Engine fault detection method

Abstract

The present invention relates to the technical field of fault detection, and more specifically, to an engine fault detection method, comprising: obtaining the concentration sequence of the engine in each historical period, calculating the noise level of each concentration value in the concentration sequence, obtaining the noise distribution of the concentration sequence, wherein the noise level of the concentration value reflects the degree of deviation between the concentration value and the fitting value, and constructing a database based on the noise distribution of all concentration sequences. The present invention screens out the current standard sequence of the concentration sequence of the current period through the noise distribution of each concentration sequence in the database, removes the concentration sequence corresponding to the noise generated by the engine fault, and then selects a more accurate current standard sequence to improve the accuracy of calculating the abnormality degree. In addition, when calculating the fault degree of the engine, the changing trend of the temperature value and concentration value of the engine in the current period is taken into account, which further improves the accuracy of calculating the fault degree of the engine.

Description

Technical Field

[0001] The present invention relates to the technical field of fault detection, and more particularly to a method for detecting engine faults. Background Art

[0002] Natural gas engines are used in modern industrial production, especially in factory equipment that requires high power and high efficiency. As a clean and efficient energy source, natural gas has gradually replaced traditional oil and coal energy sources and become one of the important sources of power in the industrial field. In the production processes of many factories, equipment often operates cyclically, that is, the equipment needs to be started, stopped, and adjusted periodically according to production needs. For example, the natural gas engines used in production lines such as certain chemical plants, steel mills, and aluminum plants often need to start or stop at specific time intervals to meet production energy needs. This type of cyclically operating equipment often requires the ability to start quickly, output stably, and operate efficiently. Therefore, natural gas engines, with their fast response speed and high combustion efficiency, have become the preferred power system.

[0003] In the related art, for example, Chinese patent application document with publication number CN116241369A discloses an engine fault detection method, controller and engine fault detection device. When the engine operating condition is a first operating condition, a methane concentration value is obtained, and the first operating condition is the operating condition of normal engine operation; it is determined whether the methane concentration value is within a preset concentration range; when the methane concentration value is within the preset concentration range, it is determined that a ventilation fault has occurred in the engine.

[0004] Currently, methane concentration is collected by a methane sensor and compared with an abnormal threshold to determine whether an engine fault has occurred. However, in some cases, noise may be generated by the methane sensor during data collection or due to normal engine wear, resulting in inaccurate concentration values. Consequently, there is a problem of low accuracy when using concentration values ​​with noisy data to determine engine faults. Summary of the Invention

[0005] The present invention provides an engine fault detection method, which aims to solve the problem in the related art that methane sensors will generate noise during the data collection process or due to normal engine wear, resulting in inaccurate collected concentration values, and thus the problem of low accuracy when using concentration values ​​with noisy data to judge engine faults.

[0006] The present invention provides an engine fault detection method, comprising: obtaining a concentration sequence of an engine in each historical period, calculating the noise level of each concentration value in the concentration sequence, obtaining a noise distribution of the concentration sequence, wherein the noise level of the concentration value reflects the degree of deviation of the concentration value from a fitted value, and constructing a database based on the noise distribution of all concentration sequences; obtaining the abnormality level of each concentration value in the current period concentration sequence according to the difference between each concentration value in the current period concentration sequence and each concentration value in the current standard sequence, and performing weighted summation of each concentration value in the current period concentration sequence using the abnormality level and the noise level of each concentration value to obtain the engine fault level, and performing a fault detection on the engine according to the fault level. The engine is used for fault detection, wherein the degree of the fault also reflects the trend of changes in the engine's temperature and concentration values ​​within the current cycle; wherein the method for obtaining the current standard sequence includes: clustering the noise distribution of all concentration sequences in the database to obtain multiple first clusters, wherein one first cluster represents a noise distribution type; clustering all concentration sequences within any first cluster to obtain two second clusters; and calculating the average value of all concentration sequences in the second cluster containing the most concentration sequences as the historical standard sequence corresponding to the noise distribution type of the first cluster, and determining the current standard sequence based on the noise distribution type of the current cycle concentration sequence. Through multi-dimensional analysis of the engine's operating status, selection of standard sequences, and weighted evaluation of comprehensive anomalies and noise, accurate fault detection can be provided.

[0007] Furthermore, engine fault detection is performed based on the severity of the fault, including issuing an alarm to alert personnel if the fault severity exceeds an abnormality threshold. By setting an abnormality threshold, the system can promptly issue an alarm when the fault severity exceeds this threshold. This rapid response helps detect problems before major engine failure occurs, preventing further damage.

[0008] Furthermore, the noise distribution type of the current period concentration sequence is used to determine the current standard sequence of the noise distribution type, including: determining the noise distribution type of the current period concentration sequence, including: calculating the similarity between the current period concentration sequence and all historical standard sequences corresponding to the first cluster clusters, and selecting the historical standard sequence with the greatest similarity as the current standard sequence, wherein the similarity is the Euclidean distance between the current period concentration sequence and the historical standard sequence corresponding to the first cluster cluster. As a simple and intuitive distance metric, the Euclidean distance can clearly reflect the difference between the two sequences, ensuring that the selection of the standard sequence can truly reflect the data characteristics of the current period.

[0009] Further, calculate the degree of engine failure , the calculation formula is: Where, is the concentration sequence of the current cycle concentration values, is the concentration sequence of the current cycle The noise level of each concentration value, Indicates the concentration sequence of the current cycle The abnormality of the concentration value, Indicates the number of concentration values ​​in the concentration sequence of the current period, is the standard normalization function.

[0010] Furthermore, the degree of failure also reflects the change trend of the temperature value and concentration value of the engine in the current cycle. The calculation formula is: Where, Indicates the Pearson correlation coefficient between the temperature series of the current cycle and the concentration series of the current cycle, which is used to reflect the changing trend of the temperature and concentration values ​​of the engine in the current cycle. is the concentration sequence of the current cycle concentration values, is the concentration sequence of the current cycle The noise level of each concentration value, Indicates the concentration sequence of the current cycle The abnormality of the concentration value, Indicates the number of concentration values ​​in the concentration sequence of the current period, is the standard normalization function.

[0011] Furthermore, the abnormality of each concentration value in the concentration sequence of the current period is calculated using the following formula: Where, Indicates the number of the concentration sequence in the current engine cycle. The abnormality of the concentration value, Indicates the concentration sequence of the current cycle The concentration value, Indicates the current standard sequence The actual value of the concentration value, Represents the mean of all concentration values ​​in the concentration series of the current period, Indicates the number of concentration values ​​in the concentration sequence of the current period, Represents the standard normalization function. By comprehensively analyzing the difference between the current cycle concentration series and the standard concentration series, as well as the degree of fluctuation of the current cycle concentration series, normal data and abnormal data can be effectively distinguished.

[0012] Furthermore, the noise level of each concentration value in the concentration sequence is calculated, including: performing a least squares fit on the concentration values ​​in the concentration sequence to obtain a first fitted value for each concentration value; removing any concentration value from the concentration sequence and performing a least squares fit on the remaining concentration values ​​to obtain a second fitted value for each remaining concentration value; calculating the cumulative sum of differences between the measured values ​​and the first fitted value, and the cumulative sum of differences between the measured values ​​and the second fitted values ​​for the remaining concentration values ​​in the concentration sequence excluding any concentration value, calculating the difference between the two cumulative sums of differences, and using the normalized difference as the noise level of the concentration value. By comparing the difference between the two fitting errors (the error with and without the removed concentration value), the difference can be used to measure the impact of the concentration value on the overall fitting result, thereby obtaining the noise level of each concentration value.

[0013] Furthermore, the noise level of each concentration value in the concentration sequence is calculated using the following formula: Where, For the The noise level of each concentration value, is the concentration sequence except The concentration value outside The measured value of the concentration value, is the concentration sequence except The concentration value outside The first fitted value for the concentration value, is the concentration sequence except The concentration value outside The second fitted value of the concentration value, Indicates the number of concentration values ​​in the concentration series.

[0014] Furthermore, obtaining the concentration sequence of the engine in each historical period includes: collecting the methane concentration value in the engine once every preset time interval within the period to obtain a concentration sequence consisting of multiple concentration values.

[0015] Beneficial effects (1) By calculating the noise level and abnormality level of each concentration value in the current period concentration sequence, the noise level and abnormality level are used to weight the fault level of the calculated engine and improve the accuracy of the calculation results.

[0016] (2) Through the noise distribution of each concentration sequence in the database, the current standard sequence of the current period concentration sequence is screened out, the concentration sequence corresponding to the noise generated by the engine failure is eliminated, and then a more accurate current standard sequence is selected to improve the accuracy of the calculation of the abnormality degree.

[0017] (3) When calculating the degree of engine failure, the changing trend of the engine's temperature value and concentration value in the current cycle is taken into account, further improving the accuracy of calculating the degree of engine failure. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The following detailed description is read with reference to the accompanying drawings, which illustrate several embodiments of the present invention by way of example and not limitation, and in which like or corresponding reference numerals designate like or corresponding parts, wherein: Figure 1 is a flowchart schematically illustrating construction of a database according to an embodiment of the present invention; Figure 2 FIG. 1 is a flowchart schematically illustrating a method for calculating an engine failure degree according to an embodiment of the present invention. DETAILED DESCRIPTION

[0019] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work shall fall within the scope of protection of the present invention.

[0020] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0021] like Figure 1 As shown, S110: acquiring engine operating data.

[0022] Specifically, engine operating data is acquired, where the operating data may include methane concentration, temperature, or exhaust and intake pressures. In this embodiment, temperature and methane concentration are used as examples of operating data. Specifically, during the engine's operating cycle, a methane sensor is used to collect methane concentration within the engine at preset time intervals, generating a concentration sequence consisting of multiple concentration values ​​within the cycle. Similarly, a temperature sensor is used to collect engine temperature at preset time intervals, generating a temperature sequence consisting of multiple temperature values ​​within the cycle for subsequent use. The preset time interval may be, for example, 1 second or 2 seconds, and may be adjusted based on specific implementation circumstances.

[0023] S111: Calculate the noise level of each concentration value in each concentration sequence and build a database.

[0024] In one embodiment, based on past engine operating cycles, concentration sequences for each historical engine cycle are obtained. The noise level of each concentration value in all concentration sequences is calculated, thereby obtaining a noise distribution for each concentration sequence. The noise level of each concentration value reflects the degree of deviation between the concentration value and the fitted value. By comparing the difference between two fitting errors (the error with and without the concentration value removed), the difference can be used to measure the impact of the concentration value on the overall fitting result. A large difference indicates that removing the concentration value significantly alters the fitting result, indicating that the concentration value may be noise. Specifically, a least squares fit is performed on the concentration values ​​in the concentration sequence to obtain a first fitted value for each concentration value. Any concentration value is removed from the concentration sequence, and a least squares fit is performed on the remaining concentration values ​​to obtain a second fitted value for each remaining concentration value. The cumulative sum of the differences between the measured values ​​and the first fitted value and the cumulative sum of the differences between the measured values ​​and the second fitted value for the remaining concentration values ​​in the concentration sequence excluding the removed concentration value is calculated. The difference between the two cumulative sums is calculated, and the normalized difference is used as the noise level for that concentration value.

[0025] In one embodiment, the noise level of each concentration value in the concentration sequence is calculated, and a calculation method is provided as follows: Where, For the The noise level of each concentration value, Divide the concentration value by the concentration value in the concentration sequence Other than The measured value of the concentration value, Divide the concentration value by the concentration value in the concentration sequence Other than The first fitted value of the concentration value, Divide the concentration value by the concentration value in the concentration sequence Other than The second fitted value of the concentration value, represents the number of concentration values ​​in the concentration sequence, is the standard normalization function. At this point, the noise level of each concentration value in each concentration sequence can be obtained, the noise distribution of the concentration sequence can be obtained, and a database can be constructed based on the noise distribution of all concentration sequences.

[0026] like Figure 2 As shown, S121: Calculate the noise level of the current cycle concentration sequence and determine the current standard sequence.

[0027] Specifically, to calculate the degree of engine failure, it is necessary to obtain the engine's operating data for the current cycle, that is, the operating data within the latest engine operating cycle, and then obtain the current cycle concentration sequence and the current cycle temperature sequence, so as to calculate the engine's failure degree after the latest use of the engine.

[0028] In one embodiment, the noise level of the concentration sequence of the current period is calculated using the above method for calculating the noise level of each concentration value in the concentration sequence.

[0029] S122: Calculate the abnormality level of each concentration value in the concentration sequence of the current cycle.

[0030] Specifically, when the engine load and working environment remain unchanged, the difference between the engine's operating data and the standard data at the same time should be small. When the current operating data and the standard data at the same time are significantly different, it indicates that the current operating data may be abnormal, which also indicates that the engine is affected by a fault. Therefore, the degree of abnormality of each concentration value in the current period concentration sequence can be calculated based on the above characteristics.

[0031] It should be noted that the noise distribution of an engine during long-term use, when its internal parts wear out normally, will also vary from that of an engine failure. When selecting standard data, the concentration sequence corresponding to the noise distribution during normal wear should be selected as the current standard sequence for the concentration sequence of the current cycle.

[0032] Specifically, after obtaining a database of noise distributions for all concentration sequences, it is necessary to classify the noise distributions of all concentration sequences in the database to obtain multiple noise distribution types and obtain the historical standard sequences corresponding to each noise distribution type. The noise distribution type of the current period concentration sequence is then determined, and the current standard sequence corresponding to the current period concentration sequence is determined based on this.

[0033] In one embodiment, the noise distributions of all concentration sequences in the database are clustered to obtain multiple first clusters, each of which represents a noise distribution type. Within each noise distribution type, noise generated by normal wear of internal engine components and noise generated by engine failure are then distinguished. Specifically, all concentration sequences within any first cluster are clustered to obtain two second clusters. The average value of all concentration sequences within the second cluster containing the most concentration sequences is used as the historical standard sequence for the noise distribution type corresponding to the first cluster. This is because, during engine operation, the probability of engine failure is relatively low, while noise generated by engine wear is far greater than the probability of engine failure. Therefore, the second cluster containing the most concentration sequences is selected as the concentration sequence generated by normal wear, while the second cluster containing the fewest concentration sequences is selected as the concentration sequence generated by engine failure. Finally, the current standard sequence is determined based on the noise distribution type of the concentration sequence in the current period. The clustering method employs the K-means clustering algorithm.

[0034] For example, the noise distribution of all concentration sequences in the database is clustered to obtain a plurality of first clusters, namely, the first clusters , the first cluster , the first cluster and the first cluster , where the first cluster The corresponding noise distribution type is , the first cluster The corresponding noise distribution type is , the first cluster The corresponding noise distribution type is and the first cluster The corresponding noise distribution type is Then the noise distribution type is calculated as The corresponding historical standard sequence is , the noise distribution type is The corresponding historical standard sequence is , the noise distribution type is The corresponding historical standard sequence is , the noise distribution type is The corresponding historical standard sequence is .

[0035] In one embodiment, the current standard sequence is determined based on the noise distribution type of the current periodic concentration sequence. This first includes determining the noise distribution type of the current periodic concentration sequence. Specifically, the similarity between the current periodic concentration sequence and all historical standard sequences corresponding to the first clustering clusters is calculated, and the historical standard sequence with the largest similarity is selected as the current standard sequence, wherein the similarity is the Euclidean distance between the current periodic concentration sequence and the historical standard sequence corresponding to the first clustering cluster.

[0036] For example, the Euclidean distance algorithm is used to calculate the current period concentration sequence and the historical standard sequence , historical standard sequence , historical standard sequence and historical standard series The similarities between them are 、 、 、 ,and , and finally select the similarity The corresponding historical standard sequence is Serves as the current standard sequence for the current cycle concentration sequence.

[0037] The above method is used to select the current standard sequence of the current period concentration sequence, and the concentration sequence corresponding to the noise generated by the engine failure is eliminated, thereby selecting an accurate current standard sequence, thereby improving the accuracy of the subsequent calculation of the abnormality degree of each concentration value in the current period concentration sequence.

[0038] In one embodiment, the abnormality degree of each concentration value in the concentration sequence of the current cycle is calculated using the following formula: Where, Indicates the number of the concentration sequence in the current engine cycle. The abnormality of the concentration value, Indicates the concentration sequence of the current cycle The concentration value, Indicates the current standard sequence The actual value of the concentration value, Represents the mean of all concentration values ​​in the concentration series of the current period, Indicates the number of concentration values ​​in the concentration sequence of the current period, represents the standard normalization function. Indicates the concentration sequence of the current cycle concentration value and the current standard sequence The difference between the concentration values, the greater the difference, the greater the The greater the abnormality of the concentration value. Indicates the degree of dispersion of the concentration values ​​in the current period concentration sequence. The greater the dispersion, the The larger the value of The greater the abnormality of the concentration value.

[0039] S123: Calculate the degree of engine failure.

[0040] In one embodiment, the abnormality and noise levels of each concentration value are used to perform weighted summation of each concentration value in the current period concentration sequence to obtain the engine failure degree. The calculation formula is: Where, is the concentration sequence of the current cycle concentration values, is the concentration sequence of the current cycle The noise level of each concentration value, Indicates the concentration sequence of the current cycle The abnormality degree of the concentration value, N represents the number of concentration values ​​in the concentration sequence of the current period, is the standard normalization function. The larger the value of The greater the possibility that the first concentration value is noise, the Concentration value The less credible, the more Concentration value A smaller weight makes it possible to calculate the i-th concentration value The value of the engine fault degree at the moment is low, thereby reducing the impact of the data collected at the moment on the fault degree of the overall moment.

[0041] It should be noted that the higher the engine's methane concentration, the greater the likelihood of engine failure. However, simply judging engine failure based on methane concentration is inaccurate. Specifically, the higher the engine's methane concentration and the higher the engine's temperature, the greater the probability of engine failure. The higher the engine's methane concentration and the lower the engine's temperature, the lower the probability of engine failure. This is because higher temperatures generally help reduce methane concentration, indicating a more complete combustion process. A higher methane concentration at this point indicates a greater probability of engine failure. In other words, the higher the engine's temperature, the lower the engine's methane concentration should be to ensure normal motor operation. Therefore, based on the changes in the engine's methane concentration, temperature, and related linear relationships, and combined with the degree of abnormality and noise level of each concentration value, the following alternative embodiment is provided for calculating the degree of engine failure.

[0042] In another embodiment, the degree of engine failure is calculated , the calculation formula is: Where, Indicates the Pearson correlation coefficient between the temperature series of the current cycle and the concentration series of the current cycle, which is used to reflect the changing trend of the temperature and concentration values ​​of the engine in the current cycle. is the concentration sequence of the current cycle concentration values, is the concentration sequence of the current cycle The noise level of each concentration value, N represents the number of concentration values ​​in the concentration sequence of the current cycle, is the standard normalization function.

[0043] At this point, the engine fault severity is determined, and engine fault detection is performed based on the severity of the fault. Specifically, if the engine fault severity exceeds an abnormality threshold, an alarm is issued to alert personnel, prompting them to perform maintenance. If the engine fault severity is less than or equal to the abnormality threshold, no alarm is issued, and normal operation resumes. In this embodiment, the empirical value of the abnormality threshold is 0.7. In other embodiments, the empirical value of the abnormality threshold can be 0.77, 0.8, or other values, and can be adjusted based on specific implementation circumstances.

[0044] Through the above steps, the noise level and abnormality level of each concentration value in the current cycle concentration sequence are calculated, which are used to weight the engine fault level, thereby improving the accuracy of the calculation results. Furthermore, based on the noise distribution of each concentration sequence in the database, the current standard sequence corresponding to the current cycle concentration sequence is screened, and the concentration sequence corresponding to the noise generated by the engine fault is eliminated. This allows the selection of a more accurate current standard sequence, thereby improving the accuracy of the abnormality calculation. Furthermore, when calculating the engine fault level, the changing trends of the engine's temperature and concentration values ​​within the current cycle are taken into account, further improving the accuracy of the engine fault level calculation.

[0045] In the description of this specification, "multiple" and "several" mean at least two, such as two, three or more, etc., unless otherwise clearly defined.

[0046] The technical features of the above-mentioned embodiments can be combined arbitrarily. In order to make the description concise, not all possible combinations of the technical features in the above-mentioned embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0047] The above-described embodiments merely illustrate several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent application. It should be noted that a person skilled in the art would be able to make numerous variations and improvements without departing from the spirit of the present invention, and all such variations and improvements fall within the scope of protection of the present invention.

Claims

1. An engine fault detection method, characterized in that: include: Acquire the concentration sequence of the engine in each historical period, calculate the noise degree of each concentration value in the concentration sequence, obtain the noise distribution of the concentration sequence, wherein the noise degree of the concentration value reflects the degree of deviation of the concentration value from the fitting value, and construct a database based on the noise distribution of all concentration sequences; According to the difference between each concentration value in the current period concentration sequence and each concentration value in the current standard sequence, the abnormality degree of each concentration value in the current period concentration sequence is obtained, and the abnormality degree and noise degree of each concentration value are used to perform weighted summation on each concentration value in the current period concentration sequence to obtain the fault degree of the engine, and the engine fault detection is performed according to the magnitude of the fault degree; The method for obtaining the current standard sequence includes: clustering the noise distribution of all concentration sequences in the database to obtain multiple first clusters, wherein one first cluster represents a noise distribution type, and clustering all concentration sequences in any first cluster to obtain two second clusters, and calculating the average value of all concentration sequences in the second cluster containing the most concentration sequences as the historical standard sequence corresponding to the noise distribution type of the first cluster, and determining the current standard sequence based on the noise distribution type of the current period concentration sequence.

2. The engine fault detection method according to claim 1, characterized in that: Performing fault detection on the engine according to the degree of the fault includes: If the fault level is greater than an abnormal threshold, an alarm is issued to alert the staff.

3. The engine fault detection method according to claim 1, characterized in that: The current standard sequence of noise distribution type is determined by the noise distribution type of the current period concentration sequence, including: Determining the noise distribution type of the current cycle concentration sequence includes: calculating the similarity between the current cycle concentration sequence and all historical standard sequences corresponding to the first cluster clusters, and selecting the historical standard sequence with the largest similarity as the current standard sequence, wherein the similarity is the Euclidean distance between the current cycle concentration sequence and the historical standard sequence corresponding to the first cluster cluster.

4. The engine failure detection method according to claim 1, characterized in that: Calculate the degree of engine failure , the calculation formula is: ； In the formula, is the concentration sequence of the current cycle concentration value, is the concentration sequence of the current cycle The noise level of each concentration value is Indicates the concentration sequence of the current cycle The abnormality of the concentration value Indicates the number of concentration values ​​in the concentration sequence of the current period, is the standard normalization function.

5. The engine fault detection method according to claim 1, characterized in that: The fault degree also reflects the change trend of the temperature value and concentration value of the engine in the current cycle. The calculation formula is: ； In the formula, It represents the Pearson correlation coefficient between the temperature series of the current cycle and the concentration series of the current cycle, and is used to reflect the changing trend of the temperature and concentration values ​​of the engine in the current cycle. is the concentration sequence of the current cycle concentration value, is the concentration sequence of the current cycle The noise level of each concentration value is Indicates the concentration sequence of the current cycle The abnormality of the concentration value Indicates the number of concentration values ​​in the concentration sequence of the current period, is the standard normalization function.

6. The engine fault detection method according to claim 1, characterized in that: Calculate the abnormal degree of each concentration value in the current period concentration sequence. The calculation formula is: ； In the formula, Indicates the concentration sequence of the current engine cycle. The abnormality of the concentration value Indicates the concentration sequence of the current cycle The concentration value of Indicates the current standard sequence The actual value of the concentration value, Represents the mean of all concentration values ​​in the concentration sequence of the current period, Indicates the number of concentration values ​​in the concentration sequence of the current period, Represents the standard normalization function.

7. The engine fault detection method according to claim 1, characterized in that: Calculate the noise level of each concentration value in the concentration sequence, including: Performing least square fitting on the concentration values ​​in the concentration sequence to obtain a first fitting value of each concentration value; removing any concentration value from the concentration sequence, performing least square fitting on the remaining concentration values ​​to obtain a second fitting value of each remaining concentration value; Calculate the remaining concentration values ​​in the concentration sequence excluding any concentration value, the cumulative sum of differences between the measured value and the first fitted value, the cumulative sum of differences between the measured value and the second fitted value, and calculate the difference between the two cumulative sums of differences, and use the normalized difference as the noise performance degree of the concentration value.

8. The engine fault detection method according to claim 1, characterized in that: Calculate the noise level of each concentration value in the concentration sequence. The calculation formula is: ； In the formula, For the The noise level of each concentration value is is the concentration sequence except The concentration value outside The measured value of the concentration, is the concentration sequence except The concentration value outside The first fitted value for the concentration value, is the concentration sequence except The concentration value outside The second fitted value of the concentration value, represents the number of concentration values ​​in the concentration sequence, is the standard normalization function.

9. The engine fault detection method according to claim 1, characterized in that: Obtain the concentration series of engines in each historical period, including: The concentration value of methane in the engine is collected once at a preset time interval within the cycle to obtain a concentration sequence consisting of a plurality of concentration values.

Images