A method for fault diagnosis of short-process electric furnace steelmaking equipment

Abstract

The present invention relates to the field of fault analysis technology, and specifically to a fault diagnosis method for short-process electric furnace steelmaking equipment, the method comprising: collecting multiple parameters such as furnace body temperature, furnace chamber pressure, and electrode current from the electric furnace steelmaking equipment, performing time series analysis on the electric furnace operation data, extracting the time interval and sequence order, and obtaining time series structure data. In the present invention, by reconstructing the phase space of the dynamic system and analyzing the sensitivity to initial conditions in chaos theory, it is possible to more accurately monitor and predict the behavior of the system, effectively identify small changes in the system, and realize the detection of early faults. In addition, by identifying chaotic areas, the monitoring capability of potential fluctuations in the system is further enhanced. By combining spectral clustering with a local anomaly factor algorithm, as well as effectively classifying and detecting anomalies in data, the accuracy of fault location is significantly improved, thereby optimizing the maintenance decision-making process.

Description

Technical Field

[0001] The invention relates to the technical field of fault analysis, and in particular to a fault diagnosis method for short-process electric furnace steelmaking equipment. Background Art

[0002] The field of failure analysis technology involves the use of a systematic approach to identify, analyze, and resolve failures in equipment, systems, or processes. Widely used in the industrial and manufacturing sectors, the core task of failure analysis is to determine the root cause of the failure so that preventive measures can be taken to avoid future recurrence of the problem. It includes data collection, failure simulation, testing and analysis, and the use of various diagnostic tools and techniques. Failure analysis is not limited to post-event diagnosis, but also includes predictive maintenance and real-time monitoring to identify potential problems in advance and repair them in a timely manner.

[0003] Among them, the short-process electric furnace steelmaking equipment fault diagnosis method refers to a system designed for the electric furnace steelmaking process in steel production, which is used to identify and diagnose equipment failures that may occur during the production process. The system monitors the equipment operating status and performance indicators, and uses algorithms and models to analyze the data, thereby discovering problems in real time and pointing out possible causes of failures. Its main purpose is to improve production efficiency, reduce downtime, and ensure the stability and safety of the production process.

[0004] However, the existing technology is difficult to effectively capture all potential fault signals when dealing with nonlinear and dynamically changing electric furnace steelmaking equipment. This results in the inability to accurately judge small changes in the system, resulting in excessive or insufficient maintenance of the equipment, affecting production efficiency and equipment life. In terms of data processing and fault area location, it is difficult to effectively distinguish and identify various complex fault modes, resulting in a lack of pertinence and accuracy in the fault diagnosis process. This not only prolongs the fault response time, but also may cause unnecessary maintenance costs due to misdiagnosis. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings in the prior art and to propose a fault diagnosis method for short-process electric furnace steelmaking equipment.

[0006] In order to achieve the above object, the present invention adopts the following technical scheme: a method for diagnosing faults of short-process electric furnace steelmaking equipment, comprising the following steps:

[0007] Collect multiple parameters such as furnace body temperature, furnace chamber pressure, and electrode current from electric furnace steelmaking equipment, perform time series analysis on the operation data of the electric furnace, extract the time interval and sequence order, and obtain time series structure data;

[0008] Based on the time series structure data, an embedding dimension and a time delay parameter are selected, and the electric furnace operation data is spatially mapped to obtain a phase space trajectory diagram;

[0009] Performing dynamic analysis on the phase space trajectory diagram, calculating the trajectory divergence of furnace temperature, pressure and current state points, identifying chaotic behavior and fluctuation areas of steelmaking equipment through quantitative analysis of the divergence results, and generating a fault analysis index value;

[0010] Based on the fault analysis index value, the equipment operation status is divided by spectral clustering, local density analysis is applied to each status, and abnormal data points are selected according to local density comparison to obtain the electric furnace fault location analysis result.

[0011] As a further solution of the present invention, the step of acquiring the time series structure data is specifically as follows:

[0012] Collect the furnace temperature of electric furnace steelmaking equipment , furnace chamber pressure and electrode current Multiple parameters, set the initial data array , using the formula:

[0013]

[0014] Calculate the parameter changes at consecutive time points to obtain time interval data ;

[0015] in, Represents the initial data array, , and are the coefficients for time series analysis, is the attenuation factor, represents the time index, represents the base of natural logarithm, and represent the data of the previous time point and the previous two time points, respectively;

[0016] Based on the time interval data , according to the formula:

[0017] Perform a weighted logarithmic transformation on the sequence data to generate sequential data of the time series ;

[0018] in, is the weight coefficient, represents the natural logarithm, The furnace temperature is The data value at a point in time, express The absolute value of To avoid When the logarithm is 0, it is undefined. is the index in the sequence, indicating that all data points are operated on. is the total number of data points;

[0019] Combined with the time interval data and time series sequential data , using the formula:

[0020] Get time series structure data ;

[0021] in, and represents the adjustment factor combining the time interval and the sequence order, The adjustment factor is used to adjust the nonlinear characteristics of the response. is an exponential function used to adjust and Influence degree.

[0022] As a further solution of the present invention, the steps of acquiring the embedding dimension and the time delay parameter are specifically as follows:

[0023] Based on the time series structure data , using the formula:

[0024] Computing time series structure data The weighted average and weighted standard deviation ;

[0025] Through the time series structure data The weighted average and weighted standard deviation , according to the formula:

[0026] Get the embedding dimension An estimated value of

[0027] in, Representing time series structured data The data points, , is the data point The weight of is the total number of data points, represents the natural logarithm, is the index into the sequence, represents the natural index;

[0028] According to the embedding dimension Estimation and time series structure data The characteristics of , using the formula:

[0029] Calculate the autocorrelation function ;

[0030] in, A value representing the time delay, is used for weighting and The coefficient of correlation, is the total number of data points, is the index in the sequence;

[0031] In the autocorrelation function Find the first zero crossing point in , using the formula:

[0032] Get the time delay parameters The value of

[0033] in, is the autocorrelation function, is satisfied Minimum value, is the time delay parameter, is the index into the sequence.

[0034] As a further solution of the present invention, the step of obtaining the phase space trajectory diagram is specifically:

[0035] Using the embedding dimension The estimated value and time delay parameter of The value of is, using the formula:

[0036] For time series structured data Reconstruct to form a state vector ;

[0037] in, is the state vector, Represents a time series in time data points, is the time delay parameter, is the embedding dimension, Represents a time series index;

[0038] Based on the state vector , using the formula:

[0039] Perform spatial mapping to obtain the mapped state vector ;

[0040] in, For the The mapping coefficient of each dimension is used to adjust the contribution of each dimension in the mapping;

[0041] Based on the mapped state vector , using the formula;

[0042] Perform normalization to obtain the processed state vector , generate phase space trajectory diagram;

[0043] in, and They are The minimum and maximum values ​​in the data series.

[0044] As a further solution of the present invention, the steps for obtaining the trajectory divergence of the furnace temperature, pressure and current state points are specifically as follows:

[0045] From the processed state vector Extract furnace temperature from the data ,pressure and current The time series is:

[0046] Calculate the first-order difference of the time series and get the first-order difference of the time series of the furnace temperature , the first difference of the pressure time series The first-order difference of the current time series ;

[0047] in, and Respectively expressed in time and The furnace temperature, and Respectively expressed in time and pressure, and Respectively expressed in time and The current;

[0048] The first-order difference of the furnace temperature time series , the first difference of the pressure time series The first-order difference of the current time series To convert to absolute value, use the formula:

[0049] Accumulate the absolute values ​​to get the total divergence value of the furnace temperature , the total divergence value of pressure , the total divergence value of the current ;

[0050] in, is the length of the time series, indicating the total number of time points in the sequence, is the summation symbol, which is used to accumulate the divergence value at each time point. is the time point index in the sequence, used to iterate over each time point.

[0051] As a further solution of the present invention, the step of obtaining the fault analysis index value is specifically as follows:

[0052] The total dispersion value of the furnace temperature , the total divergence value of pressure , the total divergence value of the current Compared with the sum of furnace temperature, pressure and current divergence values, the formula is:

[0053] Calculate the ratio of a single total divergence value to the sum of furnace temperature, pressure, and current divergence values ​​to obtain the ratio of furnace temperature to total divergence , the ratio of pressure to total divergence The ratio of the sum current to the total divergence ;

[0054] in, is the constant in the denominator;

[0055] Based on the ratio of the furnace temperature to the total scatter value , the ratio of pressure to total divergence The ratio of the sum current to the total divergence , using the formula:

[0056] Calculate the mean scatter ratio , quantifying overall volatility and potential instability;

[0057] Based on the average scatter ratio , using the formula:

[0058] Calculation of the failure index based on the maximum volatility parameter , and obtain the fault analysis index value.

[0059] As a further solution of the present invention, the steps of dividing the equipment operation states and applying local density analysis to each state are specifically as follows:

[0060] Based on the fault analysis index value, the formula is used:

[0061] Build a similarity matrix and calculate device status and The similarity between ;

[0062] in, and Status and Fault analysis indicators, is the standard deviation parameter, () represents the exponential function, is the offset in the exponential function, is an index used to identify the device status;

[0063] In the device status and The similarity between Based on this, the formula is used:

[0064] Apply local density analysis to each state to obtain the device state The local density ;

[0065] in, is the similarity threshold, is the indicator function, when the condition The value is 1 if it is true, otherwise it is 0. is an index used to identify the device status. is the total number of states.

[0066] As a further solution of the present invention, the steps for obtaining the electric furnace fault location analysis result are specifically as follows:

[0067] Based on the device status The local density , using the formula:

[0068] Calculate the local density of all device states The average , assists in identifying abnormal data points;

[0069] in, is the total number of states analyzed, is the device status;

[0070] Based on the local density of all device states The average , using the formula:

[0071] Defining anomaly thresholds ;

[0072] in, is a coefficient less than 1, used to adjust the sensitivity;

[0073] According to the abnormal threshold , using the formula:

[0074] Calculate the set of states marked as abnormal , get the electric furnace fault location analysis results;

[0075] in, Is the device status.

[0076] Compared with the prior art, the advantages and positive effects of the present invention are:

[0077] In the present invention, by reconstructing the phase space of the dynamic system and analyzing the sensitivity of initial conditions in chaos theory, the system behavior can be monitored and predicted more accurately, and small changes in the system can be effectively identified, thereby realizing the detection of early faults. In addition, by identifying the chaotic region, the monitoring capability of the potential fluctuations of the system is further enhanced. By combining spectral clustering with the local anomaly factor algorithm, as well as effectively classifying and detecting anomalies of the data, the accuracy of fault location is significantly improved, thereby optimizing the maintenance decision-making process and reducing unnecessary maintenance costs and equipment downtime. BRIEF DESCRIPTION OF THE DRAWINGS

[0078] Figure 1 is a system flow chart of the present invention;

[0079] Figure 2 A flowchart of the steps for obtaining time series structure data of the present invention;

[0080] Figure 3 A flowchart of the steps of acquiring the embedding dimension and time delay parameters of the present invention;

[0081] Figure 4 Flow chart of the steps of obtaining the phase space trajectory diagram of the present invention;

[0082] Figure 5 A flow chart of the steps for obtaining the trajectory divergence of furnace temperature, pressure and current state points of the present invention;

[0083] Figure 6 A flowchart of the steps for obtaining the fault analysis index value of the present invention;

[0084] Figure 7 A flowchart of the steps of obtaining the equipment operating states and applying local density analysis to each state for the present invention;

[0085] Figure 8 The present invention is a flow chart of the steps for obtaining the results of electric furnace fault location analysis. DETAILED DESCRIPTION

[0086] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0087] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, in the description of the present invention, "multiple" means two or more, unless otherwise clearly and specifically defined.

[0088] See also Figure 1 A method for diagnosing faults of short-process electric furnace steelmaking equipment comprises the following steps:

[0089] Collect multiple parameters such as furnace body temperature, furnace chamber pressure, and electrode current from electric furnace steelmaking equipment, perform time series analysis on the operation data of the electric furnace, extract the time interval and sequence order, and obtain time series structure data;

[0090] Based on the time series structure data, the embedding dimension and time delay parameters are selected, and the electric furnace operation data is spatially mapped to obtain the phase space trajectory diagram;

[0091] Perform dynamic analysis on the phase space trajectory diagram, calculate the trajectory divergence of furnace temperature, pressure and current state points, identify the chaotic behavior and fluctuation area of ​​steelmaking equipment through quantitative analysis of the divergence results, and generate fault analysis index values;

[0092] Based on the fault analysis index value, the equipment operation status is divided by spectral clustering, local density analysis is applied to each status, and abnormal data points are selected according to local density comparison to obtain the fault location analysis results of the electric furnace.

[0093] The time series structure data includes time tags, numerical readings and key sequence nodes extracted from furnace body temperature, furnace chamber pressure and electrode current. The phase space trajectory diagram includes the dimensional mapping representation, embedded dimension value and time delay interval configuration of the electric furnace operation data. The fault analysis index value includes the quantitatively measured state point dispersion value, the calibrated chaotic behavior type and the fluctuation area range. The electric furnace fault location analysis results include the identified abnormal data point locations, the clustering results of related states and the density parameters used for division.

[0094] See also Figure 2 , the specific steps for obtaining time series structure data are:

[0095] Collect the furnace temperature of electric furnace steelmaking equipment , furnace chamber pressure and electrode current Multiple parameters, set the initial data array , using the formula:

[0096] Calculate the parameter changes at consecutive time points to obtain time interval data ;

[0097] in, Represents the initial data array, , and For the coefficients of time series analysis, adjust the influence of each parameter in the model, is the attenuation factor, represents the time index, represents the base of natural logarithm, and Respectively represent the data of the previous time point and the previous two time points, which is used to consider the historical impact of the data;

[0098] Based on time interval data , according to the formula:

[0099] Perform a weighted logarithmic transformation on the sequence data to generate sequential data of the time series ;

[0100] in, is the weight coefficient, represents the natural logarithm, The furnace temperature is The data value at a point in time, express The absolute value of To avoid When the logarithm is 0, it is undefined. is the index in the sequence, indicating that all data points are operated on. is the total number of data points;

[0101] Combining time interval data and time series sequential data , using the formula:

[0102] Get time series structure data ;

[0103] in, and represents the adjustment factor for the comprehensive time interval and sequence order, The adjustment factor is used to adjust the nonlinear characteristics of the response. is an exponential function used to adjust and Influence to achieve nonlinear adjustment and enhance the sensitivity of the model to abnormal data.

[0104] The calculation process is as follows:

[0105] Assume that the data set furnace temperature =[1500,1520,1510] degrees; furnace chamber pressure =[101,100,102]kPa; electrode current =[1000,1010,995]A;

[0106] For the formula: ;

[0107] Assumptions , , , ,time Starting from 1, for Situation (i.e. the third time point): ,calculate:

[0108] Substituting these values ​​into the formula:

[0109] Represents the furnace temperature forecast at the third time point after adjustment by weighted average and historical data.

[0110] For the formula: ;

[0111] Assumptions , , ,calculate:

[0112]

[0113]

[0114] Represents ordinal data calculated from a weighted log transformation.

[0115] For the formula: ;

[0116] Assumptions , , ,calculate:

[0117] It represents the final time series structure data, reflecting the comprehensive data processing result combining time and series order at time point 3.

[0118] See also Figure 3 , the steps for obtaining the embedding dimension and time delay parameters are as follows:

[0119] Based on time series structure data , using the formula:

[0120]

[0121] Computing time series structure data The weighted average and weighted standard deviation ;

[0122] Structuring data through time series The weighted average and weighted standard deviation , according to the formula:

[0123] Get the embedding dimension An estimated value of

[0124] in, Representing time series structured data The data points, , is the data point The weight of is the total number of data points, represents the natural logarithm, It is the index in the sequence, used to index time series structure data The elements in represents the natural index;

[0125] According to the embedding dimension Estimation and time series structure data The characteristics of , using the formula:

[0126] Calculate the autocorrelation function ;

[0127] in, A value representing the time delay, is used for weighting and The coefficient of correlation, is the total number of data points, is the index in the sequence;

[0128] In the autocorrelation function Find the first zero crossing point in , using the formula:

[0129] Get the time delay parameters The value of

[0130] in, is the autocorrelation function, representing the time series structure data With The autocorrelation function of the value after time, is satisfied Minimum value, is the time delay parameter, is the index in the sequence;

[0131] The calculation process is as follows:

[0132] Assuming time series structure data It consists of three data points: ; Calculate the weighted average ; Assume the weight , calculated as:

[0133] set up , the weight calculation result is:

[0134] Weighted Average: ;

[0135] Calculate weighted standard deviation :

[0136]

[0137] Calculate embedding dimension : ;

[0138] Here It means that according to the characteristics of time series data, the selected embedding dimension is 22, which can affect the complexity and performance of subsequent models.

[0139] Calculate the autocorrelation function , assuming the calculation Autocorrelation of:

[0140]

[0141] Finding the zero crossing point And calculate the time delay parameters , assuming (Right now ),calculate :

[0142]

[0143] According to The autocorrelation property of the model is used to determine the correlation distance between data points, which in turn affects the response sensitivity and prediction accuracy of the model.

[0144] See also Figure 4 , the steps to obtain the phase space trajectory diagram are as follows:

[0145] Using Embedding Dimension The estimated value and time delay parameter of The value of is, using the formula:

[0146] For time series structured data Reconstruct to form a state vector ;

[0147] in, is the state vector, i.e. at time Reconstructed state vector, converting time series structure data The one-dimensional time series data is One-dimensional phase space data, Represents a time series in time data points, is the time delay parameter, is the embedding dimension, Represents a time series index;

[0148] Based on the state vector , using the formula:

[0149] Perform spatial mapping to obtain the mapped state vector ;

[0150] in, For the The mapping coefficient of the dimensions is used to adjust the contribution of each dimension in the mapping. It can be assumed The weight coefficient is simplified so that the more recent time points have higher weights;

[0151] Based on the mapped state vector , using the formula;

[0152] Perform normalization to obtain the processed state vector , generate phase space trajectory diagram;

[0153] in, and They are The minimum and maximum values ​​in the data sequence, after processing The values ​​will be mapped to the interval [0,1], which is suitable for graphical display;

[0154] The calculation process is as follows:

[0155] For the formula: ;

[0156] Assumptions ;

[0157] Assumptions

[0158] According to the given and , reconstruct the time series :

[0159]

[0160] And so on, until .

[0161] For the formula: ;

[0162] Mapping coefficient Simplify the selection to ;

[0163] Assume the mapping coefficients: .

[0164] Weighted calculation using mapping coefficients :

[0165] right of :

[0166] right of :

[0167] For the formula: ;

[0168] calculate The minimum and maximum values ​​of are:

[0169] Assumptions ; Minimum ; Maximum value ;

[0170] Normalization: of : ;

[0171] right of : ;

[0172] Interpretation of results The value of ranges from 0 to 1, indicating the relative position of the normalized data points in the phase space, where 0 and 1 represent the minimum and maximum mapping values, respectively. This normalization makes the data more uniform and interpretable when represented graphically.

[0173] See also Figure 5 , the steps for obtaining the trajectory divergence of the furnace temperature, pressure and current state points are as follows:

[0174] From the processed state vector Extract furnace temperature from the data ,pressure and current The time series is:

[0175] and and .

[0176] Calculate the first-order difference of the time series and get the first-order difference of the time series of the furnace temperature , the first difference of the pressure time series The first-order difference of the current time series ;

[0177] in, and Respectively expressed in time and The furnace temperature, and Respectively expressed in time and pressure, and Respectively expressed in time and The current;

[0178] The first-order difference of the time series of furnace temperature , the first difference of the pressure time series The first-order difference of the current time series To convert to absolute value, use the formula:

[0179]

[0180] Accumulate the absolute values ​​to get the total divergence value of the furnace temperature , the total divergence value of pressure , the total divergence value of the current ;

[0181] in, is the length of the time series, indicating the total number of time points in the sequence, is the summation symbol, which is used to accumulate the divergence value at each time point. is the time point index in the sequence, used to iterate over each time point.

[0182] The calculation process is as follows:

[0183] For the formula:

[0184] Assume that the furnace temperature is Degree, pressure psi, the current is amp.

[0185]

[0186] Calculate the absolute value:

[0187]

[0188] For the formula:

[0189]

[0190] (total furnace temperature change); (total pressure change); (Total current change).

[0191] It shows how the trajectory divergence of a state point can be calculated from given furnace temperature, pressure, and current data. The sum of these values ​​provides quantitative information about the volatility of these parameters during the observation period, which helps in making decisions during maintenance or control processes.

[0192] See also Figure 6 ,The specific steps for obtaining the fault analysis index value are:

[0193] The total dispersion value of the furnace temperature , the total divergence value of pressure , the total divergence value of the current Compared with the sum of furnace temperature, pressure and current divergence values, the formula is: and and .

[0194] Calculate the ratio of a single total divergence value to the sum of furnace temperature, pressure, and current divergence values ​​to obtain the ratio of furnace temperature to total divergence , the ratio of pressure to total divergence The ratio of the sum current to the total divergence ;

[0195] in, It is a constant in the denominator, which prevents the denominator from being zero and stabilizes the proportional calculation.

[0196] Based on the ratio of furnace temperature to total scatter , the ratio of pressure to total divergence The ratio of the sum current to the total divergence , using the formula:

[0197] Calculate the mean scatter ratio , quantifying overall volatility and potential instability;

[0198] Based on the average scatter ratio , using the formula:

[0199] Calculation of the failure index based on the maximum volatility parameter , get the fault analysis index value;

[0200] The calculation process is as follows:

[0201] Assuming furnace temperature divergence is 15, pressure divergence is 3, the current divergence is 8.

[0202] For the formula: and and .

[0203] in Set to 1 for stability.

[0204]

[0205]

[0206]

[0207] For the formula: ;

[0208]

[0209] The calculated value reflects the average volatility level of the system.

[0210] For the formula: ;

[0211]

[0212] The fault index value indicates the potential degree of failure of the system, with lower values ​​indicating a lesser likelihood of failure and higher values ​​requiring further attention.

[0213] See also Figure 7 , divide the equipment operation state, and apply local density analysis to each state to obtain the following steps:

[0214] Based on the fault analysis index value, the formula is used:

[0215] Build a similarity matrix and calculate device status and The similarity between ;

[0216] in, and The device status and Fault analysis indicators, is the standard deviation parameter, which adjusts the similarity sensitivity. () represents the exponential function, is the offset in the exponential function used to adjust the baseline of the similarity.

[0217] In device status and The similarity between Based on this, the formula is used:

[0218] Apply local density analysis to each state to obtain the device state The local density ;

[0219] in, is the similarity threshold, which is used to determine which states are similar to the state is close enough to be included in the local density calculation, is the indicator function, when the condition The value is 1 when it is established, otherwise it is 0, used for screening and status Other states that are close enough, is an index used to identify different device states and traverse all device states to calculate , is the total number of states, which indicates the number of all states that need to be considered when calculating the local density.

[0220] The calculation process is as follows:

[0221] For the formula: ;

[0222] Assume that there are three states of fault indicators, namely , and set and .

[0223] calculate :

[0224] Similarly, calculate and .

[0225] For the formula: ;

[0226] Continue to use the similarity matrix results, assuming there are similarity data for three states, and set the similarity threshold .

[0227] Assume the similarity matrix for:

[0228]

[0229] calculate :

[0230]

[0231] Similarly, calculate and .

[0232] Local Density The higher the value, the better the status. At a given threshold The higher the similarity with other states, the more likely the state is to be the center of a dense area. This indicates that state 1 has a higher local density relative to other states and may be the core area of ​​a cluster.

[0233] See also Figure 8 , the specific steps for obtaining the results of the electric furnace fault location analysis are:

[0234] Based on device status The local density , using the formula:

[0235] Calculate the local density of all device states The average , assists in identifying abnormal data points;

[0236] in, is the total number of states analyzed, is the device status;

[0237] Based on the local density of all device states The average , using the formula:

[0238]

[0239] Defining anomaly thresholds ;

[0240] in, is a coefficient less than 1, used to adjust the sensitivity, usually with a value such as The state marked as less than half of the average density is abnormal;

[0241] Based on the abnormal threshold , using the formula:

[0242]

[0243] Calculate the set of states marked as abnormal , get the electric furnace fault location analysis results;

[0244] in, Is the device status.

[0245] The calculation process is as follows:

[0246] For the formula: ;

[0247] Assume that there are five states with local density values: ,

[0248]

[0249]

[0250] For the formula: ;

[0251] Continue to use ,set up ; ;

[0252] For the formula: ;

[0253] Use the previous local density value and ;

[0254] Abnormal state identification: ;

[0255] because and Both lower than .

[0256] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A method for diagnosing faults of short-process electric furnace steelmaking equipment, characterized in that: The following steps are involved: Collect multiple parameters such as furnace body temperature, furnace chamber pressure, and electrode current from electric furnace steelmaking equipment, perform time series analysis on the operation data of the electric furnace, extract the time interval and sequence order, and obtain time series structure data; The steps for obtaining the time series structure data are specifically as follows: Collect multiple parameters of the electric furnace steelmaking equipment, including furnace body temperature T, furnace chamber pressure P and electrode current I, set the initial data array D = [T, P, I], and use the formula: Calculate the parameter changes at consecutive time points to obtain the time interval data V t ; Where D represents the initial data array, a, b and k are the coefficients of time series analysis, λ is the decay factor, t represents the time index, e represents the natural logarithm base, and D t-1 and D t-2 represent the data of the previous time point and the previous two time points, respectively; Based on the time interval data V t , according to the formula: Perform a weighted logarithmic transformation on the sequence data to generate the sequential data S of the time series t ; Among them, w i is the weight coefficient, log represents the natural logarithm, T i is the data value of the furnace temperature at the i-th time point, |T i | indicates T i The absolute value of +1 is used to avoid T i When it is 0, the logarithm is undefined, i is the index in the sequence, indicating that the operation is performed on all data points, and n is the total number of data points; Combined with the time interval data V t and the sequential data S of the time series t , using the formula: Get the time series structure data R; Among them, c and d represent the adjustment coefficients combining the time interval and the sequence order, and the g adjustment factor is used to adjust the nonlinear characteristics of the response. is an exponential function used to adjust V t and S t The degree of influence on R; Based on the time series structure data, an embedding dimension and a time delay parameter are selected, and the electric furnace operation data is spatially mapped to obtain a phase space trajectory diagram; The steps of acquiring the embedding dimension and the time delay parameter are specifically as follows: Based on the time series structure data R, the formula is adopted: and Calculate the weighted mean μ and weighted standard deviation σ of the time series structure data R; Through the weighted mean μ and weighted standard deviation σ of the time series structure data R, according to the formula: Get an estimate of the embedding dimension m; Among them, R i represents the i-th data point in the time series structure data R, is the data point R i The weight of , n is the total number of data points, log represents the natural logarithm, i is the index in the sequence, which is used to index the elements in the time series structure data R, and e represents the natural index; According to the estimated value of the embedding dimension m and the characteristics of the time series structure data R, the formula is adopted: Calculate the autocorrelation function ACF(k); Where k represents the value of time delay, w' i is used to weight R i and R i+k The coefficient of correlation, n is the total number of data points, i is the index in the sequence; Find the first zero crossing point k0 in the autocorrelation function ACF(k) using the formula: Get the value of the time delay parameter τ; Where ACF(k) is the autocorrelation function, k0 is the minimum k value that satisfies ACF(k0)≈0, τ is the time delay parameter, and i is the index in the sequence; The steps for obtaining the phase space trajectory diagram are specifically as follows: Using the estimated value of the embedding dimension m and the value of the time delay parameter τ, the formula is adopted: X(t)=[R t ,R t+τ ,R t+2τ ,...,R t+(m-1)τ ] Reconstruct the time series structure data R to form a state vector X(t); Among them, X(t) is the state vector, R t represents the data point of the time series at time t, τ is the time delay parameter, m is the embedding dimension, and t represents the time series index; Based on the state vector X(t), the formula is adopted: Perform spatial mapping to obtain the mapped state vector Y(t); Among them, C ib is the mapping coefficient of the ibth dimension, which is used to adjust the contribution of each dimension in the mapping; Based on the mapped state vector Y(t), the formula is adopted; Perform normalization processing to obtain the processed state vector Z(t) and generate a phase space trajectory diagram; Among them, min(Y) and max(Y) are the minimum and maximum values ​​in the Y(t) data sequence respectively; Performing dynamic analysis on the phase space trajectory diagram, calculating the trajectory divergence of furnace temperature, pressure and current state points, identifying chaotic behavior and fluctuation areas of steelmaking equipment through quantitative analysis of the divergence results, and generating a fault analysis index value; The steps for obtaining the trajectory divergence of the furnace temperature, pressure and current state points are specifically as follows: The time series of furnace temperature T(t), pressure P(t) and current I(t) are extracted from the processed state vector Z(t) data respectively, using the formula: ΔT(t)=T(t+1)-T(t) and ΔP(t)=P(t+1)-P(t) and ΔI(t)=I(t+1)-I(t) Calculate the first-order difference of the time series to obtain the first-order difference of the furnace temperature ΔT(t), the first-order difference of the pressure ΔP(t) and the first-order difference of the current ΔI(t); Wherein, T(t) and T(t+1) represent the furnace temperature at time t and t+1, respectively; P(t) and P(t+1) represent the pressure at time t and t+1, respectively; I(t) and I(t+1) represent the current at time t and t+1, respectively; The first-order difference of the furnace temperature time series ΔT(t), the first-order difference of the pressure time series ΔP(t) and the first-order difference of the current time series ΔI(t) are converted into absolute values ​​using the formula: and and The absolute values ​​are accumulated to obtain the total divergence value D of the furnace temperature. T , the total divergence value of pressure D P , the total divergence value of the current D I ; Where na is the length of the time series, indicating the total number of time points in the sequence, ∑ is the summation symbol, used to accumulate the divergence value of each time point, and t is the time point index in the sequence, used to iterate each time point; The steps for obtaining the fault analysis index value are specifically as follows: The total dispersion value D of the furnace temperature T , the total divergence value of pressure D P , the total divergence value of the current D I Compared with the sum of furnace temperature, pressure and current divergence values, the formula is: and and Calculate the ratio of a single total divergence value to the sum of furnace temperature, pressure, and current divergence values ​​to obtain the ratio value R' of furnace temperature to total divergence T , the ratio of pressure to total divergence R' P The ratio of the current to the total divergence R' I ; Among them, α is the constant in the denominator; Based on the ratio of the furnace temperature to the total scatter R' T , the ratio of pressure to total divergence R' P The ratio of the current to the total divergence R' I , using the formula: Calculate the mean dispersion ratio CI to quantify overall volatility and potential instability; Based on the average dispersion ratio CI, the formula is used: Calculate the fault index FI based on the maximum volatility parameter to obtain the fault analysis index value; Based on the fault analysis index value, the equipment operation status is divided by spectral clustering, local density analysis is applied to each status, and abnormal data points are selected according to local density comparison to obtain the electric furnace fault location analysis result.

2. The short-process electric furnace steelmaking equipment fault diagnosis method according to claim 1, characterized in that: The steps of dividing the equipment operation states and applying local density analysis to each state are specifically as follows: Based on the fault analysis index value, the formula is used: Construct a similarity matrix and calculate the similarity S between device states i′ and j i′j ; Among them, FI i′ and FI j are the fault analysis indicators of equipment states i′ and j, respectively, σ is the standard deviation parameter, exp() represents the exponential function, β is the offset in the exponential function, and j is the index used to identify the equipment state; The similarity S between the device states i′ and j i′j Based on this, the formula is used: Apply local density analysis to each state and obtain the local density ρ of device state i′ i c; Among them, θ is the similarity threshold, is an indicator function, which takes the value of 1 when the condition x>θ holds, otherwise it takes the value of 0. j is an index used to identify the device state, and N is the total number of states.

3. The short-process electric furnace steelmaking equipment fault diagnosis method according to claim 2 is characterized in that: The steps for obtaining the electric furnace fault location analysis result are specifically as follows: Based on the local density ρ of the device state i i c, using the formula: Calculate the local density ρ of all device states ic The average value of avg , assists in identifying abnormal data points; Where nb is the total number of analyzed states, i′ is the device state; Based on the local density ρ of all device states ic The average value of avg , using the formula: Threshold=ρ avg ×δ Define the abnormal threshold Threshold; Among them, δ is a coefficient less than 1, which is used to adjust the sensitivity; According to the abnormal threshold Threshold, the formula is adopted: Fault_Locations={i′|ρ ic <Threshold} Calculate the state set Fault_Locations marked as abnormal and obtain the fault location analysis result of the electric furnace; Where i' is the device state.

Images