Data processing method applied to multiple scenes

Abstract

The invention provides a data processing method applied to multiple scenes, which aims at abnormal values, high-frequency noise, white noise and other main noise influencing the accuracy of flight test data, detects and recognizes the noise types existing in various parameters of the flight test data based on a reasonable sequence, and improves the accuracy of the flight test data according to the detection and recognition results. And in combination with the curve characteristics of the parameters, the noise is automatically reduced or eliminated by using a corresponding method, so that the real state of the flight test data can be quickly and accurately restored. According to the invention, the working efficiency is greatly improved, the requirement of a modern flight test for rapidly obtaining a data processing result can be met, and certain expansibility is achieved.

Description

Technical Field

[0001] The invention relates to the field of data processing, in particular to a method for processing flight test data. Background Art

[0002] Flight test is a system engineering that integrates science, engineering and risk. It is an essential means to explore new concepts and verify new technologies. Flight test runs through the entire process of aviation basic research, applied technology research, advanced technology development, model engineering development and user verification. Flight test data is of great significance for evaluating aircraft performance, verifying design indicators, discovering potential problems and improving flight control systems. However, the noise of flight test data is an inevitable problem in flight tests. It comes from many aspects, such as body vibration, flight dynamic swing, sensor noise, environmental factors, etc. These noises will affect the accuracy and reliability of flight test data. Therefore, before using flight test data, certain methods and techniques need to be adopted to reduce or eliminate noise.

[0003] At present, the processing of flight test data is mostly done manually. Many flight test engineers and scientific researchers need to analyze and process these data at the same time. According to the parameter characteristics and application scenarios, combined with the experience of engineers, they choose to process different parameters using different methods. Summary of the invention

[0004] In order to overcome the deficiencies of the prior art, the present invention provides a data processing method applicable to multiple scenarios.

[0005] With the continuous improvement of the technical level of aviation weapons and equipment, the new aircraft used as flight test objects are becoming more and more complex, and the degree of integration and informatization is becoming higher and higher, which has led to a sharp increase in flight test data. The onboard acquisition parameters of a test aircraft reach tens of thousands, and the acquisition scenarios and application scenarios of different parameters are different. The amount of modern flight test data is getting larger and larger, and the scenarios are becoming more and more complex, but the processing cycle is getting shorter and shorter. Traditional flight test data processing methods and technologies can no longer meet the needs of modern flight tests to quickly obtain data processing results.

[0006] A multi-scenario flight test data processing method is proposed. This method detects and identifies the noise types existing in various parameters of the flight test data based on a reasonable order, targeting the main noises that affect the accuracy of the flight test data, such as outliers, high-frequency noise, white noise (random noise), etc. According to the detection and identification results and combined with the curve characteristics of the parameters, the corresponding method is automatically used to reduce or eliminate the noise, which can restore the true state of the flight test data very quickly and accurately.

[0007] The technical solution adopted by the present invention to solve the technical problem comprises the following steps:

[0008] Step 1: Select the raw flight test data to be processed;

[0009] Step 2: Read the parameter line of the original flight test data;

[0010] Step 3: Split the parameter line by the separator to obtain the parameter array and the number of parameters N. The separator of the text data is a comma, tab or space.

[0011] Step 4: Identify the time parameters and non-time parameters in the parameter array;

[0012] By using the names of preset time parameters, determine which column parameters belong to time parameters, and other column parameters are non-time parameters; time parameters do not need to be processed, but non-time parameters need to be processed; the names of time parameters include TIME, time, and Index;

[0013] Step 5: Read the flight test data in the column where the time parameter is located;

[0014] Step 6: Calculate the sampling rate using a sampling rate calculation algorithm to obtain the sampling rate of the current flight test data;

[0015] Step 7: Read the flight test data in the column where the nth column of non-time parameters is located, and record it as parameter n;

[0016] Step 8: Use the data processing algorithm applied to multiple scenarios to process the parameter n;

[0017] Step 9: Get the processed parameter n;

[0018] Step 10: Determine whether the parameter n is less than N; if the parameter n is less than N, return to step 7; if the parameter n is not less than N, proceed to step 11;

[0019] Step 11: The flight test data processing is completed, and the processed flight test data of all parameters have been obtained.

[0020] Further defined, the sampling rate calculation algorithm in step 6 comprises the following steps:

[0021] Step 6.1: Get the time column data;

[0022] Step 6.2: Identify the time column data format. The specific steps are as follows:

[0023] Step 6.2.1: Use colon ":" as the separator to split the time column data and get the split array;

[0024] Step 6.2.2: Get the array dimensions;

[0025] Step 6.2.3: Determine the format of the time column data. If the array dimension is 1, it means that the format of the time column data is a one-segment format, which only contains decimals, such as 169.25; if the array dimension is 3, it means that the format of the time column data is a three-segment format, which contains hours, minutes, and seconds, such as 8:20:56.028; if the array dimension is 4, it means that the format of the time column data is a four-segment format, which contains hours, minutes, seconds, and milliseconds, such as 10:04:19:728;

[0026] Step 6.3: Take the first and eleventh rows of data in the time column;

[0027] Step 6.4: Convert the first row of time column data into a period format, record it as start, and convert the eleventh row of time column data into a period format, record it as end. The conversion method is:

[0028] Step 6.4.1: For a format, no conversion is required;

[0029] Step 6.4.2: For the three-segment format, the conversion formula is: hours*3600+minutes*60+seconds

[0030] Step 6.4.3: For the four-segment format, the conversion formula is: hour*3600+minute*60+second+millisecond / 1000;

[0031] Step 6.5: Calculate the sampling rate: sampling rate Fs0 = (end-start) / 10;

[0032] Step 6.6: Matching sampling rate: Common sampling rates are built into the algorithm to form a sampling rate confidence array. Subtract Fs0 from the sampling rates in the sampling rate confidence array one by one, and take the absolute value of the result;

[0033] Step 6.7: Determine the sampling rate: The sampling rate with the smallest absolute value is the sampling rate Fs of the flight test data.

[0034] Further defined, the data processing algorithm for multiple scenarios in step 8 comprises the following steps:

[0035] Step 8.1: Obtain the original flight test data parameter n;

[0036] Step 8.2: Use the switch quantity detection method to detect whether the parameter n is a switch quantity;

[0037] Step 8.3: If the parameter n is a switch quantity, go to step 8.4; if it is not a switch quantity, go to step 8.5;

[0038] Step 8.4: Perform switch processing on parameter n by rounding each data. After the processing is completed, proceed to step 8.14;

[0039] Step 8.5: Outlier identification: Use the outlier identification algorithm to determine whether there are outliers in the parameter n;

[0040] Step 8.6: If there is an outlier in parameter n, go to step 8.7; if there is no outlier, go to step 8.8;

[0041] Step 8.7: Outlier elimination and correction: Eliminate the data belonging to outliers, and use Lagrange or other interpolation methods to correct the data at the outliers. After the processing is completed, proceed to step 8.8;

[0042] Step 8.8: Spectrum analysis: Sampling spectrum analysis algorithm, performing spectrum analysis on parameter n;

[0043] Step 8.9: Determine whether the parameter n needs filtering. If yes, proceed to step 8.10. If no, proceed to step 8.11.

[0044] Step 8.10: Low-pass filtering: low-pass filtering is used to process the parameter n to remove high-frequency noise. After the processing is completed, step 8.11 is entered;

[0045] Step 8.11: Smoothness calculation: Use the smoothness calculation algorithm to judge the smoothness of parameter n;

[0046] Step 8.12 determines whether the parameter n needs to be smoothed. If it needs to be smoothed, go to step 8.13. If not, go to step 8.14.

[0047] Step 8.13: Data smoothing:

[0048] Step 8.14: The processing is completed and the processed parameter n is obtained.

[0049] It is further defined that the switch quantity detection algorithm of step 8.2 of the above-mentioned data processing algorithm for multiple scenarios comprises the following steps:

[0050] Step 8.2.1: Merge the data with the same value of parameter n, that is, only keep one repeated value; the number of data of parameter n will change after the processing is completed; taking (1, 1, 2, 2, 3, 3, 4, 4) as an example, before the processing, there are 8 data of parameter n, and after the processing, parameter n becomes (1, 2, 3, 4), only 4 data;

[0051] Step 8.2.2: Calculate the number of data of parameter n after processing in step 8.2.1. If the number of data is less than 10, parameter n is considered to be a switch quantity. If the number of data is more than 10, parameter n is considered not to be a switch quantity.

[0052] It is further defined that the outlier identification algorithm in step 8.5 of the above-mentioned data processing algorithm for multiple scenarios comprises the following steps:

[0053] Step 8.5.1: Determine whether the i-th data yi is an outlier. The value of i is 3 to N, where N is the number of rows of the data to be determined.

[0054] Step 8.5.2: Calculate the mean y of the inth to the i-th data, where n ranges from 3 to 50;

[0055] Step 8.5.3: Calculate the standard deviation b of the in-th to i-th data;

[0056] Step 8.5.4: Calculate the residual Ci of the i-th data;

[0057] Step 8.5.5: Calculate the fitting value Yi of the ith data using the 7-point second-order algorithm forward difference method;

[0058] Step 8.5.6: Determine the Chauvelet criterion coefficient wn according to the value of n. wn is obtained by looking up the Zc coefficient table of the Chauvelet criterion.

[0059] Step 8.5.7: Determine whether the residual Ci is greater than wn×b×1.8. If so, proceed to step 8.5.8. If not, proceed to step 8.5.14.

[0060] Step 8.5.8: Determine whether Ci is greater than wn×b×3. If so, proceed to step 8.5.12. If not, proceed to step 8.5.9.

[0061] Step 8.5.9: Determine whether |yi / Yi| is greater than 2.2 or less than 0.45. If so, proceed to step 8.5.10. If not, proceed to step 8.5.11.

[0062] Step 8.5.10 determines whether |yi|+|Yi| is greater than 1. If so, proceed to step 8.5.12. If not, proceed to step 8.5.14.

[0063] Step 8.5.11 determines whether 2.2 / n*Yi is less than b. If so, proceed to step 8.5.12. If not, proceed to step 8.5.14.

[0064] Step 8.5.12: The i-th data yi is an outlier and needs to be processed, so go to step 8.5.13;

[0065] Step 8.5.13: Use Lagrange interpolation to correct the i-th data yi;

[0066] Step 8.5.14 The i-th data yi is not an outlier and does not need to be processed.

[0067] It is further defined that the spectrum analysis algorithm in step 8.8 of the above-mentioned data processing algorithm for multiple scenarios comprises the following steps:

[0068] Step 8.8.1: Perform Fourier transform on the parameter data;

[0069] Step 8.8.2: Calculate the energy spectral density;

[0070] Step 8.8.3: Calculate the sum P0 of the energy spectrum density P(i) of all frequencies Fs(i) except the DC component, and calculate the proportion of the energy spectrum density of each frequency in the total energy spectrum density p(i) = P(i) / P0;

[0071] Step 8.8.4: Accumulate and sum the p(i) before the i-th frequency that contains Fs(i), and let P be P+p(i);

[0072] Step 8.8.5: Determine whether P is greater than 90%. If so, proceed to step 8.8.6. If not, add 1 to i and return to step 8.8.4.

[0073] Step 8.8.6: Record the low-pass cutoff frequency as Fs(i) and proceed to step 8.8.7;

[0074] Step 8.8.7: Determine whether 2*Fs(i) / Fs is greater than the first 30%, where Fs is the sampling rate of the data. If so, proceed to step 8.8.8; if not, proceed to step 8.8.9;

[0075] Step 8.8.8: The data needs to be filtered;

[0076] Step 8.8.9: This data does not need to be filtered.

[0077] It is further defined that the smoothness calculation algorithm in step 8.11 of the above-mentioned data processing algorithm for multiple scenarios comprises the following steps:

[0078] Step 8.11.1: Use the mean method (one of the most commonly used algorithms for data smoothing) and take the window = sampling rate / 5 to smooth the parameters and record the results after the mean method processing;

[0079] Step 8.11.2: Calculate the relative change Δi of each row of data before and after processing, that is, Δi = |(processed i - original i) / original|;

[0080] Step 8.11.3: Calculate the average relative change A of the parameter data before and after processing, that is, add up the relative changes of all rows of data and divide by the number of data rows N, A = (Δ1 + Δ2 + ... + ΔN) / N;

[0081] Step 8.11.4: Calculate the standard deviation B1 of the original data;

[0082] Step 8.11.5: Calculate the standard deviation B2 of the processed data;

[0083] Step 8.11.6: Calculate the relative change B of the standard deviation of the data before and after processing, that is, B = |(B2-B1) / B1|;

[0084] Step 8.11.7: Determine whether A>p0 and B>q0, p0, q0 are smoothing indicators 1. If so, proceed to step 8.11.10. If not, proceed to step 8.11.8.

[0085] Step 8.11.8: Determine whether A>p1 or B>q1, where p1 and q1 are smoothing indicators 2. If so, proceed to step 8.11.10; if not, proceed to step 8.11.9;

[0086] Step 8.11.9: This parameter does not need to be smoothed;

[0087] Step 8.11.10: Parameters need to be smoothed.

[0088] The smoothing indices p0, q0, p1 and q1 are determined by the following steps:

[0089] Step a: Calculate the data duration; data duration = number of data rows * sampling rate;

[0090] Step b: Determine whether the data duration is less than 50 seconds. If so, proceed to step d. If not, proceed to step c.

[0091] Step c: The values ​​of the smoothing index are p0=0.25, q0=0.15, p1=0.65, q1=0.60;

[0092] Step d: Determine whether the data duration is ≥ 50 seconds and < 100 seconds. If so, proceed to step e. If not, proceed to step f.

[0093] Step e: The values ​​of the smoothing index are p0=0.22, q0=0.13, p1=0.60, q1=0.55;

[0094] Step f: The values ​​of the smoothing index are p0=0.20, q0=0.10, p1=0.55, q1=0.50.

[0095] It is further defined that the data smoothing in step 8.13 of the above-mentioned data processing algorithm for multiple scenarios is processed by the mean method, and the window size = sampling rate / 5.

[0096] An electronic device comprises: one or more processors; a memory; and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, and the one or more programs are configured to execute the method as described above.

[0097] The beneficial effect of the present invention is that by using algorithms such as sampling rate calculation, switch quantity identification, wild value identification, spectrum analysis, and smoothness detection, it is possible to make decisions on whether there are switch quantities, wild values, high-frequency noise, and whether smoothing is required in the flight test data, and automatically process the relevant parameters based on the decision results, replacing the work originally completed manually with computer programs, greatly improving work efficiency, meeting the needs of modern flight tests to quickly obtain data processing results, and being able to cope with increasingly complex scenarios in modern flight tests. At the same time, the present invention avoids the subjectivity of manual work in data processing through the calculation and determination of objective indicators, including the subjectivity of the selection of processing methods, the subjectivity of whether processing is required, and the subjectivity of determining the processing window, greatly improving the reliability and uniformity of data processing, and the indicators therein can also be synchronously optimized based on future aircraft model iterations and acquisition system upgrades, and have certain scalability. BRIEF DESCRIPTION OF THE DRAWINGS

[0098] Figure 1 It is a flow chart of the overall implementation of the present invention.

[0099] Figure 2 This is a schematic diagram of the automatic sampling rate recognition program.

[0100] Figure 3 This is a flow chart for implementing the outlier identification of the present invention.

[0101] Figure 4 This is the data smoothing detection method of the present invention.

[0102] Figure 5 This is a method for determining a data smoothing index according to the present invention.

[0103] Figure 6 This is the original curve of a flight test data - implementation example 1.

[0104] Figure 7 This is a comparison curve before and after processing of a flight test data - implementation example 1.

[0105] Figure 8 This is the original curve of a flight test data - Implementation Example 2.

[0106] Fig. 9 This is a comparison curve before and after processing of a flight test data - implementation example 2.

[0107] Fig.10 This is the original curve of a flight test data - implementation example 3.

[0108] Fig.11 This is a comparison curve before and after processing of a flight test data - implementation example 3.

[0109] Fig.12 This is the original curve of a flight test data - Implementation Example 4.

[0110] Fig.13 This is a comparison curve before and after processing of a flight test data - implementation example 4.

[0111] Fig.14 This is the original curve of a flight test data - Implementation Example 4.

[0112] Fig.15 This is a comparison curve before and after processing of a flight test data - implementation example 4. DETAILED DESCRIPTION

[0113] The present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0114] The present invention provides a method for processing flight test data applied to multiple scenarios, such as Figure 1 As shown, the following steps are included:

[0115] Step 1: Select the raw flight test data to be processed;

[0116] Step 2: Read the parameter line of the original flight test data;

[0117] Step 3: Split the parameter line by the separator to obtain the parameter array and the number of parameters N. The separator of the text data is a comma, tab or space.

[0118] Step 4: Identify the time parameters and non-time parameters in the parameter array;

[0119] By using the names of preset time parameters, determine which column parameters belong to time parameters, and other column parameters are non-time parameters; time parameters do not need to be processed, but non-time parameters need to be processed; the names of time parameters include TIME, time, and Index;

[0120] Step 5: Read the flight test data in the column where the time parameter is located;

[0121] Step 6: Calculate the sampling rate using a sampling rate calculation algorithm to obtain the sampling rate of the current flight test data;

[0122] Step 7: Read the flight test data in the column where the nth column of non-time parameters is located, and record it as parameter n;

[0123] Step 8: Use the data processing algorithm applied to multiple scenarios to process the parameter n;

[0124] Step 9: Get the processed parameter n;

[0125] Step 10: Determine whether the parameter n is less than N; if the parameter n is less than N, return to step 7; if the parameter n is not less than N, proceed to step 11;

[0126] Step 11: The flight test data processing is completed, and the processed flight test data of all parameters have been obtained.

[0127] Further limiting, such as Figure 2 As shown, the sampling rate calculation algorithm in step 6 includes the following steps:

[0128] Step 6.1: Get the time column data;

[0129] Step 6.2: Identify the time column data format. The specific steps are as follows:

[0130] Step 6.2.1: Use colon ":" as the separator to split the time column data and get the split array;

[0131] Step 6.2.2: Get the array dimensions;

[0132] Step 6.2.3: Determine the format of the time column data. If the array dimension is 1, it means that the format of the time column data is a one-segment format, which only contains decimals, such as 169.25; if the array dimension is 3, it means that the format of the time column data is a three-segment format, which contains hours, minutes, and seconds, such as 8:20:56.028; if the array dimension is 4, it means that the format of the time column data is a four-segment format, which contains hours, minutes, seconds, and milliseconds, such as 10:04:19:728;

[0133] Step 6.3: Take the first and eleventh rows of data in the time column;

[0134] Step 6.4: Convert the first row of time column data into a period format, record it as start, and convert the eleventh row of time column data into a period format, record it as end. The conversion method is:

[0135] Step 6.4.1: For a format, no conversion is required;

[0136] Step 6.4.2: For the three-segment format, the conversion formula is: hours*3600+minutes*60+seconds

[0137] Step 6.4.3: For the four-segment format, the conversion formula is: hour*3600+minute*60+second+millisecond / 1000;

[0138] Step 6.5: Calculate the sampling rate: sampling rate Fs0 = (end-start) / 10;

[0139] Step 6.6: Matching sampling rate: Common sampling rates are built into the algorithm to form a sampling rate confidence array. Subtract Fs0 from the sampling rates in the sampling rate confidence array one by one, and take the absolute value of the result;

[0140] Step 6.7: Determine the sampling rate: The sampling rate with the smallest absolute value is the sampling rate Fs of the flight test data.

[0141] Further limiting, such as Figure 3 As shown, the data processing algorithm for multiple scenarios in step 8 includes the following steps:

[0142] Step 8.1: Obtain the original flight test data parameter n;

[0143] Step 8.2: Use the switch quantity detection method to detect whether the parameter n is a switch quantity;

[0144] Step 8.3: If the parameter n is a switch quantity, go to step 8.4; if it is not a switch quantity, go to step 8.5;

[0145] Step 8.4: Perform switch processing on parameter n by rounding each data. After the processing is completed, proceed to step 8.14;

[0146] Step 8.5: Outlier identification: Use the outlier identification algorithm to determine whether there are outliers in the parameter n;

[0147] Step 8.6: If there is an outlier in parameter n, go to step 8.7; if there is no outlier, go to step 8.8;

[0148] Step 8.7: Outlier elimination and correction: Eliminate the data belonging to outliers, and use Lagrange or other interpolation methods to correct the data at the outliers. After the processing is completed, proceed to step 8.8;

[0149] Step 8.8: Spectrum analysis: Sampling spectrum analysis algorithm, performing spectrum analysis on parameter n;

[0150] Step 8.9: Determine whether the parameter n needs filtering. If yes, proceed to step 8.10. If no, proceed to step 8.11.

[0151] Step 8.10: Low-pass filtering: low-pass filtering is used to process the parameter n to remove high-frequency noise. After the processing is completed, step 8.11 is entered;

[0152] Step 8.11: Smoothness calculation: Use the smoothness calculation algorithm to judge the smoothness of parameter n;

[0153] Step 8.12 determines whether the parameter n needs to be smoothed. If it needs to be smoothed, go to step 8.13. If not, go to step 8.14.

[0154] Step 8.13: Data smoothing:

[0155] Step 8.14: The processing is completed and the processed parameter n is obtained.

[0156] It is further defined that the switch quantity detection algorithm of step 8.2 of the above-mentioned data processing algorithm for multiple scenarios comprises the following steps:

[0157] Step 8.2.1: Merge the data with the same value of parameter n, that is, only keep one repeated value; the number of data of parameter n will change after the processing is completed; taking (1, 1, 2, 2, 3, 3, 4, 4) as an example, before the processing, there are 8 data of parameter n, and after the processing, parameter n becomes (1, 2, 3, 4), only 4 data;

[0158] Step 8.2.2: Calculate the number of data of parameter n after processing in step 8.2.1. If the number of data is less than 10, parameter n is considered to be a switch quantity. If the number of data is more than 10, parameter n is considered not to be a switch quantity.

[0159] Further limiting, such as Figure 4 As shown, the outlier identification algorithm in step 8.5 of the above-mentioned data processing algorithm for multiple scenarios includes the following steps:

[0160] Step 8.5.1: Determine whether the i-th data yi is an outlier. The value of i is 3 to N, where N is the number of rows of the data to be determined.

[0161] Step 8.5.2: Calculate the mean y of the inth to the i-th data, where n ranges from 3 to 50;

[0162] Step 8.5.3: Calculate the standard deviation b of the in-th to i-th data;

[0163] Step 8.5.4: Calculate the residual Ci of the i-th data;

[0164] Step 8.5.5: Calculate the fitting value Yi of the ith data using the 7-point second-order algorithm forward difference method;

[0165] Step 8.5.6: Determine the Chauvelet criterion coefficient wn according to the value of n. wn is obtained by looking up the Zc coefficient table of the Chauvelet criterion.

[0166] Step 8.5.7: Determine whether the residual Ci is greater than wn×b×1.8. If so, proceed to step 8.5.8. If not, proceed to step 8.5.14.

[0167] Step 8.5.8: Determine whether Ci is greater than wn×b×3. If so, proceed to step 8.5.12. If not, proceed to step 8.5.9.

[0168] Step 8.5.9: Determine whether |yi / Yi| is greater than 2.2 or less than 0.45. If so, proceed to step 8.5.10. If not, proceed to step 8.5.11.

[0169] Step 8.5.10 determines whether |yi|+|Yi| is greater than 1. If so, proceed to step 8.5.12. If not, proceed to step 8.5.14.

[0170] Step 8.5.11 determines whether 2.2 / n*Yi is less than b. If so, proceed to step 8.5.12. If not, proceed to step 8.5.14.

[0171] Step 8.5.12: The i-th data yi is an outlier and needs to be processed, so go to step 8.5.13;

[0172] Step 8.5.13: Use Lagrange interpolation to correct the i-th data yi;

[0173] Step 8.5.14 The i-th data yi is not an outlier and does not need to be processed.

[0174] Further limiting, such as Figure 5 As shown, the spectrum analysis algorithm in step 8.8 of the above-mentioned data processing algorithm for multiple scenarios includes the following steps:

[0175] Step 8.8.1: Perform Fourier transform on the parameter data;

[0176] Step 8.8.2: Calculate the energy spectral density;

[0177] Step 8.8.3: Calculate the sum P0 of the energy spectrum density P(i) of all frequencies Fs(i) except the DC component, and calculate the proportion of the energy spectrum density of each frequency in the total energy spectrum density p(i) = P(i) / P0;

[0178] Step 8.8.4: Accumulate and sum the p(i) before the i-th frequency that contains Fs(i), and let P be P+p(i);

[0179] Step 8.8.5: Determine whether P is greater than 90%. If so, proceed to step 8.8.6. If not, add 1 to i and return to step 8.8.4.

[0180] Step 8.8.6: Record the low-pass cutoff frequency as Fs(i) and proceed to step 8.8.7;

[0181] Step 8.8.7: Determine whether 2*Fs(i) / Fs is greater than the first 30%, where Fs is the sampling rate of the data. If so, proceed to step 8.8.8; if not, proceed to step 8.8.9;

[0182] Step 8.8.8: The data needs to be filtered;

[0183] Step 8.8.9: This data does not need to be filtered.

[0184] Further limiting, such as Figure 6 As shown, the smoothness calculation algorithm in step 8.11 of the above-mentioned data processing algorithm for multiple scenes includes the following steps:

[0185] Step 8.11.1: Use the mean method (one of the most commonly used algorithms for data smoothing) and take the window = sampling rate / 5 to smooth the parameters and record the results after the mean method processing;

[0186] Step 8.11.2: Calculate the relative change Δi of each row of data before and after processing, that is, Δi = |(processed i - original i) / original|;

[0187] Step 8.11.3: Calculate the average relative change A of the parameter data before and after processing, that is, add up the relative changes of all rows of data and divide by the number of data rows N, A = (Δ1 + Δ2 + ... + ΔN) / N;

[0188] Step 8.11.4: Calculate the standard deviation B1 of the original data;

[0189] Step 8.11.5: Calculate the standard deviation B2 of the processed data;

[0190] Step 8.11.6: Calculate the relative change B of the standard deviation of the data before and after processing, that is, B = |(B2-B1) / B1|;

[0191] Step 8.11.7: Determine whether A>p0 and B>q0, p0, q0 are smoothing indicators 1. If so, proceed to step 8.11.10. If not, proceed to step 8.11.8.

[0192] Step 8.11.8: Determine whether A>p1 or B>q1, where p1 and q1 are smoothing indicators 2. If so, proceed to step 8.11.10; if not, proceed to step 8.11.9;

[0193] Step 8.11.9: This parameter does not need to be smoothed;

[0194] Step 8.11.10: Parameters need to be smoothed.

[0195] The smoothing indices p0, q0, p1 and q1 are as follows: Figure 7 As shown, it is determined by the following steps:

[0196] Step a: Calculate the data duration; data duration = number of data rows * sampling rate;

[0197] Step b: Determine whether the data duration is less than 50 seconds. If so, proceed to step d. If not, proceed to step c.

[0198] Step c: The values ​​of the smoothing index are p0=0.25, q0=0.15, p1=0.65, q1=0.60;

[0199] Step d: Determine whether the data duration is ≥ 50 seconds and < 100 seconds. If so, proceed to step e. If not, proceed to step f.

[0200] Step e: The values ​​of the smoothing index are p0=0.22, q0=0.13, p1=0.60, q1=0.55;

[0201] Step f: The values ​​of the smoothing index are p0=0.20, q0=0.10, p1=0.55, q1=0.50.

[0202] It is further defined that the data smoothing in step 8.13 of the above-mentioned data processing algorithm for multiple scenarios is processed by the mean method, and the window size = sampling rate / 5.

[0203] The examples are as follows:

[0204] Example 1: The method of the present invention is used to process a flight test data. The method of the present invention detects that the parameter has an outlier, and automatically removes and corrects the outlier. At the same time, for the result after the outlier removal and correction, it is found that the smoothness is not enough, and the data is smoothed. The original curve is as follows: Figure 8 As shown, the processed curve is Fig. 9 shown.

[0205] Example 2: The method of the present invention is used to process a flight test data. The method of the present invention detects that the parameter has high-frequency noise, so it is low-pass filtered. At the same time, for the result after low-pass filtering, it is considered that its smoothness is not enough, so it is smoothed. The original curve is as follows: Fig.10 As shown, the processed curve is Fig.11 shown.

[0206] In Example 3, the method of the present invention is used to process a flight test data. The method of the present invention determines that the parameter does not belong to a switch value, does not have wild values, does not have high-frequency noise, and has good smoothness. Therefore, no processing is performed on the data. Therefore, the processed curve completely overlaps with the original curve. The original curve is as follows Fig.12 As shown, the processed curve is Fig.13 shown.

[0207] In Example 4, the method of the present invention is used to process a flight test data. The method of the present invention identifies that the parameter belongs to a switch value, so it is automatically processed into an integer. The original curve is as follows: Fig.14 As shown, the processed curve is Fig.15 shown.

Claims

1. A data processing method applied to multiple scenarios, characterized in that The steps include: Step 1: Select the raw flight test data to be processed; Step 2: Read the parameter line of the original flight test data; Step 3: Split the parameter row by the separator to obtain the parameter array and the number of parameters N. The separator of the text data is a comma, tab or space. Step 4: Identify the time parameters and non-time parameters in the parameter array; By using the names of preset time parameters, determine which column parameters belong to time parameters, and other column parameters are non-time parameters; time parameters do not need to be processed, but non-time parameters need to be processed; the names of time parameters include TIME, time, and Index; Step 5: Read the flight test data in the column where the time parameter is located; Step 6: Calculate the sampling rate using a sampling rate calculation algorithm to obtain the sampling rate of the current flight test data; Step 7: Read the flight test data in the column where the nth column of non-time parameters is located, and record it as parameter n; Step 8: Use the data processing algorithm applied to multiple scenarios to process the parameter n; Step 9: Get the processed parameter n; Step 10: Determine whether the parameter n is less than N; If the parameter n is less than N, return to step 7; if the parameter n is not less than N, proceed to step 11; Step 11: The flight test data processing is completed, and the processed flight test data of all parameters have been obtained.

2. The data processing method applied to multiple scenarios according to claim 1, characterized in that: The sampling rate calculation algorithm in step 6 contains the following steps: Step 6.1: Get the time column data; Step 6.2: Identify the time column data format. The specific steps are as follows: Step 6.2.1: Use colon ":" as the separator to split the time column data and get the split array; Step 6.2.2: Get the array dimensions; Step 6.2.3: Determine the format of the time column data. If the array dimension is 1, it means that the format of the time column data is a one-segment format, which only contains decimals; if the array dimension is 3, it means that the format of the time column data is a three-segment format, which contains hours, minutes, and seconds; if the array dimension is 4, it means that the format of the time column data is a four-segment format, which contains hours, minutes, seconds, and milliseconds; Step 6.3: Take the first and eleventh rows of data in the time column; Step 6.4: Convert the first row of time column data into a period format, record it as start, and convert the eleventh row of time column data into a period format, record it as end. The conversion method is: Step 6.4.1: For a format, no conversion is required; Step 6.4.2: For the three-segment format, the conversion formula is: hours*3600+minutes*60+seconds Step 6.4.3: For the four-segment format, the conversion formula is: hour*3600+minute*60+second+millisecond / 1000; Step 6.5: Calculate the sampling rate: sampling rate Fs0 = (end-start) / 10; Step 6.6: Matching sampling rate: Common sampling rates are built into the algorithm to form a sampling rate confidence array. Subtract Fs0 from the sampling rates in the sampling rate confidence array one by one, and take the absolute value of the result; Step 6.7: Determine the sampling rate: The sampling rate with the smallest absolute value is the sampling rate Fs of the flight test data.

3. The data processing method applied to multiple scenarios according to claim 1 is characterized in that: The data processing algorithm for multiple scenarios in step 8 includes the following steps: Step 8.1: Obtain the original flight test data parameter n; Step 8.2: Use the switch quantity detection method to detect whether the parameter n is a switch quantity; Step 8.3: If the parameter n is a switch quantity, go to step 8.4; if it is not a switch quantity, go to step 8.5; Step 8.4: Perform switch processing on parameter n by rounding each data. After the processing is completed, proceed to step 8.14; Step 8.5: Outlier identification: Use the outlier identification algorithm to determine whether there are outliers in the parameter n; Step 8.6: If there is an outlier in parameter n, go to step 8.7; if there is no outlier, go to step 8.8; Step 8.7: Outlier elimination and correction: Eliminate the data belonging to outliers, and use Lagrange or other interpolation methods to correct the data at the outliers. After the processing is completed, proceed to step 8.8; Step 8.8: Spectrum analysis: Sampling spectrum analysis algorithm, performing spectrum analysis on parameter n; Step 8.9: Determine whether the parameter n needs filtering. If yes, proceed to step 8.

10. If no, proceed to step 8.

11. Step 8.10: Low-pass filtering: low-pass filtering is used to process the parameter n to remove high-frequency noise. After the processing is completed, step 8.11 is entered; Step 8.11: Smoothness calculation: Use the smoothness calculation algorithm to judge the smoothness of parameter n; Step 8.12 determines whether the parameter n needs to be smoothed. If it needs to be smoothed, go to step 8.

13. If not, go to step 8.

14. Step 8.13: Data smoothing: Step 8.14: The processing is completed and the processed parameter n is obtained.

4. The data processing method applied to multiple scenarios according to claim 3 is characterized in that: The switch quantity detection algorithm of step 8.2 comprises the following steps: Step 8.2.1: Merge the data with the same value of parameter n, that is, only keep one repeated value; Step 8.2.2: Calculate the number of data of parameter n after processing in step 8.2.

1. If the number of data is less than 10, parameter n is considered to be a switch quantity. If the number of data is more than 10, parameter n is considered not to be a switch quantity.

5. The data processing method applied to multiple scenarios according to claim 3 is characterized in that: The outlier identification algorithm in step 8.5 comprises the following steps: Step 8.5.1: Determine whether the i-th data yi is an outlier. The value of i is 3 to N, where N is the number of rows of the data to be determined. Step 8.5.2: Calculate the mean y of the inth to the i-th data, where n ranges from 3 to 50; Step 8.5.3: Calculate the standard deviation b of the in-th to i-th data; Step 8.5.4: Calculate the residual Ci of the i-th data; Step 8.5.5: Calculate the fitting value Yi of the ith data using the 7-point second-order algorithm forward difference method; Step 8.5.6: Determine the Chauvelet criterion coefficient wn according to the value of n. wn is obtained by looking up the Zc coefficient table of the Chauvelet criterion. Step 8.5.7: Determine whether the residual Ci is greater than wn×b×1.

8. If so, proceed to step 8.5.

8. If not, proceed to step 8.5.

14. Step 8.5.8: Determine whether Ci is greater than wn×b×3. If so, proceed to step 8.5.

12. If not, proceed to step 8.5.

9. Step 8.5.9: Determine whether |yi / Yi| is greater than 2.2 or less than 0.

45. If so, proceed to step 8.5.

10. If not, proceed to step 8.5.

11. Step 8.5.10 determines whether |yi|+|Yi| is greater than 1. If so, proceed to step 8.5.

12. If not, proceed to step 8.5.

14. Step 8.5.11 determines whether 2.2 / n*Yi is less than b. If so, proceed to step 8.5.

12. If not, proceed to step 8.5.

14. Step 8.5.12: The i-th data yi is an outlier and needs to be processed, so go to step 8.5.13; Step 8.5.13: Use Lagrange interpolation to correct the i-th data yi; Step 8.5.14 The i-th data yi is not an outlier and does not need to be processed.

6. The data processing method applied to multiple scenarios according to claim 3 is characterized in that: The spectrum analysis algorithm in step 8.8 comprises the following steps: Step 8.8.1: Perform Fourier transform on the parameter data; Step 8.8.2: Calculate the energy spectral density; Step 8.8.3: Calculate the sum P0 of the energy spectrum density P(i) of all frequencies Fs(i) except the DC component, and calculate the proportion of the energy spectrum density of each frequency in the total energy spectrum density p(i) = P(i) / P0; Step 8.8.4: Accumulate and sum the p(i) before the i-th frequency that contains Fs(i), and let P be P+p(i); Step 8.8.5: Determine whether P is greater than 90%. If so, proceed to step 8.8.

6. If not, add 1 to i and return to step 8.8.

4. Step 8.8.6: Record the low-pass cutoff frequency as Fs(i) and proceed to step 8.8.7; Step 8.8.7: Determine whether 2*Fs(i) / Fs is greater than the first 30%, where Fs is the sampling rate of the data. If so, proceed to step 8.8.8; if not, proceed to step 8.8.9; Step 8.8.8: The data needs to be filtered; Step 8.8.9: This data does not need to be filtered.

7. The data processing method applied to multiple scenarios according to claim 3 is characterized in that: The smoothness calculation algorithm in step 8.11 comprises the following steps: Step 8.11.1: Use the mean method and take the window = sampling rate / 5 to smooth the parameters and record the results after the mean method processing; Step 8.11.2: Calculate the relative change Δi of each row of data before and after processing, that is, Δi = |(processed i - original i) / original|; Step 8.11.3: Calculate the average relative change A of the parameter data before and after processing, that is, add up the relative changes of all rows of data and divide by the number of data rows N, A = (Δ1 + Δ2 + ... + ΔN) / N; Step 8.11.4: Calculate the standard deviation B1 of the original data; Step 8.11.5: Calculate the standard deviation B2 of the processed data; Step 8.11.6: Calculate the relative change B of the standard deviation of the data before and after processing, that is, B = |(B2-B1) / B1|; Step 8.11.7: Determine whether A>p0 and B>q0, where p0 and q0 are smoothing indicators 1. If so, proceed to step 8.11.

10. If not, proceed to step 8.11.

8. 、 Step 8.11.8: Determine whether A>p1 or B>q1, where p1 and q1 are smoothing indicators 2. If so, proceed to step 8.11.10; if not, proceed to step 8.11.9; Step 8.11.9: This parameter does not need to be smoothed; Step 8.11.10: Parameters need to be smoothed.

8. The data processing method applied to multiple scenarios according to claim 7 is characterized in that: The smoothing indices p0, q0, p1 and q1 are determined by the following steps: Step a: Calculate the data duration; data duration = number of data rows * sampling rate; Step b: Determine whether the data duration is less than 50 seconds. If so, proceed to step d. If not, proceed to step c. Step c: The values ​​of the smoothing index are p0=0.25, q0=0.15, p1=0.65, q1=0.60; Step d: Determine whether the data duration is ≥ 50 seconds and < 100 seconds. If so, proceed to step e. If not, proceed to step f. Step e: The values ​​of the smoothing index are p0=0.22, q0=0.13, p1=0.60, q1=0.55; Step f: The values ​​of the smoothing index are p0=0.20, q0=0.10, p1=0.55, q1=0.

50.

9. The data processing method applied to multiple scenarios according to claim 3 is characterized in that: The data smoothing in step 8.13 is processed by the mean method, and the window size = sampling rate / 5.

10. An electronic device, characterized in that: include: one or more processors; Memory; One or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, and the one or more programs are configured to execute the method according to any one of claims 1 to 9.

Images