Method for identifying data deviation in ship structure monitoring, medium and electronic equipment
The method employs ARIMA models to detect and correct data deviations in ship structure monitoring, addressing sensor and proxy model inaccuracies, ensuring reliable monitoring results.
Patent Information
- Application Number
- CN202411180750.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-27
- Publication Date
- 2024-12-20
AI Technical Summary
Existing ship structure monitoring systems fail to accurately detect data deviations between sensor-collected data and proxy model outputs, leading to inaccurate monitoring results due to sensor failures and proxy model inaccuracies, which can cause irreversible errors.
A method using time series prediction models, specifically ARIMA, to identify data deviations by calculating residuals and statistical features from sensor and proxy model data, with thresholds to distinguish between sensor drift, failure, and proxy model inaccuracies.
Ensures accurate detection of data deviations, preventing error accumulation and maintaining reliable ship structure monitoring by correcting identified issues.
Smart Images
Figure CN119167215A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of ship engineering technology, and in particular to a method, a medium and an electronic device for identifying data deviation in ship structure monitoring. Background Art
[0002] Ship structure monitoring technology was first proposed by Lindemann in 1978, but it was only a theoretical design study without a specific application plan. With the development of electronic and computer technology, Lloyd's Register of Shipping (LR) and Ship & Marine Data Systems (Ship & Marine Data Systems) jointly developed a voyage recorder for very large crude carriers (VLCC) in 1989 to record hull structure stress data, but it did not have data processing and calculation capabilities.
[0003] In recent years, a variety of ship structure monitoring systems have been developed at home and abroad. Foreign systems include Italy's MON-HULL and CETNA, the UK's StressAlert and S&MDS, Norway's LightStruct, and South Korea's Global MaritimeEngineering. These systems use sensors to collect data and can achieve long-term real-time monitoring of hull structures. Domestic systems include the Long-term Monitoring System for Ship Structures (LOTEMS) developed by CSIC 702 Institute and the Real-time Monitoring and Evaluation System for Ship Structures developed by COSCO SHIPBUILDINGS.
[0004] Current ship structure monitoring technology assumes that both the data collected by sensors and the output data of proxy models are accurate. However, this is not always true in actual environments. Inaccurate input data will affect the output accuracy of ship structure monitoring.
[0005] Therefore, there is a need for a technical solution that can detect data deviations between the ship response data collected by sensors and the output data of a proxy model used to predict the ship structure response data. Summary of the invention
[0006] One purpose of the present application is to provide a method for identifying data deviations in ship structure monitoring, so as to solve the problem that it is difficult to detect data deviations of sensors and proxy models under the existing technology.
[0007] To achieve the above objectives, some embodiments of the present application provide a method for identifying data deviations in ship structure monitoring, the method comprising:
[0008] Acquire a first data sequence collected by a sensor in real time;
[0009] Determine, according to the first data sequence and a preset statistical period, a data statistical feature corresponding to the first data sequence, where the data statistical feature includes one or more combinations of the following: mean, standard deviation, skewness, and kurtosis;
[0010] Inputting the first data sequence into a pre-trained time series prediction model to obtain a predicted second data sequence, wherein the time series prediction model is used to predict the current data sequence based on the historical data sequence;
[0011] determining a residual between the second data sequence and the first data sequence;
[0012] The data deviation identification result is determined based on the data statistical characteristics and residuals corresponding to the first data sequence.
[0013] Furthermore, the time series prediction model is an autoregressive integrated moving average model.
[0014] Furthermore, the first data sequence is input into a pre-trained time series prediction model to obtain a predicted second data sequence, including:
[0015] Determining that the first data sequence is a stationary random sequence;
[0016] According to the calculation results of the autocorrelation function and the partial autocorrelation function, the parameter values corresponding to the autoregressive integrated moving average model are determined;
[0017] According to the parameters corresponding to the autoregressive integrated moving average model, train the autoregressive integrated moving average model;
[0018] According to the trained autoregressive integrated moving average model, a predicted second data sequence is obtained.
[0019] Furthermore, according to the calculation results of the autocorrelation function and the partial autocorrelation function, the parameter values corresponding to the autoregressive integrated moving average model are determined, including:
[0020] When the calculation result of the autocorrelation function is close to zero, the current value of the first parameter is determined as the value of the order of the autoregressive term in the parameter corresponding to the autoregressive integrated moving average model; or
[0021] When the calculation result of the partial autocorrelation function is close to zero, the current value of the second parameter is determined as the value of the order of the sliding average term in the parameter corresponding to the autoregressive integrated sliding average model.
[0022] Further, according to the data statistical characteristics and residuals corresponding to the first data sequence, a data deviation identification result is determined, including:
[0023] When the data statistical feature corresponding to the first data sequence deviates from the historical data statistical feature and the residual is less than a preset first threshold, determining that the data deviation identification result is a sensor signal drift; or
[0024] When the data statistical feature corresponding to the first data sequence deviates from the historical data statistical feature and the residual is greater than a preset first threshold, it is determined that the data deviation identification result is a sensor system failure.
[0025] Some embodiments of the present application also provide a method for identifying data deviations in ship structure monitoring, the method comprising:
[0026] Acquire a third data sequence collected in real time by the verification sensor, wherein the data collected by the verification sensor has a high correlation with the data collected by the non-verification sensor;
[0027] According to the data sequence collected in real time by the non-calibration sensor and the pre-established proxy model for predicting the ship structure response data, a fourth data sequence for predicting the collected data of the calibration sensor is obtained;
[0028] determining a residual between the third data sequence and the fourth data sequence;
[0029] When the residual is greater than a preset second threshold, a data deviation identification result is determined.
[0030] Further, determining the data deviation identification result includes:
[0031] Performing data deviation judgment on the third data sequence;
[0032] In the case where there is a data deviation in the third data sequence, the data deviation identification result is determined as a sensor failure, otherwise the data deviation identification result is determined as a poor fit of the proxy model.
[0033] Furthermore, performing data deviation judgment on the third data sequence includes:
[0034] Determine, according to the third data sequence and a preset statistical period, a data statistical feature corresponding to the third data sequence, where the data statistical feature includes one or more of the following combinations: mean, standard deviation, skewness, and kurtosis;
[0035] Inputting the third data sequence into a pre-trained time series prediction model to obtain a predicted fifth data sequence;
[0036] determining a residual between the third data sequence and the fifth data sequence;
[0037] The data deviation judgment result is determined based on the data statistical characteristics and residuals corresponding to the third data sequence.
[0038] Some embodiments of the present application further provide a computer-readable medium having computer-readable instructions stored thereon, wherein the computer-readable instructions can be executed by a processor to implement the aforementioned method for identifying data deviations in ship structure monitoring.
[0039] Some embodiments of the present application also provide an electronic device, which includes a memory for storing computer program instructions and a processor for executing the computer program instructions, wherein when the computer program instructions are executed by the processor, the electronic device executes the aforementioned method for identifying data deviations in ship structure monitoring.
[0040] Compared with the prior art, the solution provided by the present application can obtain a first data sequence collected by a sensor in real time, and determine the data statistical features corresponding to the first data sequence based on the first data sequence and the statistical period, and then input the first data sequence into a pre-trained time series prediction model to obtain a predicted second data sequence, determine the residual between the second data sequence and the first data sequence, and determine the data deviation identification result based on the data statistical features and residuals corresponding to the first data sequence. It can also obtain a third data sequence collected by a verification sensor in real time, and obtain a fourth data sequence for predicting the collected data of the verification sensor based on the data sequence collected by the non-verification sensor in real time and the proxy model for predicting the response data of the ship structure, determine the residual between the third data sequence and the fourth data sequence, and determine the data deviation identification result when the residual is greater than the second threshold, thereby realizing the detection of data deviations of the collected data of the sensor and the output data of the proxy model, avoiding the occurrence of an error accumulation effect that causes an irreversible deviation in the overall estimation result of the ship structure monitoring, and determining the cause of the data deviation based on the detection result, so that the user can correct the data deviation according to the corresponding cause, thereby ensuring the performance and reliability of the ship structure monitoring process. BRIEF DESCRIPTION OF THE DRAWINGS
[0041] Other features, objects and advantages of the present application will become more apparent by reading the detailed description of non-limiting embodiments made with reference to the following drawings:
[0042] Figure 1 A flowchart of a method for identifying data deviations in ship structure monitoring is provided for some embodiments of the present application.
[0043] Figure 2 A flowchart is provided for some embodiments of the present application for inputting a first data sequence into a pre-trained time series prediction model to obtain a predicted second data sequence.
[0044] Figure 3 A flowchart of another method for identifying data deviations in ship structure monitoring is provided for some embodiments of the present application.
[0045] Figure 4 A flowchart for performing data deviation judgment on a third data sequence is provided for some embodiments of the present application. DETAILED DESCRIPTION
[0046] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. This embodiment is implemented based on the technical solution of the present invention, and provides a detailed implementation method and specific operation process, but the protection scope of the present invention is not limited to the following embodiments.
[0047] Here, the method for identifying data deviation in ship structure monitoring of an embodiment of the present application is suitable for scenarios in which data deviation detection is performed on sensor collected data and output data of a proxy model for predicting ship structure response data in ship structure monitoring.
[0048] In this scenario, ship structure monitoring first obtains structural response data such as stress data, strain data, and motion response data such as hull attitude data collected by sensors installed at multiple locations on the ship, and then inputs the collected response data into an agent model using artificial intelligence technology. The agent model outputs the predicted response data at various locations on the entire ship, and uses the predicted response data to calculate the fatigue load and remaining healthy life of the ship structure.
[0049] Since sensors are high-precision electronic devices, they are prone to electrical failures and functional failures in the high-salt, high-temperature and high-humidity marine environment, resulting in data errors and distortions, which in turn lead to false alarms or missed alarms in the entire ship structure monitoring system. Similarly, the proxy model is an engineering model established through training based on a finite sample data set. Under certain extreme working conditions, it is very likely that the predicted response data of the ship structure will differ greatly from the actual response data.
[0050] The numerical calculation and judgment of the health status of ship structure in ship structure monitoring is the cumulative statistics of ship response data based on time series. If the above two major data deviations are not discovered in time, there will be an error accumulation effect, resulting in irreversible deviations in the overall estimation results of ship structure monitoring.
[0051] The method for identifying data deviation in ship structure monitoring provided in an embodiment of the present application can obtain a first data sequence collected by a sensor in real time, and determine the data statistical features corresponding to the first data sequence based on the first data sequence and the statistical period, then input the first data sequence into a pre-trained time series prediction model to obtain a predicted second data sequence, determine the residual between the second data sequence and the first data sequence, and determine the data deviation identification result based on the data statistical features and the residual corresponding to the first data sequence, and can also obtain a third data sequence collected by a verification sensor in real time, and obtain a fourth data sequence for predicting the collected data of the verification sensor based on the data sequence collected by the non-verification sensor in real time and a proxy model for predicting the response data of the ship structure, determine the residual between the third data sequence and the fourth data sequence, and determine the data deviation identification result when the residual is greater than a second threshold, thereby realizing the detection of data deviation of the collected data of the sensor and the output data of the proxy model, avoiding the occurrence of an error accumulation effect that causes an irreversible deviation in the overall estimation result of the ship structure monitoring, and can determine the cause of the data deviation based on the detection result, so that the user can correct the data deviation according to the corresponding cause, thereby ensuring the performance and reliability of the ship structure monitoring process.
[0052] In some embodiments of the present application, the electronic device executes the process of the method for identifying data deviation in ship structure monitoring, such as Figure 1 As shown, the method may include the following steps:
[0053] Step S101, acquiring a first data sequence collected by a sensor in real time.
[0054] It will be appreciated that electronic devices may include, but are not limited to, laptop computers, desktop computers, tablet computers, mobile phones, wearable devices, head-mounted displays, servers, mobile email devices, portable game consoles, portable music players, reader devices, televisions having one or more processors embedded or coupled thereto, or other electronic devices capable of accessing a network, etc.
[0055] Here, the ship may include but is not limited to various types of marine transport or engineering ships, oil, gas, and mineral mining and storage floating devices, as well as floating wind power, photovoltaic, hydrogen and ammonia production platforms, etc. As long as it is a device that can operate at sea, the embodiments of the present application do not impose specific restrictions on this.
[0056] The sensors are arranged at multiple positions on the ship to collect response data of the ship under wave conditions, and the response data is used to describe the stress, strain, motion state change, etc. of the ship structure position caused by the wave conditions. The sensors continuously collect the response data of the arranged positions in real time to form a first data sequence, which includes multiple response data at the arranged positions sorted by time.
[0057] Step S102: determining data statistical features corresponding to the first data sequence according to the first data sequence and a preset statistical period.
[0058] The statistical period is used to describe the length of time for calculating statistical features, for example, it can be 12 hours, 24 hours, 48 hours, etc. In some embodiments of the present application, the statistical period is set to 24 hours, that is, the data statistical features are calculated once a day, so that the calculation is convenient.
[0059] Here, the data statistical features may include but are not limited to: mean, standard deviation, skewness, kurtosis, etc. The data statistical features corresponding to the first data sequence are used to describe the relevant statistical characteristics of the first data sequence within a certain time range.
[0060] For example, the first data sequence collected by the sensor is y1, y2, y3, ...y i , then the data statistical features corresponding to the first data sequence can be {μ, σ, γ1, γ2}, where the mean can be expressed by the formula as follows: The standard deviation can be expressed by the formula as follows: The skewness can be expressed by the formula as follows: Kurtosis can be expressed by the formula as follows:
[0061] Step S103: input the first data sequence into a pre-trained time series prediction model to obtain a predicted second data sequence.
[0062] Here, the time series prediction model is a pre-trained model used to predict the current data sequence based on the historical data sequence.
[0063] In some embodiments of the present application, the time series prediction model may be an autoregressive integrated moving average model. The autoregressive integrated moving average model (ARIMA) is a time series model, which is a mathematical model obtained by curve fitting and parameter estimation based on observed time series data.
[0064] The basic idea of the ARIMA model is to use the historical information of the data itself to predict the future. The label value at a point in time is affected by both the label value in the past period of time and the accidental events in the past period of time. In other words, the ARIMA model assumes that the label value fluctuates around the general trend of time, where the trend is affected by historical labels, and the fluctuation is affected by accidental events in a period of time, and the general trend itself is not necessarily stable. In short, the ARIMA model attempts to extract the time series patterns hidden behind the data through the autocorrelation and difference of the data, and then use these patterns to predict future data.
[0065] The ARIMA model can be expressed by the formula as follows:
[0066] ARIMA(p, d, q):
[0067] Among them, p, d, and q are three hyperparameters of the model. p is the order of the autoregressive (AR) term, which is used to indicate the number of previous historical values related to the current value of Y. d is the minimum difference order that makes the data sequence stable, which is usually 1, 2, or 3. For periodic functions such as sensors, it is generally taken as order 1. q is the order of the moving average (MA) term, which is used to indicate the number of autoregressive prediction errors of previous historical values related to the current value of Y. and θ are non-zero unknown coefficients, ε t is an independent error term, Y t is a time series that has been processed to be stationary.
[0068] Figure 2 The process of inputting a first data sequence into a pre-trained time series prediction model to obtain a predicted second data sequence in some embodiments of the present application is shown, such as Figure 2 As shown, the following steps may be specifically included:
[0069] Step S201, determining whether the first data sequence is a stationary random sequence.
[0070] Here, the first data sequence is subjected to differential processing, and whether the first data sequence is a stationary random sequence is determined based on the result of the differential processing. If it is a stationary random sequence, subsequent processing is performed; if it is not a stationary random sequence, the first data sequence needs to be stabilized until the processed first data sequence is a stationary random sequence.
[0071] The first data sequence collected by the sensor can approximately fit a sine function with a period of t. In some embodiments of the present application, the first data sequence can be subjected to first-order difference processing: Y i =yi -y i+t (d = 1), and then use the stationary sequence test method to test Y i Conduct inspection.
[0072] In some embodiments of the present application, the KPSS (Kwiatkowski Phillips Schmidt Shin Test) test method can be used to determine the Y i Is it a stationary sequence? If it is a stationary sequence, enter step S202 for processing. If it is not a stationary sequence, it means that there is too much noise in the first data sequence, and the first data sequence needs to be cleaned or further processed by filtering and denoising methods, until the processed first data sequence is verified to be a stationary random sequence, and then enter step S202 for processing.
[0073] KPSS (Kwiatkowski Phillips Schmidt Shin Test) is a method used to test whether a time series is stationary. It is one of the hypothesis testing methods. The null hypothesis of the KPSS test is that the time series is stationary. If the test result rejects the null hypothesis, it indicates that the time series is non-stationary. The KPSS test makes a judgment by comparing the value of the test statistic with the critical value corresponding to the significance level. If the value of the test statistic is greater than the critical value, the null hypothesis is considered to be invalid, that is, the time series is non-stationary.
[0074] Step S202: Determine the parameter values corresponding to the autoregressive integrated moving average model according to the calculation results of the autocorrelation function and the partial autocorrelation function.
[0075] Here, for the stationary first data sequence {Y i}, calculate the autocorrelation function (ACF) and partial autocorrelation function (PACF), and determine the parameters p and q of the ARIMA model.
[0076] Autocorrelation function ρ k It can be expressed as follows:
[0077]
[0078] Wherein, μ is the first data sequence {Y i}, N is the length of the first data sequence.
[0079] Partial autocorrelation function φ k It can be expressed as follows:
[0080]
[0081] Among them, Cov() is the covariance and Var() is the conditional variance.
[0082] In some embodiments of the present application, when the calculation result of the autocorrelation function is close to zero, the current value of the first parameter can be determined as the value of the order of the autoregressive term in the parameter corresponding to the autoregressive integrated moving average model. Here, the first parameter is the autocorrelation function ρ k The parameter k in the autocorrelation function ρ k In the calculation process, as the first parameter k increases, the autocorrelation function ρ k The result value decreases rapidly. When k increases to a certain value, the autocorrelation function ρ k The resulting value decreases to close to zero, at which point the current value of the first parameter k is used as the value of the autoregressive order parameter p of the ARIMA model.
[0083] In some embodiments of the present application, when the calculation result of the partial autocorrelation function is close to zero, the current value of the second parameter can be determined as the value of the sliding average term order in the parameter corresponding to the autoregressive integrated sliding average model. Here, the second parameter is the partial autocorrelation function φ k The parameter k in the partial autocorrelation function φ k In the calculation process, as the second parameter k increases, the partial autocorrelation function φ k The result value decreases rapidly. When k increases to a certain value, the partial autocorrelation function φ k The resulting value decreases to close to zero, at which point the current value of the second parameter k is used as the value of the sliding average order parameter q of the ARIMA model.
[0084] In addition, it is understood that the calculation of the autocorrelation function and the partial autocorrelation function can be implemented using different software function packages written in different programming languages, and the present application embodiment does not specifically limit this. For example, a function tool written in Python language can be used to calculate the autocorrelation function and the partial autocorrelation function.
[0085] Step S203: training the autoregressive integrated moving average model according to the parameters corresponding to the autoregressive integrated moving average model.
[0086] After determining the parameters p, d, and q corresponding to the autoregressive integrated moving average model, the autoregressive integrated moving average model is trained according to these parameters and the first data sequence to obtain a trained autoregressive integrated moving average model. The trained autoregressive integrated moving average model can output a predicted corresponding data sequence based on the input data sequence.
[0087] It is understood that the training of the autoregressive integrated moving average model can be implemented using different software function packages written in different programming languages, and the embodiments of the present application do not impose specific restrictions on this. For example, the data sequence collected by the sensor can be trained through the function library of the Python language, and the specific training instruction is model = ARIMA (y, order = (p, d, q)).
[0088] Step S204, obtaining a predicted second data sequence according to the trained autoregressive integrated moving average model.
[0089] Here, the first data sequence is predicted by the trained autoregressive integrated moving average model, and the obtained data sequence as the second data sequence.
[0090] Step S104, determining the residual between the second data sequence and the first data sequence.
[0091] Get the predicted second data sequence Then, the residual between the second data sequence and the first data sequence is calculated. The residual r i It can be expressed as follows:
[0092]
[0093] Step S105, determining a data deviation identification result according to the data statistical characteristics and residuals corresponding to the first data sequence.
[0094] Here, whether the data statistical characteristics corresponding to the first data sequence are normal can be judged by comparing with the historical data statistical characteristics of the sensor. The statistical period of the historical data statistical characteristics is the same as the statistical period of the data statistical characteristics corresponding to the first data sequence, for example, both are 24 hours. The historical data is the data sequence collected by the sensor in the past period of time.
[0095] In addition, the data deviation identification results may include but are not limited to: sensor working normally, sensor signal drift, sensor system failure, etc., and the embodiments of the present application do not impose specific restrictions on this.
[0096] In some embodiments of the present application, when the data statistical features corresponding to the first data sequence deviate from the historical data statistical features and the residual is less than a preset first threshold, the data deviation identification result can be determined to be a sensor signal drift.
[0097] It is understood that the first threshold can be set according to the specific situation of the sensor collecting data, and can be used to define the maximum range of the residual, and the embodiment of the present application does not impose specific restrictions on this. For example, the size of the first threshold can be set to three times the standard deviation of the data statistical feature.
[0098] Here, the data statistical characteristics of the first data sequence are abnormal, while the residual between the first data sequence and the predicted second data sequence is normal, indicating that there is likely signal drift in the sensor hardware. The signal drift problem can be solved by resetting the zero position of the sensor hardware.
[0099] In some embodiments of the present application, when the data statistical features corresponding to the first data sequence deviate from the historical data statistical features and the residual is greater than a preset first threshold, the data deviation identification result can be determined to be a sensor system failure.
[0100] Here, the data statistical characteristics of the first data sequence are abnormal, and the residual between the first data sequence and the predicted second data sequence is also abnormal, indicating that there is a high probability that the sensor has a systemic failure, which can be solved by replacing the sensor.
[0101] In addition, if the data statistical characteristics of the first data sequence are normal, it means that the sensor is in a normal working state and there is no problem of data deviation.
[0102] In some embodiments of the present application, another process of a method for identifying data deviations in ship structure monitoring by executing electronic equipment is provided, which is used to detect data deviations of a proxy model used in ship structure monitoring, such as Figure 3 As shown, the method may include the following steps:
[0103] Step S301, obtaining a third data sequence collected in real time by a verification sensor.
[0104] Here, since the ship is a large and complex rigid structure, the use of standard finite element method to analyze the structural response data will result in a long calculation time due to the huge solution space, which is difficult to meet the real-time monitoring requirements of the ship structure. Therefore, the current ship structure monitoring technology uses machine learning technology to train an agent model. The agent model can predict the response data of all structural positions of the entire ship based on the response data collected by sensors set at a certain number of structural positions in the ship. For example, sensors can be set at 10 to 30 structural positions in a ship, and the response data collected by these sensors can be used to use the agent model to predict the response data of all structural positions of the entire ship.
[0105] Here, a sensor is set at a structural position of the ship to collect response data. The position serves as a calculation verification point for the response data, and the sensor set at the position is a calibration sensor.
[0106] In addition, multiple sensors installed at other structural positions in the ship other than the calculation verification points are non-calibration sensors, and the response data collected by the non-calibration sensors are used as input data of the proxy model.
[0107] In some embodiments of the present application, the data collected by the verification sensor and the data collected by the non-verification sensor have a high correlation. The high correlation between the two data can make the proxy model more accurate when predicting the collected data of the verification sensor based on the response data collected by the non-verification sensor. To determine the correlation between the data collected by the verification sensor and the data collected by the non-verification sensor, a sensitivity analysis can be performed to obtain the correlation result between the two data.
[0108] Therefore, multiple candidate calculation verification points can be selected in the structural position of the ship, the correlation between the response data of the candidate calculation verification points and the response data collected by the non-calibration sensor is calculated, and the candidate calculation verification point with the highest correlation is determined as the final calculation verification point.
[0109] After the calculation verification point is determined, a verification sensor is set at the point to obtain a sequence of response data collected in real time by the verification sensor, and the sequence of response data is the third data sequence.
[0110] Step S302: acquiring a fourth data sequence for predicting the data collected by the calibration sensor according to the data sequence collected by the non-calibration sensor in real time and a pre-established proxy model for predicting the ship structure response data.
[0111] Here, the proxy model is a pre-trained model for predicting the response data of ship structures. It can predict the response data corresponding to the structural position in the ship based on the data sequence collected in real time by multiple non-calibration sensors, for example, predict the response data corresponding to the calculation verification point.
[0112] The proxy models used may include but are not limited to: Kriging model, Artificial Neural Network (ANN), Radial Basis Functions (RBF), Support Vector Machine (SVM), Polynomial Regression (PRG), etc.
[0113] In some embodiments of the present application, a data sequence collected in real time by multiple non-calibration sensors can be used as input data of a proxy model. The proxy model receives the input data and outputs prediction results of response data of multiple structural positions in the ship. The predicted response data corresponding to the calculation verification points are then extracted from the prediction results of the proxy model. The predicted response data is the predicted collection data of the calibration sensor, that is, the fourth data sequence.
[0114] Step S303, determining the residual between the third data sequence and the fourth data sequence.
[0115] Here, the residual between the third data sequence and the fourth data sequence is calculated, and reference may be made to the process of calculating the residual between the second data sequence and the first data sequence in step S104, which will not be described in detail here.
[0116] Step S304: when the residual is greater than a preset second threshold, determine the data deviation identification result.
[0117] It can be understood that the second threshold can be set according to the allowable error range of the proxy model and can be used to define the maximum error range of the proxy model. The embodiment of the present application does not impose any specific limitation on this.
[0118] Here, the residual is greater than the preset second threshold, indicating that there is a large error between the predicted response data of the verification sensor output by the proxy model and the response data actually collected by the verification sensor, and there is a data deviation problem, and the cause of the problem needs to be further determined.
[0119] In addition, if the residual is smaller than the preset second threshold, it means that the third data sequence and the fourth data sequence are within a reasonable error range and there is no data deviation problem.
[0120] It can be understood that if the residual is greater than the preset second threshold, it indicates that there is a data deviation, and the data deviation identification result may include but is not limited to sensor failure, poor proxy model fit, etc.
[0121] In some embodiments of the present application, to determine the data deviation identification result, the data deviation judgment can be first performed on the third data sequence, and when there is a data deviation in the third data sequence, the data deviation identification result is determined as a sensor failure, otherwise the data deviation identification result is determined as a poor fit of the proxy model.
[0122] Here, the third data sequence is the response data collected by the verification sensor in real time. The verification sensor may also cause the third data sequence collected in real time to be abnormal due to related faults, which in turn causes the residual abnormality between the third data sequence and the fourth data sequence predicted by the proxy model. Therefore, by determining whether there is data deviation in the third data sequence, it can be determined whether the residual abnormality is caused by sensor failure or poor fitting of the proxy model.
[0123] Figure 4 The process of determining the data deviation of the third data sequence in some embodiments of the present application is shown as follows: Figure 4 As shown, the following steps may be specifically included:
[0124] Step S401, according to the third data sequence and the preset statistical period, determine the data statistical features corresponding to the third data sequence. The data statistical features may include but are not limited to: mean, standard deviation, skewness and kurtosis. The data statistical features corresponding to the third data sequence are used to describe the relevant statistical characteristics of the third data sequence within a certain time range.
[0125] The data statistical features corresponding to the third data sequence may be determined by referring to the specific process of determining the data statistical features corresponding to the first data sequence in step S102, which will not be described in detail here.
[0126] Step S402: input the third data sequence into a pre-trained time series prediction model to obtain a predicted fifth data sequence.
[0127] Similarly, the predicted fifth data sequence can be obtained by referring to the specific process of inputting the first data sequence into the pre-trained time series prediction model in step S103 to obtain the predicted second data sequence, which will not be repeated here.
[0128] Step S403, determining the residual between the third data sequence and the fifth data sequence.
[0129] Similarly, the residual between the third data sequence and the fifth data sequence may be determined by referring to the specific process of determining the residual between the second data sequence and the first data sequence in step S104, which will not be described in detail here.
[0130] Step S404: determine the data deviation judgment result according to the data statistical characteristics and residuals corresponding to the third data sequence.
[0131] Similarly, the data deviation judgment result can be determined by referring to the specific process of determining the data deviation identification result according to the data statistical characteristics and residuals corresponding to the first data sequence in step S105, which will not be repeated here.
[0132] Here, the data deviation judgment result may include but is not limited to: the sensor is working normally, the sensor signal drifts, the sensor system fails, etc., and the embodiment of the present application does not impose specific restrictions on this.
[0133] Some embodiments of the present application further provide a computer-readable medium having computer-readable instructions stored thereon, wherein the computer-readable instructions can be executed by a processor to implement the aforementioned method for identifying data deviations in ship structure monitoring.
[0134] Some embodiments of the present application also provide an electronic device, which includes a memory for storing computer program instructions and a processor for executing the computer program instructions, wherein when the computer program instructions are executed by the processor, the electronic device executes the aforementioned method for identifying data deviations in ship structure monitoring.
[0135] In summary, the solution provided by the present application can obtain the first data sequence collected by the sensor in real time, and determine the data statistical features corresponding to the first data sequence based on the first data sequence and the statistical period, and then input the first data sequence into the pre-trained time series prediction model to obtain the predicted second data sequence, determine the residual between the second data sequence and the first data sequence, and determine the data deviation identification result based on the data statistical features and residuals corresponding to the first data sequence. It can also obtain the third data sequence collected by the verification sensor in real time, and obtain the fourth data sequence for predicting the collected data of the verification sensor based on the data sequence collected by the non-verification sensor in real time and the proxy model for predicting the response data of the ship structure, determine the residual between the third data sequence and the fourth data sequence, and determine the data deviation identification result when the residual is greater than the second threshold value, so as to realize the detection of data deviation of the collected data of the sensor and the output data of the proxy model, avoid the occurrence of error accumulation effect leading to irreversible deviation in the overall estimation result of ship structure monitoring, and determine the cause of the data deviation based on the detection result, so that the user can correct the data deviation according to the corresponding cause, thereby ensuring the performance and reliability of the ship structure monitoring process.
[0136] It should be noted that the present application can be implemented in software and / or a combination of software and hardware, for example, can be implemented using an application specific integrated circuit (ASIC), a general purpose computer or any other similar hardware device. In one embodiment, the software program of the present application can be executed by a processor to implement the steps or functions described above. Similarly, the software program of the present application (including relevant data structures) can be stored in a computer-readable recording medium, for example, a RAM memory, a magnetic or optical drive or a floppy disk and similar devices. In addition, some steps or functions of the present application can be implemented using hardware, for example, as a circuit that cooperates with a processor to perform each step or function.
[0137] In a typical configuration of the present application, the terminal and the network device each include one or more processors (CPU), input / output interface, network interface and memory.
[0138] The memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.
[0139] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, read-only compact disk (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include non-transitory media such as modulated data signals and carrier waves.
[0140] In addition, a part of the present application may be applied as a computer program product, such as a computer program instruction, which, when executed by a computer, can call or provide the method and / or technical solution according to the present application through the operation of the computer. The program instruction for calling the method of the present application may be stored in a fixed or removable recording medium, and / or transmitted through a data stream in a broadcast or other signal-bearing medium, and / or stored in a working memory of a computer device that runs according to the program instruction. Here, according to an embodiment of the present application, a device is included, the device including a memory for storing computer program instructions and a processor for executing program instructions, wherein, when the computer program instruction is executed by the processor, the device is triggered to run the method and / or technical solution based on the aforementioned multiple embodiments according to the present application.
[0141] It is obvious to those skilled in the art that the present application is not limited to the details of the above exemplary embodiments, and that the present application can be implemented in other specific forms without departing from the spirit or basic features of the present application. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-restrictive, and the scope of the present application is limited by the attached claims rather than the above description, so it is intended to include all changes that fall within the meaning and scope of the equivalent elements of the claims in the present application. Any figure mark in the claims should not be regarded as limiting the claims involved. In addition, it is obvious that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or devices stated in the device claim can also be implemented by one unit or device through software or hardware. The words first, second, etc. are used to indicate names, and do not indicate any particular order.
Claims
1. A method for identifying data deviation in ship structure monitoring, characterized in that: The method includes: Acquire a first data sequence collected by a sensor in real time; Determine, according to the first data sequence and a preset statistical period, a data statistical feature corresponding to the first data sequence, wherein the data statistical feature includes one or more combinations of the following: mean, standard deviation, skewness, and kurtosis; Inputting the first data sequence into a pre-trained time series prediction model to obtain a predicted second data sequence, wherein the time series prediction model is used to predict a current data sequence based on a historical data sequence; determining a residual between the second data sequence and the first data sequence; A data deviation identification result is determined based on the data statistical characteristics corresponding to the first data sequence and the residual.
2. The method according to claim 1, characterized in that The time series prediction model is an autoregressive integrated moving average model.
3. The method according to claim 2, characterized in that Inputting the first data sequence into a pre-trained time series prediction model to obtain a predicted second data sequence includes: Determining that the first data sequence is a stationary random sequence; According to the calculation results of the autocorrelation function and the partial autocorrelation function, the parameter values corresponding to the autoregressive integrated moving average model are determined; Training the autoregressive integrated moving average model according to the parameters corresponding to the autoregressive integrated moving average model; According to the trained autoregressive integrated moving average model, a predicted second data sequence is obtained.
4. The method according to claim 3, characterized in that: According to the calculation results of the autocorrelation function and the partial autocorrelation function, the parameter values corresponding to the autoregressive integrated moving average model are determined, including: When the calculation result of the autocorrelation function is close to zero, determining the current value of the first parameter as the value of the order of the autoregressive term in the parameter corresponding to the autoregressive integrated moving average model; or When the calculation result of the partial autocorrelation function is close to zero, the current value of the second parameter is determined as the value of the order of the sliding average term in the parameter corresponding to the autoregressive integrated sliding average model.
5. The method according to claim 1, characterized in that Determining a data deviation identification result according to the data statistical features corresponding to the first data sequence and the residual, including: When the data statistical feature corresponding to the first data sequence deviates from the historical data statistical feature and the residual is less than a preset first threshold, determining that the data deviation identification result is a sensor signal drift; or When the data statistical feature corresponding to the first data sequence deviates from the historical data statistical feature and the residual is greater than a preset first threshold, it is determined that the data deviation identification result is a sensor system failure.
6. A method for identifying data deviation in ship structure monitoring, characterized in that: The method includes: Acquire a third data sequence collected in real time by the verification sensor, wherein the data collected by the verification sensor has a high correlation with the data collected by the non-verification sensor; According to the data sequence collected in real time by the non-calibration sensor and the pre-established proxy model for predicting the ship structure response data, a fourth data sequence for predicting the collected data of the calibration sensor is obtained; determining a residual between the third data sequence and the fourth data sequence; When the residual is greater than a preset second threshold, a data deviation identification result is determined.
7. The method according to claim 6, characterized in that Determine the data deviation identification results, including: Performing data deviation judgment on the third data sequence; In the case where there is a data deviation in the third data sequence, the data deviation identification result is determined as a sensor failure; otherwise, the data deviation identification result is determined as a poor fit of the proxy model.
8. The method according to claim 7, characterized in that Performing data deviation judgment on the third data sequence includes: Determine, according to the third data sequence and a preset statistical period, a data statistical feature corresponding to the third data sequence, wherein the data statistical feature includes one or more combinations of the following: mean, standard deviation, skewness, and kurtosis; Inputting the third data sequence into a pre-trained time series prediction model to obtain a predicted fifth data sequence; determining a residual between the third data sequence and the fifth data sequence; A data deviation judgment result is determined based on the data statistical characteristics corresponding to the third data sequence and the residual.
9. A computer-readable medium having computer-readable instructions stored thereon, wherein the computer-readable instructions can be executed by a processor to implement the method for identifying data deviations in ship structure monitoring according to any one of claims 1 to 5 or 6 to 8.
10. An electronic device comprising a memory for storing computer program instructions and a processor for executing the computer program instructions, wherein: When the computer program instructions are executed by the processor, the electronic device is caused to perform the method for identifying data deviations in ship structure monitoring according to any one of claims 1 to 5 or 6 to 8.