Diesel engine safety state evaluation method based on parameter prediction

Abstract

This invention relates to a method for assessing the safety status of diesel engines based on parameter prediction, and pertains to the field of diesel engine safety protection. The invention aims to address the current inability to predict and assess the safety status of diesel engines. The invention includes: acquiring diesel engine safety protection thermal parameters and assembling an initial data set S. * For the initial data set S * Preprocessing is performed to obtain S; long-term prediction is performed on S to obtain long-term prediction data of diesel engine operating parameters S0. l Short-term forecasting of S yields short-term forecast data for diesel engine operating parameters. s ; For S, S l S s Normalization was performed separately to obtain S X S Xl S Xs ; Using S X S Xl S Xs This invention obtains the diesel engine safety status prediction assessment conclusion P1 and the remaining duration of the diesel engine safety status operation prediction P2. It is used to assess the safety status of a diesel engine.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of diesel engine safety protection, and particularly relates to a diesel engine safety state evaluation method based on parameter prediction. BACKGROUND

[0002] The diesel engine is the power equipment of most ships, and the safe running state of the diesel engine is crucial to the navigation of the ship. At present, the safety protection measures for the running of the diesel engine are mainly based on the monitored diesel engine running data. When it is judged that a fault occurs or the monitored parameters appear obvious abnormalities, emergency safety protection measures are taken. However, for special situations, such as when the naval vessels perform tasks, the safety running protection measures of the diesel engine will interrupt the original decision, and the sudden increase and decrease of the load state will also cause great damage to the diesel engine, and even cause more serious consequences. Therefore, before triggering the safety running protection measures of the diesel engine, the safety running state of the main engine and its nearby system is predicted and evaluated, the passive disposal is changed into active prevention, early warning and appropriate measures are taken, which can effectively reduce or avoid the loss.

[0003] The prediction and evaluation of the safety state of the diesel engine is an important part of the safety protection of the diesel engine, and the accuracy of the prediction and the authenticity of the evaluation are of great significance to the operation and maintenance decision of the diesel engine. According to the monitored diesel engine running parameters, the safety state of the diesel engine is predicted and evaluated by predicting the change trend and the deterioration state of each parameter. However, due to the complex structure of the diesel engine and the strong nonlinearity of the running parameters, it is currently impossible to accurately predict the long-term running thermal parameters of the diesel engine with multiple states and multiple trends, so as to comprehensively predict the information and evaluate the safety running state of the diesel engine. SUMMARY

[0004] The present application aims to solve the problem that the safety state of the diesel engine cannot be predicted and evaluated at present, and proposes a diesel engine safety state evaluation method based on parameter prediction.

[0005] The diesel engine safety state evaluation method based on parameter prediction specifically includes the following steps:

[0006] Step 1, obtaining the safety protection thermal parameters of the diesel engine and forming an initial data set

[0007] Among them, is a time series data sequence of the main engine lubricating oil inlet pressure, is a time series data sequence of the crankcase pressure, is a time series data sequence of the main engine inlet air pressure, is a time series data sequence of the exhaust manifold temperature, is a time series data sequence of the main engine fresh water inlet temperature, is a time series data sequence of the main engine lubricating oil inlet temperature, is a time series data sequence of the main engine inlet air temperature, is a turbocharger speed time series data sequence;

[0008] Step two, using the Laiyida criterion to preprocess the initial data set, obtaining the preprocessed initial data set S;

[0009] Step three, long-term prediction of S, obtaining long-term prediction data S of diesel engine operating parameters l ;

[0010] Step four, short-term prediction of S, obtaining short-term prediction data S of diesel engine operating parameters s ;

[0011] Step five, normalize S, S l , S s respectively, and obtain S X , S Xl , S Xs ;

[0012] Step six, using S X , S Xl , S Xs obtained in step five to obtain diesel engine safety state prediction evaluation conclusion P1:

[0013] First, using S X , S Xl , S Xs to obtain the values of parameter long-term prediction attribute containing features F l , short-term prediction attribute containing features F s , monitoring parameter attribute containing features F f , threshold value frequency domain attribute containing features F t , threshold value time domain attribute containing features F tr ; Then, using F l , F s , F f , F t , F tr to obtain diesel engine safety state prediction evaluation conclusion P1;

[0014] Wherein, the greater P1, the more dangerous the state of the diesel engine;

[0015] Step seven, using the threshold value time domain attribute containing features F tr obtained in step six to obtain the remaining time P2 of diesel engine safety state operation prediction.

[0016] Further, in the step two, the initial data set is preprocessed by using the Laiyida criterion to obtain the preprocessed initial data set S, specifically:

[0017] Step two one, divide each parameter time series data sequence in the initial data set into n equal intervals, and obtain the average value of each interval

[0018] Step two two, use the average value of each interval time series data Obtain the residual error v of each interval time series data i And the standard deviation σ:

[0019]

[0020]

[0021]

[0022] Wherein, m is the total number of data points in each interval time series data, M is the total number of data in each parameter time series data sequence, Is the ith time series data in a certain interval time series data segment;

[0023] Step two three, use v i And σ to obtain the outliers in each interval time series data, and remove the outliers, then fill in the position after removing the outliers, obtain the pretreated initial data set S = [S1 S2 S3 S4 S5 S6 S7 S8] T :

[0024] First, judge Whether to meet If so, it is an outlier;

[0025] Then, remove the outliers The position of removing the outliers is empty;

[0026] Then, Newton interpolation is performed on the null position to replace the null data, and the pretreated initial data set S = [S1 S2 S3 S4 S5 S6 S7 S8] T ;

[0027] Wherein, S1 is the pretreated host oil inlet pressure time series data sequence, S2 is the pretreated crankcase pressure time series data sequence, S3 is the pretreated host inlet air pressure time series data sequence, S4 is the pretreated exhaust manifold temperature time series data sequence, S5 is the pretreated host fresh water inlet temperature time series data sequence, S6 is the pretreated host oil inlet temperature time series data sequence, S7 is the pretreated host inlet air temperature time series data sequence, and S8 is the pretreated supercharger speed time series data sequence.

[0028] ​Furthermore, in step three, long-term prediction of S is performed to obtain long-term prediction data S of the diesel engine operating parameters. l This includes the following steps:

[0029] Step 3: 1. Use the unit root test to check the stationarity of each preprocessed parameter time series in S. If it passes the stationarity test, then S is considered a stationary series. f Proceed to step 32; if the stationarity test is not passed, perform differencing on S until the differencing sequence S passes the stationarity test, thus obtaining a stationary sequence S. f Then proceed to step three two;

[0030] Step 3.2: Establish a long-term prediction model, and use the stationary series S f The data is input into the long-term parameter prediction model to obtain the long-term prediction data S of the diesel engine operating parameters. l ;

[0031] The long-term prediction model for the parameters is as follows:

[0032]

[0033] Among them, sf t Let ε be the parameter value at time t in the stationary sequence. t ,...,ε t-q It is white noise; These are the autoregressive coefficients, i = 1, 2, ..., p, ψ j q are the moving average coefficients, j = 1, 2, ..., q; p and q are the orders of the autoregressive model and the moving average model, respectively.

[0034] Furthermore, in step four, a short-term prediction of S is performed to obtain short-term prediction data S of the diesel engine operating parameters. s This includes the following steps:

[0035] Step 41: Perform an accumulation operation on S to obtain...

[0036]

[0037]

[0038] Where γ∈[1,8], s γ (i') is S γ The i'th element in yes The k-th element in It is for S γ The result of summing the elements in the middle row;

[0039] Step four two, using the S obtained in step four one (1) Construct whitening sequence

[0040]

[0041]

[0042] ω = (B T B) -1 B T Y

[0043]

[0044] Wherein, ω is the development coefficient, is the kth element in the middle matrix B, Y is the intermediate matrix;

[0045] Step four three, using the S obtained in step four one (1) and W obtained in step four two (1) Establish GM (1, 1) gray prediction model, as follows:

[0046]

[0047]

[0048] Wherein, α is the gray action amount, is the sequence change prediction value, is the parameter sequence prediction value;

[0049] Step four four, input S into the GM (1, 1) gray prediction model, diesel engine running parameter short prediction data S s .

[0050] Further, the step five in the article S, S l , S s respectively, S X , S Xl , S Xs , specific for:

[0051] S, S l , S s respectively, S X = [S X1 S X2 S X3 S X4 S X5 S X6 S X7 S X8 ]T , Xl = [S Xl1 , Xl2 S Xl3 , Xl4 S Xl5 , Xl6 S Xl7 , Xl8 ] T , Xs = [S Xs1 , Xs2 S Xs3 , Xs4 S Xs5 , Xs6 S Xs7 , Xs8 ] T ;

[0052] wherein S X is the normalized S, S Xl is the normalized S l , S Xs is the normalized S s ;

[0053] The sequence X is normalized according to the formula to obtain the normalized sequence as follows:

[0054]

[0055] wherein k = 1, 2, …, M, is the kth element of the normalized sequence , x min is the minimum value in the sequence X, x max is the maximum value in the sequence X, X = [x1, x2, x3, …, x M ], X is S, S l or S s , and x k is the kth element in X.

[0056] Further, the S X , S Xl , S Xs obtained in step five is used in step six to obtain a diesel engine safety state prediction evaluation conclusion P1, including the following steps:

[0057] Step six one, establish parameter long-term prediction attribute containing features F l = F1 = [f kl , f tkl , f frl ], short-term prediction attribute containing features F s = F2 = [fks ,f tks ,f frs ], monitoring parameter attribute contains feature F f =F3=[f kf ,f tkf ,f frf ], threshold value frequency domain attribute contains feature F t =F4=[f tl ,f ts ,f tf ], threshold value time domain attribute contains feature F tr =F5=[t rtl ,t rts ,t rtf ];

[0058] Wherein, f kl is the average line slope feature in long prediction feature, f tkl is the tangent slope feature in long prediction feature, f frl is the difference-root mean square feature in long prediction feature, f ks is the average line slope feature in short-term prediction data feature, f tks is the tangent slope feature in short-term prediction data feature, f frs is the difference-root mean square feature in short-term prediction data feature, f kf is the average line slope feature in parameter monitoring feature, f tkf is the tangent slope feature in parameter monitoring feature, f frf is the difference-root mean square feature in parameter monitoring feature, f tl is the parameter long prediction feature in threshold value frequency domain feature, f ts is the short prediction feature in threshold value frequency domain feature, f tf is the monitoring data feature in threshold value frequency domain feature, t rtl is the parameter long prediction threshold value fitting time feature in threshold value time domain feature, t rts is the parameter short prediction threshold value fitting time feature in threshold value time domain feature, t rtf is the monitoring data threshold value fitting time feature in threshold value time domain feature;

[0059] Step six two, for diesel engine safety protection thermal parameter input parameter and the feature set by step six one relative importance value, and build judgment matrix A*, A 1 , A 2 , A 3 , A 4 , A 5 , A:

[0060] Relative importance value is an integer between 1-9;

[0061]

[0062]

[0063] wherein, is a turbocharger speed feature relative importance value, is a main engine inlet air temperature feature relative importance value, is a main engine lube oil inlet temperature feature relative importance value, is a main engine fresh water inlet temperature feature relative importance value, is an exhaust gas manifold temperature feature relative importance value, is a main engine inlet air pressure feature relative importance value, is a crankcase pressure feature relative importance value, is a main engine lube oil inlet pressure feature relative importance value;

[0064] wherein

[0065] wherein, is a long term prediction feature mean line slope feature relative importance value, is a long term prediction feature tangent line slope feature relative importance value, is a long term prediction feature difference-standard deviation feature relative importance value;

[0066] wherein

[0067] wherein, is a short term prediction feature mean line slope feature relative importance value, is a short term prediction feature tangent line slope feature relative importance value, is a short term prediction feature difference-standard deviation feature relative importance value;

[0068] wherein

[0069] wherein, is a parameter monitoring feature mean line slope feature relative importance value, is a parameter monitoring feature tangent line slope feature relative importance value, is a parameter monitoring feature difference-standard deviation feature relative importance value;

[0070] wherein

[0071] wherein, is a relative importance value of the long-term predicted frequency domain feature in the threshold frequency domain feature, is a relative importance value of the short-term predicted frequency domain feature in the threshold frequency domain feature, is a relative importance value of the monitoring parameter frequency domain feature in the threshold frequency domain feature;

[0072] wherein

[0073] wherein, is a relative importance value of the long-term predicted time domain feature in the threshold time domain feature, is a relative importance value of the short-term predicted time domain feature in the threshold time domain feature, is a relative importance value of the monitoring parameter time domain feature in the threshold time domain feature;

[0074] wherein a1

[0075] wherein, a1, a2, a3, a4, a5 are in turn a relative importance value of a parameter long-term predicted predicted feature, a relative importance value of a short-term predicted data feature, a relative importance value of a parameter monitoring feature, a relative importance value of a threshold frequency domain feature, and a relative importance value of a threshold time domain feature;

[0076] Step six three, consistency check is performed on the judgment matrix constructed in step six two, if the judgment matrix passes the consistency check, step six four is executed, if the judgment matrix does not pass the consistency check, step six two is returned to reset the relative importance value;

[0077] The judgment matrix constructed in step six two is subjected to consistency check, specifically:

[0078] The consistency ratio CR of each judgment matrix constructed in step six two is calculated respectively, if CR<0.1, the current judgment matrix passes the consistency check, if CR≥0.1, the current judgment matrix does not pass the consistency check;

[0079]

[0080]

[0081] wherein, RI is a consistency index of the judgment matrix, υ is the number of evaluation indexes involved in the current judgment matrix, λ max is the maximum eigenvalue in the judgment matrix, CI is an intermediate variable;

[0082] Step six four, the maximum eigenvalue λ max of the judgment matrix passing the consistency check is obtained, and the standardized eigenvector C corresponding to λ max is obtained.

[0083] Where C is c*, c, c 1 c 2 c 3 c 4 or c 5 , c = [c1 c2 … c5], to c* is the standardized vector corresponding to the largest eigenvalue in the judgment matrix A*, and c is the standardized vector corresponding to the largest eigenvalue in the judgment matrix A. 1 To c 5 It is A 1 To A 5 The largest eigenvalue in the vector corresponds to the standardized vector.

[0084] Step 65: Using the weighted statistical method, utilize the standardized eigenvector C obtained in Step 64 and the S obtained in Step 5. X S Xl S Xs Obtain attribute evaluation feature F η :

[0085] Step 66: Utilize the attribute evaluation feature F obtained in Step 65. η Obtain the attribute evaluation index BF = [B1, B2, B3, B4, B5]:

[0086] Step 67: Use the BF obtained in Step 66 to obtain the diesel engine safety status prediction and assessment conclusion P1.

[0087] Furthermore, in step six-five, the weighted statistical method utilizes the standardized feature vector C obtained in step six-four and the S obtained in step five. X S Xl S Xs Obtain attribute evaluation feature F η This includes the following steps:

[0088] Where η = 1, 2, 3, 4, 5, the long-term prediction feature F is... l =F1=[f kl ,f tkl ,f frl Short-term forecast data characteristics F s =F2=[f ks ,f tks ,f frs ] Parameter monitoring characteristics F f =F3=[f kf ,f tkf ,f frf Threshold value frequency domain characteristics F t =F4=[ftl ,f ts ,f tf Threshold value time-domain characteristics F tr =F5=[t rtl ,t rts ,t rtf ];

[0089] f kf f tkf f frf f tl f ts f tf t rtl t rts t rtf Form and f kl f tkl f frl same;

[0090] in, It is S Xl The moving average slope group for each sequence. The slope of the moving average of the predicted characteristic of the main engine lubricating oil inlet pressure field is given by parameter. The slope of the average line for the long-term prediction data of the turbocharger speed parameter. It is S Xl The set of tangent slopes for each sequence. It is S Xl The difference-standard deviation combination for each sequence; It is S Xs The moving average slope group for each sequence. It is S Xs The set of tangent slopes for each sequence. It is S Xs The difference-standard deviation combination for each sequence;

[0091] Slope f of moving average k , Tangent slope f tk Difference-root mean square f fr Threshold value breach f t Threshold fit / breakout time t rt It can be obtained through the following formula:

[0092]

[0093] f tk =max(s μ (i1+1)-s μ (i1)), i1=1,2,……,M

[0094]

[0095]

[0096]

[0097] Wherein, μ takes Xγ, Xlγ or Xsγ, i1, i2 are integers, Δs μ (i2) is an intermediate variable, s rt is a parameter threshold value.

[0098] Further, the attribute evaluation features F η The attribute evaluation index BF = [B1, B2, B3, B4, B5] is obtained as follows:

[0099]

[0100] Wherein, B1 is the long prediction data attribute evaluation result, B2 is the short prediction data attribute evaluation result, B3 is the monitoring data attribute evaluation result, B4 is the threshold frequency domain attribute evaluation result, and B5 is the threshold time domain attribute evaluation result; F η is the calculation value of the ηth attribute, the weight coefficient of each feature, is the a th feature value of the ηth attribute, and a takes 1, 2, 3.

[0101] Further, the BF obtained in step six is used to obtain the diesel engine safety state prediction evaluation conclusion P1 in step seven, specifically:

[0102]

[0103] Wherein, c η is an element of c, and B η is an element in BF.

[0104] Further, the threshold time domain attribute containing feature F tr The diesel engine safety state running prediction remaining time P2 is obtained, specifically:

[0105]

[0106] Wherein, is the main engine lubricating oil inlet pressure threshold value fitting step number, is the crankcase pressure threshold value fitting step number, is the main engine inlet air pressure threshold value fitting step number, is the main engine fresh water inlet temperature threshold value fitting step number.

[0107] The beneficial effects of the present application are:

[0108] The diesel engine safety state prediction and evaluation method provided by the present application perfects the current diesel engine safety protection mechanism, can provide diesel engine safety protection parameter prediction information, thereby long-term and accurate prediction and evaluation of the diesel engine safety state is realized, diesel engine operation and maintenance decisions can be effectively supported, and the safety of diesel engine operation is improved. In the diesel engine safety protection parameter prediction, the ARMA and grey prediction methods are respectively used for long-term prediction of the operation parameter and short-term prediction of the operation parameter according to the target prediction length, appropriate parameter length is reserved, the operation parameter prediction precision and prediction accuracy are improved, meanwhile, the present application also extracts the targeted characteristics of the monitoring parameter and the prediction parameter, quantizes the parameter degradation characteristics, and makes the subsequent safety state evaluation more convenient. The present application uses the analytic hierarchy process, constructs a three-layer hierarchical calculation relationship of the diesel engine safety state prediction and evaluation, comprehensively and systematically covers the diesel engine safety protection parameter monitoring data and the prediction data characteristics, and the pointing relationship between the characteristics and the diesel engine safety state, so that the prediction result of the diesel engine safety operation state is more accurate and real. BRIEF DESCRIPTION OF DRAWINGS

[0109] Figure 1 The flowchart of the present application is shown in the figure;

[0110] Figure 2 The diesel engine abnormal operation parameter monitoring data and long and short prediction result data curve-main engine lubricating oil inlet pressure is shown in the figure;

[0111] Figure 3 The diesel engine abnormal operation parameter monitoring data and long and short prediction result data curve-crankcase pressure is shown in the figure;

[0112] Figure 4 The diesel engine abnormal operation parameter monitoring data and long and short prediction result data curve-main engine inlet air pressure is shown in the figure;

[0113] Figure 5 The diesel engine abnormal operation parameter monitoring data and long and short prediction result data curve-exhaust manifold temperature is shown in the figure;

[0114] Figure 6 The diesel engine abnormal operation parameter monitoring data and long and short prediction result data curve-main engine fresh water inlet temperature is shown in the figure;

[0115] Figure 7 The diesel engine abnormal operation parameter monitoring data and long and short prediction result data curve-main engine lubricating oil inlet temperature is shown in the figure;

[0116] Figure 8 The diesel engine abnormal operation parameter monitoring data and long and short prediction result data curve-main engine inlet air temperature is shown in the figure;

[0117] Figure 9 For diesel engine abnormal operation parameter monitoring data and long, short prediction result data curve-pressor speed;

[0118] Figure 10 For diesel engine safety state prediction and evaluation based on parameter prediction AHP hierarchical structure. DETAILED DESCRIPTION

[0119] Specific implementation one: as shown in the embodiment, the specific process of the diesel engine safety state evaluation method based on parameter prediction is as follows: Figure 1

[0120] Step one, obtain diesel engine safety protection thermal parameters, and form an initial data set

[0121] Among them, is the time series data sequence of the main engine lubricating oil inlet pressure, is the time series data sequence of the crankcase pressure, is the time series data sequence of the main engine inlet air pressure, is the time series data sequence of the exhaust manifold temperature, is the time series data sequence of the main engine fresh water inlet temperature, is the time series data sequence of the main engine lubricating oil inlet temperature, is the time series data sequence of the main engine inlet air temperature, is the time series data sequence of the supercharger speed.

[0122] In this step, the database parameters are called in time sequence through the port, and the calling length is M, and in this embodiment, M=3000.

[0123] Step two, pre-process the initial data set by using the Laplace criterion to obtain the pre-processed initial data set:

[0124] Step two one, divide each parameter time series data sequence in the initial data set into n=30 segments at equal intervals, and obtain the average value of each segment of time series data

[0125] Step two two, use the average value of each segment of time series data to obtain the residual error v of each segment of time series data i and the standard deviation σ:

[0126]

[0127]

[0128] Among them, m is the total number of data points in each segment of time series data, s i0 ​is the ith0 time series data in a time series data segment, and M is the total number of data in each parameter time series data sequence;

[0129] Step two three, using the residual error v of each time series data i and the standard deviation σ to obtain outliers, and the outliers are removed, and the positions of the removed outliers are completed to obtain the complete and smooth parameter time series data sequence S = [S1 S2 S3 S4 S5 S6 S7 S8] T :

[0130] First, if the parameter value at a certain time satisfies , it is determined that the value is an outlier, which is removed , and the time series position is left empty;

[0131] Then, integrate each time series data, replace the null value with Newton interpolation, complete the null value data, and obtain the complete and smooth parameter time series data set S = [S1 S2 S3 S4 S5 S6 S7 S8] T ;

[0132] Among them, S1 is the pre-processed main engine inlet oil pressure time series data sequence, S2 is the pre-processed crankcase pressure time series data sequence, S3 is the pre-processed main engine inlet air pressure time series data sequence, S4 is the pre-processed exhaust manifold temperature time series data sequence, S5 is the pre-processed main engine fresh water inlet temperature time series data sequence, S6 is the pre-processed main engine oil inlet temperature time series data sequence, S7 is the pre-processed main engine inlet air temperature time series data sequence, and S8 is the pre-processed supercharger speed time series data sequence.

[0133] Step three, long-term prediction of S based on ARMA method, to obtain long-term prediction data S of diesel engine operation parameters l , specifically:

[0134] Step three one, using the unit root test method (ADF) to test the stationarity of each pre-processed parameter time series data sequence in the parameter time series data set S, if the stationarity test is passed, the current S is taken as a stationary sequence S f , and step three two is executed; if the stationarity test is not passed, the S is subjected to difference processing until the sequence S after difference processing passes the stationarity test, and a stationary sequence S f is obtained, and then step three two is executed;

[0135] Step three two, establish a parameter long-term prediction model, input the stationary sequence S f into the parameter long-term prediction model, and obtain the parameter time series data sequence S l in the future time period n l = 300;

[0136] The parameter long-term prediction model is as follows:

[0137]

[0138] Wherein, sf t is the value of the parameter at time t in the stationary sequence, ε t ,...,ε t-q is white noise; is an autoregressive coefficient, i = 1, 2,..., p, ψ j is a moving average coefficient, j = 1, 2,..., q; p and q are the autoregressive model order and the moving average model order respectively, and the order is determined by using the public method AIC method, and p and q corresponding to the minimum AIC(p, q) are expected to be obtained.

[0139] Step four, short-term prediction of S to obtain diesel engine operation parameter short-term prediction data S s , comprising the following steps:

[0140] Step four, I. Once the accumulation operation is performed on S to obtain

[0141]

[0142]

[0143] Wherein, γ ∈ [1, 8], s γ (i') is the i'th element in S γ , is the k'th element in S , is the result of row accumulation of elements in S γ ;

[0144] Step four, II. Constructing the whitening sequence (1) using S T obtained in step four, I.

[0145]

[0146]

[0147] ω = (B T B) -1 B T Y

[0148]

[0149] Wherein, ω is the development coefficient, and the value of ω represents the development trend of the original data retained in the whitening data, is the The k-th element, B and Y are intermediate matrices, when k=1, Set to 0;

[0150] Step 43: Using the S obtained in Step 41 (1) And W obtained in step four two (1) Establish the GM(1,1) differential equation, and thus establish the GM(1,1) grey prediction model:

[0151] Establish the GM(1,1) differential equation:

[0152]

[0153] Where α is the grey action quantity, the development coefficient ω = (B) is calculated using the least squares method. T B) -1 B T Y;

[0154] Establish a GM(1,1) grey prediction model based on sequence Predicted values ​​of sequence changes and parameter sequence predicted values

[0155]

[0156] Step 4: Input S into the GM(1,1) grey prediction model to obtain the future time period n. s =100 parameter time series data sequence;

[0157] Diesel engine operating parameter monitoring data and long-term and short-term prediction results data curves are as follows: Figures 2-9 As shown.

[0158] Step 5: For S, S l S s Perform max-min normalization on each side to obtain S X S Xl S Xs :

[0159] For S, S l S s Perform max-min normalization on each data set to obtain the normalized data set S. X =[S X1 S X2 S X3 S X4 S X5 S X6 S X7 S X8 ] T ,S Xl= [S Xl1 S Xl2 S Xl3 S Xl4 S Xl5 S Xl6 S Xl7 S Xl8 ] T ,S Xs = [S Xs1 S Xs2 S Xs3 S Xs4 S Xs5 S Xs6 S Xs7 S Xs8 ] T ;

[0160] The sequence X is normalized according to the formula, and the normalized sequence is obtained The formula is as follows:

[0161]

[0162] Wherein, k = 1, 2, …, M, is the kth value of the normalized sequence x min The minimum value in the sequence X, x max The maximum value in the sequence X, X = [x1, x2, x3, …, x M ], X is any parameter time series data in S, S l , S s x k is the kth element in X.

[0163] Step six, as Figure 3 shown, according to the analytic hierarchy process AHP, the diesel engine safety state prediction and evaluation is divided into target layer, attribute layer, feature layer, based on AHP layer using step five obtained S X , S Xl , S Xs obtain diesel engine safety state prediction and evaluation conclusion P1, and based on P1 obtain diesel engine safety state, including the following steps:

[0164] Step six, as Figure 10 shown, according to the analytic hierarchy process AHP, the diesel engine safety state prediction and evaluation is divided into target layer, attribute layer, feature layer;

[0165] The target layer output is diesel engine safety state prediction and evaluation conclusion P1, and the target layer input is the evaluation results of each attribute of the attribute layer BF = [B1, B2, B3, B4, B5].

[0166] The attribute layer contains parameters of the diesel engine safety protection system, and the input is the features contained in each attribute: the long-term prediction attribute contains features F l = F1 = [f kl , f tkl , f frl ], the short-term prediction attribute contains features F s = F2 = [f ks , f tks , f frs ], the monitoring parameter attribute contains features F f = F3 = [f kf , f tkf , f frf ], the threshold frequency domain attribute contains features F t = F4 = [f tl , f ts , f tf ], and the threshold time domain attribute contains features F tr = F5 = [t rtl , t rts , t rtf ];

[0167] Wherein, f kl is the average line slope feature in the long prediction feature, f tkl is the tangent slope feature in the long prediction feature, f frl is the difference-root-mean-square feature in the long prediction feature, f ks is the average line slope feature in the short-term prediction data feature, f tks is the tangent slope feature in the short-term prediction data feature, f frs is the difference-root-mean-square feature in the short-term prediction data feature, f kf is the average line slope feature in the parameter monitoring feature, f tkf is the tangent slope feature in the parameter monitoring feature, f frf is the difference-root-mean-square feature in the parameter monitoring feature, f tl is the parameter long prediction feature in the threshold frequency domain feature, f ts is the short prediction feature in the threshold frequency domain feature, f tf is the monitoring data feature in the threshold frequency domain feature, t rtl is the parameter long prediction threshold value fitting time feature in the threshold time domain feature, t rts is the parameter short prediction threshold value fitting time feature in the threshold time domain feature, and t rtf is the monitoring data threshold value fitting time feature in the threshold time domain feature.

[0168] The feature layer is the key feature of each attribute; and the diesel engine safety state prediction evaluation conclusion P1 is calculated and obtained.

[0169] Step six two, the analytic hierarchy process will be the relative importance of each feature into 1-9 scale, according to the diesel engine factory test outline, diesel engine use and repair manual, diesel engine experimental data and diesel engine simulation running model experimental data, according to the relative importance of each feature index, and according to the relative importance of each feature value to build judgment matrix A*, A 1 , A 2 , A 3 , A 4 , A 5 , A:

[0170]

[0171]

[0172] Among them, is the relative importance value of the supercharger speed characteristic, is the relative importance value of the main engine inlet air temperature characteristic, is the relative importance value of the main engine lubricating oil inlet temperature characteristic, is the relative importance value of the main engine fresh water inlet temperature characteristic, is the relative importance value of the exhaust manifold temperature characteristic, is the relative importance value of the main engine inlet air pressure characteristic, is the relative importance value of the crankcase pressure characteristic, is the relative importance value of the main engine lubricating oil inlet pressure characteristic;

[0173] The relative importance of each feature is "1", which is equally important, "3 / 5 / 7 / 9", which is increasing in importance, and "2 / 4 / 6 / 8", which is equally important and is used as a compromise;

[0174] Definition of judgment matrix A 1 is the relative importance value of the main engine inlet air temperature characteristic,

[0175] Among them

[0176] Among them, is the relative importance value of the main engine inlet air temperature characteristic, is the relative importance value of the main engine inlet air temperature characteristic, is the relative importance value of the main engine inlet air temperature characteristic;

[0177] Definition of judgment matrix A 2For the short-term prediction feature of the attribute layer parameter, the relative importance indicators of each feature are in the order of the moving average slope, the tangent slope, and the difference-standard deviation:

[0178] wherein

[0179] wherein, is the relative importance value of the moving average slope in the short-term prediction feature, is the relative importance value of the tangent slope in the short-term prediction feature, is the relative importance value of the difference-standard deviation in the short-term prediction feature;

[0180] The judgment matrix A is defined 3 For the monitoring feature of the attribute layer parameter, the relative importance indicators of each feature are in the order of the moving average slope, the tangent slope, and the difference-standard deviation:

[0181] wherein

[0182] wherein, is the relative importance value of the moving average slope in the parameter monitoring feature, is the relative importance value of the tangent slope in the parameter monitoring feature, is the relative importance value of the difference-standard deviation in the parameter monitoring feature;

[0183] The judgment matrix A is defined 4 For the threshold value frequency domain feature of the attribute layer, the relative importance indicators of each feature are in the order of the long-term prediction frequency domain feature, the short-term prediction frequency domain feature, and the monitoring parameter frequency domain feature:

[0184] wherein

[0185] wherein, is the relative importance value of the long-term prediction frequency domain feature in the threshold value frequency domain feature, is the relative importance value of the short-term prediction frequency domain feature in the threshold value frequency domain feature, is the relative importance value of the monitoring parameter frequency domain feature in the threshold value frequency domain feature;

[0186] The judgment matrix A is defined 5 For the threshold value time domain feature of the attribute layer, the relative importance indicators of each feature are in the order of the long-term prediction time domain feature, the short-term prediction time domain feature, and the monitoring parameter time domain feature:

[0187] wherein

[0188] wherein, is a relative importance value of the long-term prediction time domain feature in the threshold value time domain feature, is a relative importance value of the short-term prediction time domain feature in the threshold value time domain feature, is a relative importance value of the monitoring parameter time domain feature in the threshold value time domain feature;

[0189] A is the threshold value time domain feature in the step 1 to A 5 is

[0190] The judgment matrix A is defined as the relative importance index of each attribute under the target layer, and the corresponding order is the parameter long-term prediction prediction feature, the short-term prediction data feature, the parameter monitoring feature, the threshold value frequency domain feature and the threshold value time domain feature:

[0191] Wherein a1

[0192] Wherein, a1, a2, a3, a4 and a5 are the relative importance value of the parameter long-term prediction prediction feature, the relative importance value of the short-term prediction data feature, the relative importance value of the parameter monitoring feature, the relative importance value of the threshold value frequency domain feature and the relative importance value of the threshold value time domain feature:

[0193] In the present application

[0194] Step six three, the consistency test is carried out on the judgment matrix constructed in step six two, if the consistency test is passed, step six four is executed, if not, the relative importance appears logical error, and then step six two is returned to modify the relative importance order:

[0195] The consistency ratio CR of each judgment matrix constructed in step six two is calculated respectively, if CR<0.1, the current judgment matrix passes the consistency test, if CR≥0.1, the current judgment matrix does not pass the consistency test, and then step six two is returned:

[0196]

[0197]

[0198] Wherein, RI is the consistency index of the judgment matrix, which can be obtained according to the public table of the consistency test method, υ is the number of evaluation indexes involved in the current judgment matrix, λ max is the maximum eigenvalue in the judgment matrix, that is, the spectral radius of the judgment matrix, and CI is an intermediate variable;

[0199] Step six four, the maximum eigenvalue λ max of the judgment matrix passing the consistency test is calculated, and the corresponding normalized eigenvector C is calculated: max ​

[0200] wherein C is c*, c, c 1 , c 2 , c 3 , c 4 or c 5 , c = [c1c2…c5], to c* is the normalized vector corresponding to the largest eigenvalue of the judgment matrix A*, c is the normalized vector corresponding to the largest eigenvalue of the judgment matrix A, c 1 to c 5 is the normalized vector corresponding to the largest eigenvalue of A 1 to A 5 ; there is a relationship: AC = λ max C.

[0201] In this step, c* = (0.0550 0.2202 0.2752 0.4954 0.1651 0.3303 0.4954 0.4954), c = (0.1601 0.3203 0.4804 0.6405 0.4804), c 1 to c 5 = (0.2673 0.5345 0.8018);

[0202] Step six, based on the weighted statistical method, using the normalized eigenvector C obtained in step six four and S X , S Xl , S Xs obtained in step five, the attribute evaluation feature F η is calculated:

[0203] wherein η = 1, 2, 3, 4, 5, the parameter long-term prediction prediction feature F l = F1 = [f kl , f tkl , f frl ], short-term prediction data feature F s = F2 = [f ks , f tks , f frs ], parameter monitoring feature F f = F3 = [f kf , f tkf , f frf ], threshold frequency domain feature F t = F4 = [f tl , f ts , f tf ], threshold time domain feature F tr = F5 = [t rtl , trts t rtf ];

[0204] Parameter long-term prediction prediction feature F l = F1 = [f kl , f tkl , f frl ] :

[0205]

[0206]

[0207]

[0208] wherein, is the group of mean line slopes of each sequence in S Xl is the mean line slope of the parameter main engine lubricant inlet pressure field prediction feature, is the mean line slope of the parameter turbocharger speed long-term prediction data, is the group of tangent line slopes of each sequence in S Xl is the group of difference-standard deviation combinations of each sequence in S Xl

[0209] Short-term prediction data feature F s = F2 = [f ks , f tks , f frs ] :

[0210]

[0211]

[0212]

[0213] wherein, is the group of mean line slopes of each sequence in S Xs is the group of tangent line slopes of each sequence in S Xs is the group of difference-standard deviation combinations of each sequence in S Xs

[0214] Parameter monitoring feature F f = F3 = [f kf , f tkf , f frf ] :

[0215]

[0216] ​​​​​​

[0217]

[0218] is the set of moving average slopes of each sequence in S X is the set of tangent slopes of each sequence in S X is the set of difference-standard deviation combinations of each sequence in S X

[0219] Threshold frequency domain feature F t = F4 = [f tl , f ts , f tf ] :

[0220]

[0221]

[0222]

[0223] is the set of threshold breakouts of each sequence in S Xl is the set of threshold breakouts of each sequence in S Xs is the set of threshold breakouts of each sequence in S X

[0224] Threshold time domain feature F tr = F5 = [t rtl , t rts , t rtf ] :

[0225]

[0226]

[0227]

[0228] is the set of threshold fit / breakout times of each sequence in S Xl is the set of threshold breakouts of each sequence in S Xs is the set of threshold breakouts of each sequence in S X

[0229] Moving average slope f k , tangent slope f tk , difference-root mean square f​​​​​​​​​fr , threshold value break f t , threshold value fit / break time t rt , obtained by the following formula:

[0230] Moving average slope

[0231] Wherein, μ takes Xγ, Xlγ or Xsγ;

[0232] Tangent slope f tk = max(s μ (i1+1)-s μ (i1)), i1 = 1, 2, …, M;

[0233] Difference-standard deviation

[0234] Wherein, Δs μ (i2) is an intermediate variable, i1, i2 are integers;

[0235] Parameter threshold value break Wherein s rt is a parameter threshold value;

[0236] Threshold value fit time If the parameter threshold value is not broken, the fit time output is null.

[0237] Step six, using the attribute evaluation features F obtained in step six η Get attribute layer evaluation index BF = [B1, B2, B3, B4, B5]:

[0238]

[0239] Wherein, B η is the evaluation result of the ηth attribute of the attribute layer, B1 is the long prediction data attribute evaluation result, B2 is the short prediction data attribute evaluation result, B3 is the monitoring data attribute evaluation result, B4 is the threshold value frequency domain attribute evaluation result, and B5 is the threshold value time domain attribute evaluation result; F η is the calculation value of each feature of the ηth attribute, The weight coefficient of each feature, is the a th feature value of the ηth attribute, a takes 1, 2, 3.

[0240] Step six, using the BF obtained in step six to construct the target layer diesel engine safety state prediction evaluation structure, obtain the diesel engine safety state prediction evaluation conclusion P1, and obtain the diesel engine safety state based on P1:

[0241]

[0242] Wherein, c η is the weight coefficient of each attribute layer, B η is the evaluation result of each attribute layer.

[0243] When P1≤0.1, it indicates that the diesel engine safety state is normal.

[0244] When 0.1

[0245] When 0.2

[0246] When P1>0.7, it indicates that the diesel engine safety state is dangerous.

[0247] Step seven, using the pre-processed initial data set S to construct the threshold time domain feature, and using the threshold time domain feature to obtain the diesel engine safety state running prediction remaining time P2:

[0248]

[0249] Wherein, is the parameter main engine oil inlet pressure threshold step number, is the crankcase pressure threshold step number, is the main engine inlet air pressure threshold step number, is the main engine fresh water inlet temperature threshold step number, if the threshold value is not broken, then P2 value is empty.

[0250] In this step, the diesel engine safety state running prediction remaining time P2 refers to the minimum value of the threshold step number in the I type alarm parameter.

[0251] The present application constructs a diesel engine safety state prediction and evaluation model based on data driving. According to the monitored thermal parameters, the model is pre-processed, and according to the data characteristics and the prediction length requirements, ARMA is used for long-term prediction and GM(1,1) is used for short-term prediction; the monitoring data and the prediction data are extracted, the degradation state is analyzed; the diesel engine safety state prediction and evaluation model is constructed based on AHP, the diesel engine safety protection parameter prediction and the diesel engine safety state prediction and evaluation functions are realized.

[0252] Embodiment:

[0253] Taking 620V12 type diesel engine as an embodiment, according to the safety operation protection requirements of this embodiment, the threshold values of each parameter are shown in table 1-table 3:

[0254] Table 1. 620V12 type diesel engine safety evaluation parameter threshold

[0255]

[0256] The embodiment completes AHP calculation, calculates the target layer result of the application according to the diesel engine monitoring parameters and the simulation running model data, tests the AHP output result, and determines the diesel engine safety state prediction evaluation conclusion interval when the AHP score is less than 0.1 in the diesel engine safety state running:

[0257] Table 2. Determining the diesel engine safety state prediction evaluation conclusion interval

[0258]

[0259] Table 3. Diesel engine safety state prediction evaluation AHP output result

[0260]

[0261]

[0262] It is proved by the evaluation result and the verification result that the diesel engine safety state prediction and evaluation method based on parameter prediction can realize the diesel engine safety protection parameter prediction, and the AHP is used to predict and evaluate the diesel engine safety state. Compared with the prior art, the application first realizes the long-term prediction and short-term prediction of the diesel engine safety protection parameter; then the monitoring and prediction data are feature extracted; finally, based on the AHP method, the long-term and short-term prediction features, the monitoring features and the security threshold time-frequency domain features are comprehensively used to construct the diesel engine safety running state prediction and evaluation structure. The application optimizes the limitation of the prediction and evaluation in the current diesel engine safety protection system, provides the prediction and evaluation information before the diesel engine protection shutdown, and improves the diesel engine running safety.

Claims

1. A method for assessing the safety status of diesel engines based on parameter prediction, characterized in that... The specific process of the method is as follows: Step 1: Obtain the thermal parameters for diesel engine safety protection and assemble an initial data set. in, It is a timing data sequence of the main engine lubricating oil inlet pressure. It is a crankcase pressure timing data sequence. It is a time-series data sequence of the main unit's inlet air pressure. It is a time-series data sequence of exhaust manifold temperature. It is a time-series data sequence of the freshwater inlet temperature of the main unit. It is a time-series data sequence of the main engine lubricating oil inlet temperature. It is a time-series data sequence of the air temperature at the main unit's inlet. It is a timing data sequence of the turbocharger speed; Step 2: Preprocess the initial data set using the Laida criterion to obtain the preprocessed initial data set S; Step 3: Perform long-term forecasting of S to obtain long-term forecast data of diesel engine operating parameters S. l This includes the following steps: Step 3:

1. Use the unit root test to check the stationarity of each preprocessed parameter time series in S. If it passes the stationarity test, then S is considered a stationary series. f Proceed to step 32; if the stationarity test is not passed, perform differencing on S until the differencing sequence S passes the stationarity test, thus obtaining a stationary sequence S. f Then proceed to step three two; Step 3.2: Establish a long-term prediction model, and use the stationary series S f The data is input into the long-term parameter prediction model to obtain the long-term prediction data S of the diesel engine operating parameters. l ; The long-term prediction model for the parameters is as follows: Among them, sf t Let ε be the parameter value at time t in the stationary sequence. t ,...,ε t-q It is white noise; These are the autoregressive coefficients, i = 1, 2, ..., p, ψ j These are the moving average coefficients, j = 1, 2, ..., q; p and q are the orders of the autoregressive model and the moving average model, respectively; Step 4: Perform short-term forecasting of S to obtain short-term forecast data of diesel engine operating parameters S. s This includes the following steps: Step 41: Perform an accumulation operation on S to obtain... Where γ∈[1,8], s γ (i') is S γ The i'th element in yes The k-th element in It is for S γ The result of summing the elements in the middle row; Step 42: Using the S obtained in Step 41 (1) Constructing whitening sequences ω=(B T B) -1 B T Y Where ω is the development coefficient, It is in the middle The k-th element, B and Y are intermediate matrices; Step 43: Using the S obtained in Step 41 (1) And W obtained in step four two (1) Establish the GM(1,1) grey prediction model as follows: Where α is the gray action quantity. It is a predicted value of sequence change. These are the predicted values ​​of the parameter sequence; Step 4: Input S into the GM(1,1) grey prediction model to obtain short prediction data S of diesel engine operating parameters. s ; Step 5: For S, S l S s Normalization was performed separately to obtain S X S Xl S Xs ; Step Six: Utilize the S obtained in Step Five X S Xl S Xs Obtain the diesel engine safety condition prediction and assessment conclusion P1: First, using S X S Xl S Xs Obtain parameters for long-term prediction of attributes including features F l Short-term predictive attributes include feature F s The monitoring parameter attributes include feature F f Threshold value frequency domain attribute includes feature F t Threshold value time-domain attribute includes feature F tr The value; then, using F l F s F f F t F tr The value of P1 is used to obtain the prediction and assessment conclusion of the diesel engine's safety status. Among them, the larger P1 is, the more dangerous the diesel engine condition; Step 7: Utilize the threshold value obtained in Step 6 to include the temporal attribute containing feature F. tr Obtain the remaining duration P2 of the diesel engine's safe operating condition prediction.

2. The diesel engine safety status assessment method based on parameter prediction according to claim 1, characterized in that: In step two, the initial data set is preprocessed using the Laida criterion to obtain the preprocessed initial data set S, specifically as follows: Step 2: Divide the time series data sequence of each parameter in the initial data set into n equally spaced segments, and obtain the average value of the time series data in each segment. Step 22: Utilize the average value of each time series data segment Obtain the residual error v of each time series data segment i Sum of standard deviations σ: Where m is the total number of data points in each time series data segment, and M is the total number of data points in each parameter time series data sequence. It is the i0th time series data within a certain time series data segment; Steps two and three: using v i The outliers in each time series data segment are obtained using σ, and then removed. The positions of the removed outliers are then filled in to obtain the preprocessed initial data set S = [S1 S2 S3 S4 S5 S6 S7 S8]. T : First, determine Does it meet the requirements? If satisfied, then Outlier; Then, outliers were removed. Leave the positions for removing outliers empty; Then, Newton interpolation is used to fill in the missing values, resulting in the preprocessed initial data set S = [S1 S2 S3 S4 S5 S6 S7 S8] T ; Among them, S1 is the pre-processed main engine lubricating oil inlet pressure time sequence data, S2 is the pre-processed crankcase pressure time sequence data, S3 is the pre-processed main engine inlet air pressure time sequence data, S4 is the pre-processed exhaust manifold temperature time sequence data, S5 is the pre-processed main engine fresh water inlet temperature time sequence data, S6 is the pre-processed main engine lubricating oil inlet temperature time sequence data, S7 is the pre-processed main engine inlet air temperature time sequence data, and S8 is the pre-processed turbocharger speed time sequence data.

3. The diesel engine safety status assessment method based on parameter prediction according to claim 2, characterized in that: In step five, S, S l S s Normalization was performed separately to obtain S X S Xl S Xs Specifically: For S, S l S s Perform max-min normalization on each data set to obtain the normalized data set S. X =[S X1 S X2 S X3 S X4 S X5 S X6 S X7 S X8 ] T ,S Xl =[S Xl1 S Xl2 S Xl3 S Xl4 S Xl5 S Xl6 S Xl7 S Xl8 ] T ,S Xs =[S Xs1 S Xs2 S Xs3 S Xs4 S Xs5 S Xs6 S Xs7 S Xs8 ] T ; Among them, S X It is the normalized S, S Xl It is the normalized S l S Xs It is the normalized S s ; The normalization formula for sequence X is used to obtain the normalized sequence. As shown in the following formula: Where k = 1, 2, ..., M, It is a normalized sequence The k-th element, x min The minimum value in sequence X, x max Let X be the maximum value in the sequence X = [x1, x2, x3, ..., x...]. M ], X is S, S l or S s x k It is the k-th element in X.

4. The diesel engine safety status assessment method based on parameter prediction according to claim 3, characterized in that: The S obtained in step five is used in step six. X S Xl S Xs To obtain the diesel engine safety condition prediction assessment conclusion P1, the following steps are included: Step 61: Establish the long-term predictive attribute containing feature F. l =F1=[f kl ,f tkl ,f frl Short-term predictive attributes include feature F s =F2=[f ks ,f tks ,f frs The monitoring parameter attributes include feature F. f =F3=[f kf ,f tkf ,f frf Threshold value frequency domain attribute includes feature F t =F4=[f tl ,f ts ,f tf Threshold value time-domain attribute includes feature F tr =F5=[t rtl ,t rts ,t rtf ]; Among them, f kl It is the moving average slope feature in long-term prediction features, f tkl It is the tangent slope feature in long prediction features, f frl It is the difference-root mean square feature in long predictive features, f ks It is the slope feature of the moving average in short-term forecast data, f tks It is the tangent slope feature in short-term forecast data features, f frs It is the difference-root mean square feature in short-term predictive data features, f kf It is the moving average slope feature in the parameter monitoring characteristics, f tkf It is the tangent slope feature in the parameter monitoring features, f frf It is the difference-root mean square feature in parameter monitoring features, f tl It is a parameter-long prediction feature in the threshold value frequency domain features, f ts It is a short prediction feature in the threshold value frequency domain features, f tf It is a monitoring data feature in the frequency domain characteristics of the threshold value, t rtl It is the parameter long prediction threshold fitting time feature in the time domain feature of the threshold value, t rts It is the parameter short prediction threshold fitting time feature in the threshold time domain feature, t rtf It refers to the monitoring data threshold value matching time characteristic in the time domain characteristics of the threshold value; Step 62: Input parameters for the thermal parameters of diesel engine safety protection and assign relative importance values ​​to the features established in Step 61, and construct judgment matrices A* and A*. 1 A 2 A 3 A 4 A 5 A: The relative importance value is an integer between 1 and 9; in, The relative importance value of the turbocharger speed characteristics. The relative importance value of the air temperature characteristics at the main unit inlet. The relative importance value of the main engine lubricating oil inlet temperature characteristic. The relative importance value of the freshwater inlet temperature characteristic of the main unit. The relative importance value of the exhaust manifold temperature characteristics. The relative importance value of the main unit's inlet air pressure characteristics. This represents the relative importance value of the crankcase pressure characteristics. The relative importance value of the main engine lubricating oil inlet pressure characteristic; in in, It is the relative importance value of the moving average slope feature in long-term prediction characteristics. It is the relative importance value of the tangent slope feature in long-term predictive features. It is the relative importance value of the difference-standard deviation feature in long-term predictive features; in in, It is the relative importance value of the slope feature in short-term prediction features. It represents the relative importance of the tangent slope feature among short-term predictive features. It is the relative importance value of the difference-standard deviation feature in short-term predictive features; in in, It represents the relative importance of the moving average slope feature among the parameter monitoring characteristics. Among the parameter monitoring features, the relative importance value of the tangent slope feature is... It is the relative importance value of the difference-standard deviation feature in the parameter monitoring characteristics; in in, It is the relative importance value of the long-term predicted frequency domain features in the threshold frequency domain features. It is the relative importance value of the short-term prediction frequency domain features in the threshold frequency domain features. It is the relative importance value of the frequency domain characteristics of the monitoring parameters in the frequency domain characteristics of the threshold value; in in, It is the relative importance value of the long-term predictive time-domain features among the threshold time-domain features. It is the relative importance value of short-term prediction time-domain features in the threshold time-domain features. It is the relative importance value of the time-domain characteristics of the monitoring parameters in the time-domain characteristics of the threshold value; Where a1 < a2 < a3 < a5 < a4; Wherein, a1, a2, a3, a4, and a5 are, respectively, the relative importance values ​​of the long-term prediction feature of the parameter, the relative importance values ​​of the short-term prediction data feature, the relative importance values ​​of the parameter monitoring feature, the relative importance values ​​of the threshold value frequency domain feature, and the relative importance values ​​of the threshold value time domain feature. Step 63: Perform a consistency check on the judgment matrix constructed in Step 62. If the judgment matrix passes the consistency check, proceed to Step 64. If the judgment matrix fails the consistency check, return to Step 62 to reset the relative importance value. A consistency check is performed on the judgment matrix constructed in step six-two, specifically as follows: Calculate the consistency ratio CR of each judgment matrix constructed in step six-two. If CR < 0.1, the current judgment matrix passes the consistency test; if CR ≥ 0.1, the current judgment matrix fails the consistency test. Where RI is the consistency index of the judgment matrix, υ is the number of evaluation indicators involved in the current judgment matrix, and λ max CI is an intermediate variable used to determine the largest eigenvalue in a matrix; Step 64: Obtain the largest eigenvalue λ of the judgment matrix that passes the consistency test. max and obtain λ max The corresponding standardized feature vector C; Where C is c*, c, c 1 c 2 c 3 c 4 or c 5 , c = [c1c2…c5], to c* is the standardized vector corresponding to the largest eigenvalue in the judgment matrix A*, and c is the standardized vector corresponding to the largest eigenvalue in the judgment matrix A. 1 To c 5 It is A 1 To A 5 The largest eigenvalue in the vector corresponds to the standardized vector. Step 65: Using the weighted statistical method, utilize the standardized eigenvector C obtained in Step 64 and the S obtained in Step 5. X S Xl S Xs Obtain attribute evaluation feature F η : Step 66: Utilize the attribute evaluation feature F obtained in Step 65. η Obtain the attribute evaluation index BF = [B1, B2, B3, B4, B5]: Step 67: Use the BF obtained in Step 66 to obtain the diesel engine safety status prediction and assessment conclusion P1.

5. The diesel engine safety status assessment method based on parameter prediction according to claim 4, characterized in that: In step six-five, the weighted statistical method utilizes the standardized feature vector C obtained in step six-four and the S obtained in step five. X S Xl S Xs Obtain attribute evaluation feature F η , Includes the following steps: Where η = 1, 2, 3, 4, 5, the long-term prediction feature F is... l =F1=[f kl ,f tkl ,f frl Short-term forecast data characteristics F s =F2=[f ks ,f tks ,f frs ] Parameter monitoring characteristics F f =F3=[f kf ,f tkf ,f frf Threshold value frequency domain characteristics F t =F4=[f tl ,f ts ,f tf Threshold value time-domain characteristics F tr =F5=[t rtl ,t rts ,t rtf ]; f kf f tkf f frf f tl f ts f tf t rtl t rts t rtf Form and f kl f tkl f frl same; in, It is S Xl The moving average slope group for each sequence. The slope of the moving average of the predicted characteristic of the main engine lubricating oil inlet pressure field is given by parameter. The slope of the average line for the long-term prediction data of the turbocharger speed parameter. It is S Xl The set of tangent slopes for each sequence. It is S Xl The difference-standard deviation combination for each sequence; It is S Xs The moving average slope group for each sequence. It is S Xs The set of tangent slopes for each sequence. It is S Xs The difference-standard deviation combination for each sequence; Slope f of moving average k , Tangent slope f tk Difference-root mean square f fr Threshold value breach f t Threshold fit / breakout time t rt It can be obtained through the following formula: f tk =max(s μ (i1+1)-s μ (i1)),i1=1,2,……,M Where μ takes the values ​​Xγ, Xlγ, or Xsγ, i1 and i2 are integers, and Δs μ (i2) is an intermediate variable, s rt This is the parameter threshold value.

6. The diesel engine safety status assessment method based on parameter prediction according to claim 5, characterized in that: The attribute evaluation feature F obtained in step 65 is used in step 66. η Obtain the attribute evaluation index BF = [B1, B2, B3, B4, B5], as shown in the following formula: Among them, B1 is the evaluation result of long-term forecast data attributes, B2 is the evaluation result of short-term forecast data attributes, B3 is the evaluation result of monitoring data attributes, B4 is the evaluation result of threshold value frequency domain attributes, and B5 is the evaluation result of threshold value time domain attributes; F η Let be the calculated values ​​of each feature of the ηth attribute. Weight coefficients of each feature Let a be the a-th feature value of the η-th attribute, where a takes the values ​​1, 2, or 3.

7. The diesel engine safety status assessment method based on parameter prediction according to claim 6, characterized in that: The step six-seven, which utilizes the safety status prediction and assessment conclusion P1 obtained in step six-six, specifically involves: Among them, c η It is an element of c, B η It is an element in BF.

8. The diesel engine safety status assessment method based on parameter prediction according to claim 7, characterized in that: The temporal attribute of the threshold value obtained in step six in step seven includes feature F. tr The remaining time P2 for predicting the safe operating condition of the diesel engine is obtained as follows: in, The number of steps required to match the main engine lubricating oil inlet pressure threshold. It is the crankcase pressure threshold fitting step number. It is the number of fitting steps for the main unit's inlet air pressure threshold. The number of steps for fitting the freshwater inlet temperature threshold of the main unit.

Images