystem 2 at time t, represents the third input variable of overflow subsystem 2 at time t, is the measurement noise of overflow subsystem 2 at time t, T s It is the sampling interval of the shipboard data acquisition system.
9. The method for estimating overflow flow and overflow density dispersion of a trailing suction hopper dredger according to claim 1, characterized in that: The denoising process is specifically performed by using a Savitzky-Golay filtering method in the time domain.
10. The method for estimating overflow flow and overflow density dispersion of a trailing suction hopper dredger according to claim 1, characterized in that: It also includes the use of normalized mean square error and absolute error as evaluation functions for evaluating the performance of the cascade observer.