A method and system for dividing operating conditions of ship main engine

Abstract

The present invention provides a method and system for dividing operating conditions of a ship main engine. The method first collects ship data, and calculates the main engine power average value, main engine speed average value, seawater temperature average value, supercharger speed average value and cylinder exhaust temperature average value in the ship data within a unit time, then uses a k-means clustering algorithm to cluster and divide the main engine power average value and the main engine speed average value to obtain multiple clusters, then uses a convolution smoothing algorithm to fill in the missing data, and finally uses a Gaussian mixture model clustering algorithm to divide the working conditions of the supplemented seawater temperature average value, supercharger speed average value and cylinder exhaust temperature average value, and divides multiple main engine working conditions under each cluster, which can effectively improve the ability to divide the main engine working conditions.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a method and system for dividing operating conditions of a ship main engine. Background Art

[0002] The daily operation cost of a ship is very high, and the effective operation time determines the profit level of the shipowner. The ship's main engine is the core power device of the ship, and some failures will inevitably occur during its use. The occurrence of these failures will seriously affect the normal operation of the ship, not only affecting the normal operation of the equipment, but also causing accidents in serious cases, and even endangering personal safety.

[0003] The operating status of the ship's main engine is an important indicator of the ship's navigation status. At present, the research on ship main engines at home and abroad mainly focuses on the fault analysis and prediction of the main engine. There are few studies on how to identify different working conditions and conduct fault analysis according to different working conditions. The limited research still has problems such as the inability to finely divide the main engine working conditions under normal ship operation, the slow division of working conditions and the lack of consideration.

[0004] The main engine operating condition classification is the main basis for the subsequent realization of main engine efficiency, fault prediction, and operation analysis. According to the ship equipment information and navigation information collected from the actual ship, and according to the physical prototype of the main engine operation, the main engine operating condition is reasonably divided, which requires consideration of multiple equipment parameters, including parameters of complex mechanisms such as turbocharger parameters and cooling water parameters.

[0005] In actual engineering applications, many equipment parameters are difficult to obtain and sometimes data is lost. Changes in the main engine operating conditions are also affected by temperature and machine aging. For example, the operating conditions of a ship just starting to operate are different from those of a ship that has been in operation for many years. Therefore, dividing the reasonable ship main engine operating conditions based on actual ship data can lay the foundation for determining ship pollutant emissions, estimating fuel consumption, evaluating main engine performance, and diagnosing and predicting faults of key main engine equipment, and provides a reference for ship equipment management and maintenance.

[0006] The main engine operating condition is a coupled process. The main equipment will affect each other and cause changes in the operating conditions. However, it is difficult for an actual ship to fully output the parameters required by the main engine operation physical model. Therefore, according to the existing methods, this operating condition division is somewhat difficult. Summary of the invention

[0007] In order to solve the problems of insufficient division and low efficiency in the existing main engine working condition division process, the present invention provides a ship main engine working condition division method, which uses a convolution smoothing algorithm to make up for missing data and uses a cluster analysis method to mine various main engine working conditions, which can effectively improve the main engine working condition division capability and more finely divide the main engine working conditions under normal ship operation. The present invention also relates to a ship main engine working condition division system.

[0008] The technical solution of the present invention is as follows:

[0009] A method for dividing operating conditions of a ship main engine, characterized in that it comprises the following steps:

[0010] Data collection and calculation steps: collecting ship data, the ship data including main engine power, main engine speed, seawater temperature, supercharger speed and cylinder exhaust temperature, and calculating the average value of main engine power, main engine speed, seawater temperature, supercharger speed and cylinder exhaust temperature per unit time;

[0011] Data completion steps: Use the k-means clustering algorithm to cluster and divide multiple host power averages and multiple host speed averages within a certain period of time to obtain multiple clusters. According to the data collection of multiple data points in each cluster at a certain moment, it is judged whether the average seawater temperature, the average supercharger speed or the average cylinder exhaust temperature is missing. If a certain data is missing, the convolution smoothing algorithm is used to complete the missing data.

[0012] Working condition division steps: Gaussian mixture model clustering algorithm is used to divide the completed average seawater temperature, average turbocharger speed and average cylinder exhaust temperature into working conditions, and multiple host working conditions under each cluster are divided.

[0013] Preferably, in the data completion step, using a convolution smoothing algorithm to complete missing data specifically includes:

[0014] If the data of a data point in the previous sub-time period at a certain moment is complete, the data in the previous sub-time period is used to complete it;

[0015] If the data of a data point in the previous sub-time period at a certain moment is incomplete, the data of the data point in the next sub-time period at a certain moment is used to complete it;

[0016] If the data of a data point in the next sub-time period after a certain moment is incomplete, all the data of the data point will be removed from the cluster.

[0017] Preferably, in the data completion step, the k-means clustering algorithm uses Euclidean distance measurement to perform distance calculation.

[0018] Preferably, in the working condition division step, the Gaussian mixture model clustering algorithm is trained using the expectation maximization method.

[0019] A ship main engine operating condition classification system, characterized by comprising a data acquisition and calculation module, a data completion module and an operating condition classification module connected in sequence,

[0020] Data collection and calculation module: collects ship data, including main engine power, main engine speed, seawater temperature, supercharger speed and cylinder exhaust temperature, and calculates the average value of main engine power, main engine speed, seawater temperature, supercharger speed and cylinder exhaust temperature per unit time;

[0021] Data completion module: The k-means clustering algorithm is used to cluster the average power values ​​and the average speed values ​​of multiple main engines within a certain period of time to obtain multiple clusters. The data collection of multiple data points in each cluster at a certain moment is used to determine whether the average seawater temperature, the average supercharger speed, or the average cylinder exhaust temperature is missing. If a certain data is missing, the convolution smoothing algorithm is used to complete the missing data.

[0022] Working condition division module: The Gaussian mixture model clustering algorithm is used to divide the completed average seawater temperature, average turbocharger speed and average cylinder exhaust temperature into working conditions, and multiple host working conditions are divided under each cluster.

[0023] Preferably, in the data completion module, using the convolution smoothing algorithm to complete the missing data specifically includes:

[0024] If the data of a data point in the previous sub-time period at a certain moment is complete, the data in the previous sub-time period is used to complete it;

[0025] If the data of a data point in the previous sub-time period at a certain moment is incomplete, the data of the data point in the next sub-time period at a certain moment is used to complete it;

[0026] If the data of a data point in the next sub-time period after a certain moment is incomplete, all the data of the data point will be removed from the cluster.

[0027] Preferably, the k-means clustering algorithm uses Euclidean distance measurement for distance calculation.

[0028] Preferably, the Gaussian mixture model clustering algorithm is trained using the expectation maximization method.

[0029] The beneficial effects of the present invention are:

[0030] The present invention provides a method for dividing the working conditions of a ship main engine. The method adopts a k-means clustering algorithm to cluster and divide a plurality of main engine power average values ​​and a plurality of main engine speed average values ​​in the collected ship data to obtain a plurality of clusters, so as to divide a stable power range interval; and uses a convolution smoothing algorithm to make up for the missing data in a plurality of data points under each cluster at a certain moment, and then uses a Gaussian mixture model clustering algorithm to mine out various working conditions of the main engine. The various working conditions can be distinguished without considering a complex physical model of the main engine operation. The method does not require too many or overly detailed signal points to supplement the analysis of the main engine operation conditions, can effectively improve the main engine working condition division capability, solves the problems of insufficient division and low efficiency in the current main engine working condition division process, and divides the main engine working conditions under normal operation of the ship in a more refined manner.

[0031] The present invention also relates to a ship main engine operating condition classification system, which corresponds to the above-mentioned ship main engine operating condition classification method, and can be understood as a system for implementing the above-mentioned ship main engine operating condition classification method, including a data acquisition and calculation module, a data completion module and an operating condition classification module connected in sequence, and each module works together to make up for the missing data by using a convolution smoothing algorithm, and uses a Gaussian mixture model clustering algorithm to divide the working conditions of the completed seawater temperature average value, supercharger speed average value and cylinder exhaust temperature average value, that is, using a clustering analysis method to mine various main engine working conditions, which can effectively improve the main engine working condition classification capability and more finely divide the main engine working conditions under normal operation of the ship. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 It is a flow chart of the method for dividing the operating conditions of a ship main engine of the present invention.

[0033] Figure 2 It is a preferred flow chart of the method for dividing the operating conditions of the main engine of a ship according to the present invention. DETAILED DESCRIPTION

[0034] The present invention will be described below in conjunction with the accompanying drawings.

[0035] The present invention relates to a method for extracting characteristic values ​​of ship main engine operating conditions. The flow chart of the method is as follows: Figure 1 As shown, the following steps are included in sequence:

[0036] Data collection and calculation steps: collect ship data, that is, second-level data, including main engine power, main engine speed, seawater temperature, supercharger speed and cylinder exhaust temperature, and calculate the average main engine power, main engine speed, seawater temperature, supercharger speed and cylinder exhaust temperature per unit time (that is, per minute).

[0037] Data completion steps: The k-means clustering algorithm is used to cluster the average power values ​​of multiple main engines and the average speed values ​​of multiple main engines within a certain period of time to obtain multiple clusters. The data collection of multiple data points in each cluster at a certain moment is used to determine whether the average seawater temperature, the average supercharger speed, or the average cylinder exhaust temperature is missing. If any data is missing, the convolution smoothing algorithm is used to complete the missing data.

[0038] Specifically, Figure 2 In the preferred flowchart shown in FIG. 1 , the k-means clustering algorithm is used to cluster the average power value of the host and the average speed value of the host. The k value is positioned as n (assuming it is 4), that is, n clusters are obtained. In other words, the stable power range interval is divided based on the k-means clustering algorithm. The k-means clustering algorithm can use the Euclidean distance measurement to calculate the distance. The specific calculation steps are as follows:

[0039] 1) Select a suitable k value and input the sample set as D = {x1, x2, ... x m}, the output cluster is divided into C = {C1, C2, ... C k};

[0040] 2) Randomly select k samples from the sample set D as the initial k centroid vectors {μ1,μ2,…,μ k};

[0041] 3) Calculate sample x i and each centroid vector μ j The distance d between (j=1,2,…k) ij :

[0042]

[0043] 4) Output cluster C j Recalculate the new centroid vector μ for all sample points in j :

[0044]

[0045] In the above formula, x is a single sample.

[0046] 5) If all k centroid vectors remain unchanged, the output cluster partition C = {C1, C2, ... C k}, if there is any change, repeat the above steps until convergence.

[0047] Then, the missing data are supplemented for the data under the n clusters in turn: by judging whether the average seawater temperature, the average supercharger speed or the average cylinder exhaust temperature is missing through the data collection situation of multiple data points under each cluster at a certain moment, for example, at a certain moment A, the average supercharger speed is not collected in the j-th data point under the i-th cluster (i∈1~n), then the average supercharger speed of the j-th data point under the i-th cluster is empty (that is, a null value), that is, it is judged that the average supercharger speed of the j-th data point under the i-th cluster is missing.

[0048] If there is missing data, the data in the data points under n clusters are supplemented in turn. Assume that the missing data is supplemented for the i-th cluster (i∈1~n), and there are m data points under the i-th cluster, for example, m=1000, that is, there are 1000 data points under the i-th cluster.

[0049] First, determine whether the average seawater temperature, the average speed of all superchargers, or the average exhaust temperature of all cylinders is missing for each data point in the i-th cluster. Assume that the average seawater temperature of the j-th data point (j∈1~m) in this cluster is missing. If the average seawater temperature data of the previous sub-time period of the j-th data point, such as the previous 5 minutes, is complete, then use the average seawater temperature data of the previous 5 minutes to complete it.

[0050] Assuming that the average seawater temperature data of the previous 5 minutes is A = [a1 a2 a3 a4 a5], and the smoothing window B = [b1 b2b3 b4 b5], then the average seawater temperature data of the jth data point is the value of A convolved with B, that is, the missing data value is filled;

[0051] If the average seawater temperature data of the 5 minutes before the jth data point is incomplete, the average seawater temperature data of the next sub-time period, such as the next 5 minutes, is used to complete it;

[0052] Assuming that the average seawater temperature data of the last 5 minutes is A = [a6 a7 a8 a9 a10], and the smoothing window B = [b5 b4b3 b2 b1], the average seawater temperature data of the jth data point is the value of A convolved with B, that is, the missing data value is filled;

[0053] If the average seawater temperature data 5 minutes after the jth data point is incomplete, all data of the jth data point will be removed from the i-th cluster.

[0054] Working condition division steps: Use the Gaussian mixture model clustering algorithm to divide the three types of data, namely the completed average seawater temperature, the average turbocharger speed and the average cylinder exhaust temperature, into working conditions, and divide multiple host working conditions under each cluster. Set the number of working conditions to k, and get n*k host working conditions.

[0055] The Gaussian mixture model clustering algorithm (GMM) assumes that all data are generated from a mixture of Gaussian distributions of finite data and unknown parameters. This is a probability model based on maximum likelihood estimation. The Gaussian mixture model can be considered as a combination of M single Gaussian probability density functions, each with a corresponding mean and covariance to incorporate information about the data covariance structure and potential Gaussian centers.

[0056] Its expression is as follows:

[0057]

[0058] In the above formula, p(x) is M Gaussian probability density functions, x is a single sample, N(x|μ k ,∑ k ) is the Gaussian distribution density function of the kth sub-model; μ k is the sample mean of the kth sub-model, ∑ k is the covariance of the kth sub-model; π k is the weight of the kth Gaussian distribution and satisfies the constraint of formula (4).

[0059]

[0060] Solve formula (3) and find the probability of each data point in the M models, which is the weight. The distribution range of the M models is determined based on the weights of all data points.

[0061] The present invention also relates to a ship main engine operating condition classification system, which corresponds to the above-mentioned ship main engine operating condition classification method and can be understood as a system for implementing the above-mentioned method. The system includes a data acquisition and calculation module, a data completion module and an operating condition classification module connected in sequence. Specifically,

[0062] Data collection and calculation module: collects ship data, including main engine power, main engine speed, seawater temperature, supercharger speed and cylinder exhaust temperature, and calculates the average value of main engine power, main engine speed, seawater temperature, supercharger speed and cylinder exhaust temperature per unit time;

[0063] Data completion module: The k-means clustering algorithm is used to cluster the average power values ​​and the average speed values ​​of multiple main engines within a certain period of time to obtain multiple clusters. The data collection of multiple data points in each cluster at a certain moment is used to determine whether the average seawater temperature, the average supercharger speed, or the average cylinder exhaust temperature is missing. If a certain data is missing, the convolution smoothing algorithm is used to complete the missing data.

[0064] Working condition division module: The Gaussian mixture model clustering algorithm is used to divide the completed average seawater temperature, average turbocharger speed and average cylinder exhaust temperature into working conditions, and multiple host working conditions are divided under each cluster.

[0065] Preferably, in the data completion module, using the convolution smoothing algorithm to complete the missing data specifically includes:

[0066] If the data of a data point in the previous sub-time period at a certain moment is complete, the data in the previous sub-time period is used to complete it;

[0067] If the data of a data point in the previous sub-time period at a certain moment is incomplete, the data of the data point in the next sub-time period at a certain moment is used to complete it;

[0068] If the data of a data point in the next sub-time period after a certain moment is incomplete, all the data of the data point will be removed from the cluster.

[0069] Preferably, the k-means clustering algorithm uses the Euclidean distance measure for distance calculation.

[0070] Preferably, the Gaussian mixture model clustering algorithm is trained using the expectation maximization method.

[0071] The present invention provides an objective and scientific method and system for dividing the operating conditions of a ship's main engine. By using a convolution smoothing algorithm to make up for missing data and using a clustering analysis method to mine out various main engine operating conditions, the main engine operating condition division capability can be effectively improved, and the main engine operating conditions under normal ship operation can be divided more finely.

[0072] It should be noted that the above-described specific implementations can enable those skilled in the art to more fully understand the invention, but do not limit the invention in any way. Therefore, although this specification has described the invention in detail with reference to the drawings and embodiments, those skilled in the art should understand that the invention can still be modified or replaced by equivalents. In short, all technical solutions and improvements that do not deviate from the spirit and scope of the invention should be included in the protection scope of the patent for the invention.

Claims

1. A method for dividing operating conditions of a ship main engine, characterized in that: The following steps are involved: Data collection and calculation steps: collecting ship data, the ship data including main engine power, main engine speed, seawater temperature, supercharger speed and cylinder exhaust temperature, and calculating the average value of main engine power, main engine speed, seawater temperature, supercharger speed and cylinder exhaust temperature per unit time; Data completion step: use k-means clustering algorithm to cluster and divide multiple host power average values ​​and multiple host speed average values ​​within a certain period of time, obtain multiple clusters to divide the stable power range interval, and judge whether the average seawater temperature, the average supercharger speed or the average cylinder exhaust temperature is missing through the data collection of multiple data points under each cluster at a certain moment. If a certain data is missing, the convolution smoothing algorithm is used to complete the missing data. The convolution smoothing algorithm is used to complete the missing data. Specifically, it is judged whether the data of a certain data point under a certain cluster in the previous sub-time period at a certain moment is complete. If the data of a certain data point in the previous sub-time period at a certain moment is complete, the data in the previous sub-time period is used to complete the data. The average seawater temperature data A of the previous sub-time period =[a1, a2, a3, a4, a5], smoothing window B=[b1, b2, b3, b4, b5], then the average seawater temperature data of a certain data point is the value of A convolved with B to complete the missing data value; if the data of a certain data point in the previous sub-time period at a certain moment is incomplete, then the data of the data point in the next sub-time period at a certain moment is used to complete it, and the average seawater temperature data of the next sub-time period A=[a6, a7, a8, a9, a10], smoothing window B=[b5, b4, b3, b2, b1], then the average seawater temperature data of a certain data point is the value of A convolved with B to complete the missing data value; if the data of a certain data point in the next sub-time period at a certain moment is incomplete, then all the data of the data point are removed from the cluster where it is located; Working condition division steps: Use the Gaussian mixture model clustering algorithm to mine the various operating conditions of the main engine, divide the working conditions of the completed average seawater temperature, average turbocharger speed and average cylinder exhaust temperature, and finely divide multiple main engine working conditions under each cluster under normal operation of the ship.

2. The method for dividing the operating conditions of a ship main engine according to claim 1, characterized in that: In the data completion step, the k-means clustering algorithm uses the Euclidean distance measure to perform distance calculation.

3. The method for dividing operating conditions of a ship main engine according to claim 1, characterized in that: In the working condition division step, the Gaussian mixture model clustering algorithm is trained using the expectation maximization method.

4. A ship main engine operating condition classification system, characterized in that: It includes a data acquisition and calculation module, a data completion module and a working condition division module connected in sequence. Data collection and calculation module: collects ship data, including main engine power, main engine speed, seawater temperature, supercharger speed and cylinder exhaust temperature, and calculates the average value of main engine power, main engine speed, seawater temperature, supercharger speed and cylinder exhaust temperature per unit time; Data completion module: The k-means clustering algorithm is used to cluster and divide the average power values ​​of multiple main engines and the average speed values ​​of multiple main engines within a certain period of time, and multiple clusters are obtained to divide the stable power range interval. The data collection of multiple data points under each cluster at a certain moment determines whether the average seawater temperature, the average supercharger speed or the average cylinder exhaust temperature is missing. If a certain data is missing, the convolution smoothing algorithm is used to complete the missing data. The convolution smoothing algorithm is used to complete the missing data. Specifically, it is determined whether the data of a certain data point under a certain cluster in the previous sub-time period at a certain moment is complete. If the data of a certain data point in the previous sub-time period at a certain moment is complete, the data in the previous sub-time period is used to complete the data. The average seawater temperature data A of the previous sub-time period =[a1, a2, a3, a4, a5], smoothing window B=[b1, b2, b3, b4, b5], then the average seawater temperature data of a certain data point is the value of A convolved with B to complete the missing data value; if the data of a certain data point in the previous sub-time period at a certain moment is incomplete, then the data of the data point in the next sub-time period at a certain moment is used to complete it, and the average seawater temperature data of the next sub-time period A=[a6, a7, a8, a9, a10], smoothing window B=[b5, b4, b3, b2, b1], then the average seawater temperature data of a certain data point is the value of A convolved with B to complete the missing data value; if the data of a certain data point in the next sub-time period at a certain moment is incomplete, then all the data of the data point are removed from the cluster where it is located; Working condition division module: The Gaussian mixture model clustering algorithm is used to mine the various working conditions of the main engine, and the completed average seawater temperature, average turbocharger speed and average cylinder exhaust temperature are divided into working conditions, and multiple main engine working conditions under each cluster under normal operation of the ship are finely divided.

5. The ship main engine operating condition classification system according to claim 4, characterized in that: The k-means clustering algorithm uses the Euclidean distance measure for distance calculation.

6. The ship main engine operating condition classification system according to claim 4, characterized in that: The Gaussian mixture model clustering algorithm is trained using the expectation maximization method.

Images