Gearbox fault diagnosis system based on one-dimensional convolutional nerve and algorithm optimization

Abstract

The invention relates to the technical field of gearbox fault diagnosis, in particular to a gearbox fault diagnosis system based on one-dimensional convolutional nerves and algorithm optimization, which comprises a signal unit used for periodically acquiring one-dimensional time sequence data of a gearbox and carrying out noise reduction processing, filtering processing and normalization processing on the one-dimensional time sequence data, the extraction unit is used for extracting fault features in the processed one-dimensional time sequence data to generate a fault feature map and screening and optimizing the fault feature map through a feature activation threshold to output a corresponding probability vector, and the analysis unit is used for judging the running state of the gearbox based on the probability vector. And the control unit is used for adjusting corresponding parameters or sending corresponding notifications. According to the invention, high-precision generation of the gearbox fault diagnosis result is effectively realized, and the diagnosis efficiency of the gearbox fault is effectively improved.

Description

Technical Field

[0001] This invention relates to the field of gearbox fault diagnosis technology, and in particular to a gearbox fault diagnosis system based on one-dimensional convolutional neural networks and algorithm optimization. Background Technology

[0002] In the aviation field, gearboxes are widely used and crucial transmission devices. During operation, gearboxes efficiently transmit power through gear meshing, while simultaneously drawing power from the engine to power various pumps and control systems. They play a vital role in the safe and reliable operation of aircraft engines. Gearboxes withstand various loads in harsh environments with high temperatures, high pressures, and high speed ratios. Failure to diagnose gearbox faults in a timely manner can lead to high maintenance costs or even major accidents. Therefore, accurate fault diagnosis of gearboxes is extremely important to ensure the normal operation of aviation equipment. Traditional fault diagnosis relies on expert experience, which has certain limitations. One-dimensional convolutional neural networks (CNNs) combined with algorithm optimization can process one-dimensional time-series data of gearboxes, enabling rapid and accurate fault diagnosis. With the development of technology, the use of one-dimensional convolutional neural networks and algorithm optimization for gearbox fault diagnosis has significant practical implications.

[0003] Chinese Patent Publication No. CN113029559A discloses a gearbox fault diagnosis method and system, which acquires the time-domain vibration signal of the gearbox; applies the maximum correlation kurtosis deconvolution algorithm to the acquired time-domain vibration signal of the gearbox to enhance the fault features, and obtains the time-domain vibration signal with enhanced fault features; performs fault feature extraction and dimensionality reduction on the time-domain vibration signal with enhanced fault features, and inputs the dimensionality-reduced fault features into a preset fault diagnosis model to obtain the fault diagnosis result.

[0004] Therefore, the above-mentioned scheme solves the problem of rapid fault diagnosis under conditions of strong noise and multiple coupled faults by rapidly diagnosing single / coupled faults in the gearbox. However, the above-mentioned scheme cannot achieve high-precision generation of gearbox fault diagnosis results, thus failing to guarantee the diagnostic efficiency for gearbox faults. Summary of the Invention

[0005] To address this issue, the present invention provides a gearbox fault diagnosis system based on one-dimensional convolutional neural networks and algorithm optimization, which overcomes the problem in the prior art that the high-precision generation of gearbox fault diagnosis results cannot be achieved, resulting in low diagnostic efficiency for gearbox faults.

[0006] To achieve the above objectives, this invention provides a gearbox fault diagnosis system based on one-dimensional convolutional neural networks and algorithm optimization, comprising: The signal unit is used to periodically acquire one-dimensional time-series data of the gearbox and perform noise reduction, filtering and normalization processing on the one-dimensional time-series data. The one-dimensional time-series data includes vibration signals, temperature signals and pressure signals. An extraction unit, connected to the signal unit, is used to extract fault features from the processed one-dimensional time-series data to generate a fault feature map. The fault feature map is then filtered and optimized using a feature activation threshold, and the corresponding probability vector is output. An analysis unit, connected to the extraction unit, is used to determine the operating state of the gearbox based on the probability vector. The analysis unit is also used to determine, based on the normal maximum confidence level, whether the diagnostic result indicating that the gearbox operating state is normal is incorrect. Alternatively, based on the maximum confidence level of the fault, determine whether the diagnostic results regarding the fault in the gearbox's operating state are incorrect. Alternatively, based on signal consistency, determine whether the diagnostic result indicating that the gearbox is operating normally is incorrect. Alternatively, based on signal consistency, determine whether the diagnostic result indicating a fault in the gearbox's operating status is incorrect. Alternatively, generate corresponding processing instructions. Alternatively, it can complete the periodic determination of the gearbox's operating status and determine the gearbox's operating status for the next cycle. The control unit, which is connected to the signal unit, the extraction unit and the analysis unit respectively, is used to adjust the preset maximum confidence level based on historical false alarm cases, adjust the cutoff frequency based on spectrum analysis, adjust the passband based on spectrum analysis, adjust the feature activation threshold based on the confidence difference, or issue a notification to add the current one-dimensional time series data to the model training.

[0007] Furthermore, the analysis unit is used to determine the operating state of the gearbox based on the probability vector, and to determine whether the diagnostic result of the gearbox operating state being normal is erroneous based on the normal maximum confidence level, or to determine whether the diagnostic result of the gearbox operating state having a fault is erroneous based on the fault maximum confidence level.

[0008] Furthermore, the analysis unit is also used to determine whether the diagnostic result indicating that the gearbox operating state is normal is erroneous based on the normal maximum confidence level, and to determine whether the diagnostic result indicating that the gearbox operating state is normal is erroneous based on the signal consistency according to the determination result, or to adjust the preset normal maximum confidence level based on the historical false alarm cases; the analysis unit is also used to determine whether the diagnostic result indicating that the gearbox operating state has a fault is erroneous based on the fault maximum confidence level, and to determine whether the diagnostic result indicating that the gearbox operating state has a fault is erroneous based on the signal consistency according to the determination result, or to adjust the passband based on the spectrum analysis.

[0009] Furthermore, the analysis unit is also used to determine whether the diagnostic result for a normal gearbox operating state is erroneous based on the signal consistency, and to determine that the gearbox operating state is normal based on the determination result, or to adjust the preset maximum confidence level of normal based on the historical false alarm cases; the analysis unit is also used to determine whether the diagnostic result for a faulty gearbox operating state is erroneous based on the signal consistency, and to issue a corresponding fault notification based on the determination result, or to adjust the passband based on the spectrum analysis.

[0010] Furthermore, the control unit is used to reduce the preset maximum confidence level based on the historical false alarm cases, and reduce the preset maximum confidence level to the corresponding preset maximum confidence level.

[0011] Furthermore, the analysis unit is also used to determine whether the diagnostic result for the gearbox operating state being normal is incorrect based on the normal maximum confidence level after the preset normal maximum confidence level has been reduced, and to determine that the gearbox operating state is normal based on the determination result, or to locate and adjust the cutoff frequency based on the spectrum analysis.

[0012] Furthermore, the control unit is also used to locate and adjust the cutoff frequency based on the spectrum analysis, and to reduce the high-pass cutoff frequency and retain the fault characteristic frequency band to the maximum extent according to the determination result, or to increase the low-pass cutoff frequency and retain the fault characteristic frequency band to the maximum extent; the control unit is also used to locate and adjust the passband based on the spectrum analysis to ensure that the fault characteristic frequency band is within the passband and to filter out high-frequency noise and low-frequency noise to the maximum extent.

[0013] Furthermore, the analysis unit is also used to determine whether the diagnostic result indicating that the gearbox operating state is normal is erroneous based on the normal maximum confidence level after the cutoff frequency adjustment is completed, and to determine that the gearbox operating state is normal based on the determination result, or to issue a notification to add the current one-dimensional time series data to the model training; the analysis unit is also used to determine whether the diagnostic result indicating that the gearbox operating state has a fault is erroneous based on the fault maximum confidence level after the passband adjustment is completed, and to issue a corresponding fault notification based on the determination result, or to adjust the feature activation threshold based on the confidence difference.

[0014] Furthermore, the control unit is also used to increase the feature activation threshold based on the confidence difference, and the increase in the feature activation threshold is proportional to the confidence difference.

[0015] Furthermore, the analysis unit is also used to determine whether the diagnostic result for the fault in the gearbox operating state is incorrect based on the maximum confidence of the fault after the feature activation threshold is increased, and to issue a corresponding fault notification based on the determination result, or to issue a notification to add the current one-dimensional time series data to the model training.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: By setting up an analysis unit and a control unit, the present invention further determines the operating state of the gearbox based on probability vectors, determines whether the diagnostic result of the gearbox operating state being normal is erroneous based on the maximum confidence level of normal operation, determines whether the diagnostic result of the gearbox operating state being faulty is erroneous based on the maximum confidence level of fault, determines whether the diagnostic result of the gearbox operating state being normal is erroneous based on signal consistency, and determines whether the diagnostic result of the gearbox operating state being faulty is erroneous based on signal consistency. This enables timely and accurate determination of whether the diagnostic result of the gearbox operating state is erroneous, ensuring the accuracy of the system's determination. The control unit adjusts or issues corresponding notifications for the corresponding parameters, completing the adjustment of the corresponding parameters in a timely and accurate manner. While further realizing the high-precision generation of gearbox fault diagnosis results, the present invention further improves the diagnostic efficiency of gearbox faults.

[0017] Furthermore, the analysis unit set in this invention is used to determine the operating state of the gearbox based on the probability vector, and promptly determines whether the diagnostic result for the gearbox operating state as normal is incorrect or whether the diagnostic result for the gearbox operating state as faulty is incorrect, so as to avoid misjudgment. While further realizing the high-precision generation of gearbox fault diagnosis results, it also further improves the diagnostic efficiency for gearbox faults.

[0018] Furthermore, the analysis unit of this invention is also used to determine whether the diagnostic result for a normal gearbox operating state is erroneous based on the normal maximum confidence level, accurately determine whether the diagnostic result for a normal gearbox operating state is erroneous based on signal consistency, or whether the preset normal maximum confidence level needs to be adjusted based on historical false alarm cases. The analysis unit is also used to determine whether the diagnostic result for a faulty gearbox operating state is erroneous based on the fault maximum confidence level, accurately determine whether the diagnostic result for a faulty gearbox operating state is erroneous based on signal consistency, or whether the passband needs to be adjusted based on spectrum analysis. This further achieves high-precision generation of gearbox fault diagnosis results and further improves the diagnostic efficiency for gearbox faults.

[0019] Furthermore, the analysis unit of this invention is also used to determine whether the diagnostic result for a normal gearbox operating state is erroneous based on signal consistency, effectively determining whether the preset maximum confidence level needs to be adjusted based on historical false alarm cases. The analysis unit is also used to determine whether the diagnostic result for a faulty gearbox operating state is erroneous based on signal consistency, accurately determining whether a corresponding fault notification needs to be issued or whether the passband needs to be adjusted based on spectrum analysis. This further realizes the high-precision generation of gearbox fault diagnosis results and further improves the diagnostic efficiency for gearbox faults.

[0020] Furthermore, the control unit provided in this invention is also used to reduce the preset maximum confidence level based on historical false alarm cases, reducing the preset maximum confidence level to the corresponding preset maximum confidence level. This effectively adjusts the preset maximum confidence level and avoids the situation where the maximum confidence level is lower than the preset maximum confidence level due to the preset maximum confidence level not meeting the standard. This further improves the efficiency of gearbox fault diagnosis while achieving high-precision generation of gearbox fault diagnosis results.

[0021] Furthermore, the analysis unit provided in this invention is also used to determine whether the diagnostic result for the normal operating state of the gearbox is incorrect based on the normal maximum confidence level after the preset normal maximum confidence level has been reduced, and to accurately determine whether the cutoff frequency needs to be adjusted based on spectrum analysis. This further improves the efficiency of gearbox fault diagnosis while achieving high-precision generation of gearbox fault diagnosis results.

[0022] Furthermore, the control unit provided in this invention is also used to adjust the cutoff frequency or the passband based on spectrum analysis, effectively avoiding the situation where the normal maximum confidence level is less than the preset normal maximum confidence level due to the cutoff frequency not meeting the standard, and effectively avoiding the situation where the maximum confidence level of the fault is less than the preset maximum confidence level of the fault due to the passband not meeting the standard. While further realizing the high-precision generation of gearbox fault diagnosis results, it further improves the diagnostic efficiency for gearbox faults.

[0023] Furthermore, the analysis unit of this invention is also used to determine whether the diagnostic result of the gearbox operating state being normal is erroneous based on the normal maximum confidence level after the cutoff frequency adjustment is completed, and to accurately determine whether it is necessary to issue a notification to add the current one-dimensional time series data to the model training. The analysis unit is also used to determine whether the diagnostic result of the gearbox operating state having a fault is erroneous based on the fault maximum confidence level after the passband adjustment is completed, and to accurately determine whether a corresponding fault notification needs to be issued or whether the feature activation threshold needs to be adjusted based on the confidence difference. This further realizes the high-precision generation of gearbox fault diagnosis results and further improves the diagnostic efficiency for gearbox faults.

[0024] Furthermore, the control unit set in this invention is also used to increase the feature activation threshold based on the confidence difference, which effectively avoids the situation where the maximum confidence of the fault is less than the preset maximum confidence of the fault due to the feature activation threshold not meeting the standard. While further realizing the high-precision generation of gearbox fault diagnosis results, it further improves the diagnostic efficiency for gearbox faults.

[0025] Furthermore, the analysis unit set in this invention is also used to determine whether the diagnostic result for the gearbox operating state is incorrect based on the maximum confidence of the fault after the feature activation threshold is increased. It can promptly and accurately determine whether a corresponding fault notification needs to be issued or a notification to add the current one-dimensional time series data to the model training needs to be issued. This not only further realizes the high-precision generation of gearbox fault diagnosis results, but also further improves the diagnostic efficiency for gearbox faults. Attached Figure Description

[0026] Figure 1 This is a structural block diagram of a gearbox fault diagnosis system based on one-dimensional convolutional neural networks and algorithm optimization, according to an embodiment of the present invention. Figure 2 This is a flowchart illustrating the workflow of a gearbox fault diagnosis system based on one-dimensional convolutional neural networks and algorithm optimization, as described in an embodiment of the present invention. Figure 3 This is a flowchart illustrating how to determine the operating status of a gearbox and whether the diagnostic results are erroneous, according to an embodiment of the present invention. Figure 4 This is a flowchart illustrating how an embodiment of the present invention determines whether a diagnostic result for the operating status of a gearbox is erroneous. Detailed Implementation

[0027] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0028] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0029] Furthermore, it should be noted that, in the description of this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0030] Please see Figure 1 The diagram shown is a structural block diagram of a gearbox fault diagnosis system based on one-dimensional convolutional neural networks and algorithm optimization according to an embodiment of the present invention. The gearbox fault diagnosis system based on one-dimensional convolutional neural networks and algorithm optimization according to an embodiment of the present invention includes a signal unit, an extraction unit, an analysis unit, and a control unit; wherein, The signal unit is used to periodically acquire one-dimensional time-series data of the gearbox, and to perform noise reduction, filtering and normalization processing on the one-dimensional time-series data. The one-dimensional time-series data includes vibration signals, temperature signals and pressure signals. The extraction unit is connected to the signal unit and is used to extract fault features from the processed one-dimensional time series data to generate a fault feature map. The fault feature map is then filtered and optimized using a feature activation threshold, and the corresponding probability vector is output. The analysis unit is connected to the extraction unit and is used to determine the operating state of the gearbox based on the probability vector. The analysis unit is also used to determine, based on the normal maximum confidence level, whether the diagnostic result indicating that the gearbox operating state is normal is incorrect. Alternatively, based on the maximum confidence level of the fault, determine whether the diagnostic results regarding the fault in the gearbox's operating state are incorrect. Alternatively, based on signal consistency, determine whether the diagnostic result indicating that the gearbox is operating normally is incorrect. Alternatively, based on signal consistency, determine whether the diagnostic result indicating a fault in the gearbox's operating status is incorrect. Alternatively, generate corresponding processing instructions. Alternatively, it can complete the periodic determination of the gearbox's operating status and determine the gearbox's operating status for the next cycle. The control unit is connected to the signal unit, the extraction unit and the analysis unit respectively, and is used to adjust the preset maximum confidence level based on historical false alarm cases, adjust the cutoff frequency based on spectrum analysis, adjust the passband based on spectrum analysis, adjust the feature activation threshold based on confidence level difference, or issue a notification to add the current one-dimensional time series data to the model training. Specifically, the noise reduction process can remove noise signals, the filtering process can filter fault characteristic frequency bands, and the normalization process can unify the data scale of the one-dimensional time series data. The filtering process sets the passband and cutoff frequency of the bandpass filter according to the fault characteristic frequency band, so as to effectively separate and retain the fault characteristics. The feature activation threshold is set to filter out noisy and weak activations during the extraction process and optimize feature saliency, thereby outputting a high-quality probability vector. In this embodiment, a single cycle is 1 time / 1 second.

[0031] Please see Figure 2 The diagram shows the workflow of a gearbox fault diagnosis system based on one-dimensional convolutional neural networks and algorithm optimization according to an embodiment of the present invention. When the gearbox fault diagnosis system based on one-dimensional convolutional neural networks and algorithm optimization is running, the signal unit periodically collects the one-dimensional time-series data of the gearbox and performs noise reduction, filtering, and normalization processing on the one-dimensional time-series data. The one-dimensional time-series data includes vibration signals, temperature signals, and pressure signals. The extraction unit extracts the fault features from the processed one-dimensional time-series data to generate a fault feature map. The fault feature map is then filtered and optimized using the feature activation threshold, and the corresponding probability vector is output. The analysis unit determines the operating state of the gearbox based on the probability vector. The analysis unit also determines whether the diagnostic result indicating a normal operating state of the gearbox is incorrect based on the normal maximum confidence level. Alternatively, based on the maximum confidence level of the fault, determine whether the diagnostic result for a fault in the gearbox operating state is incorrect; or, based on the signal consistency, determine whether the diagnostic result for a normal gearbox operating state is incorrect; or, based on the signal consistency, determine whether the diagnostic result for a fault in the gearbox operating state is incorrect; or, generate corresponding processing instructions; or, complete the periodic determination of the gearbox operating state and determine the gearbox operating state for the next cycle. The control unit adjusts the preset maximum confidence level of normal operation based on the historical false alarm cases, adjusts the cutoff frequency based on the spectrum analysis, adjusts the passband based on the spectrum analysis, adjusts the feature activation threshold based on the confidence level difference, or, issues a notification to add the current one-dimensional time series data to the model training.

[0032] Please see Figure 3 The diagram shows a flowchart illustrating how an embodiment of the present invention determines the operating state of a gearbox and whether the diagnostic results are erroneous. The analysis unit described in this embodiment is used to determine the operating state of the gearbox based on the probability vector. If the confidence level of the normal state is the highest in the probability vector, the confidence level is recorded as the maximum confidence level of the normal state, and the diagnosis result of the normal gearbox operating state is determined based on the maximum confidence level of the normal state. If the confidence level of a certain fault state in the probability vector is the highest, the confidence level is recorded as the maximum confidence level of the fault, and the diagnosis result of the fault in the gearbox operating state is determined based on the maximum confidence level of the fault. Specifically, the probability vector includes operating status parameters and confidence levels. The operating status parameters include normal or several specific fault names, and the confidence level is the probability corresponding to each operating parameter. The maximum value of the probability is determined among all probabilities, and the operating status corresponding to this probability is the diagnostic result. After the diagnostic results are generated, it is necessary to further determine whether the diagnostic results are erroneous based on the probability corresponding to them, thus ensuring that the system generates diagnostic results with high accuracy.

[0033] Please continue reading. Figure 3 As shown, the analysis unit of the present invention is also used to determine whether the diagnostic result indicating that the gearbox operating state is normal is erroneous based on the normal maximum confidence level, and to determine whether the diagnostic result indicating that the gearbox operating state has a fault is erroneous based on the fault maximum confidence level: If the normal maximum confidence level is greater than or equal to the preset normal maximum confidence level W1 set in the analysis unit, the analysis unit determines whether the diagnostic result of determining the gearbox operating state as normal based on the signal consistency is incorrect. In this embodiment, the preset normal maximum confidence level W1 = 80%. If the normal maximum confidence level is less than the preset normal maximum confidence level W1, the analysis unit determines to adjust the preset normal maximum confidence level based on the historical false alarm cases; If the maximum confidence level of the fault is greater than or equal to the preset maximum confidence level W2 set in the analysis unit, the analysis unit determines whether the diagnostic result of the gearbox operating state being faulty based on the signal consistency is incorrect. In this embodiment, the preset maximum confidence level W2 is 95%. If the maximum confidence level of the fault is less than the preset maximum confidence level of the fault W2, the analysis unit determines to adjust the passband based on the spectrum analysis. Specifically, the preset maximum confidence level for normal operation and the preset maximum confidence level for fault are taken as empirical thresholds. After determining that the diagnostic result for the gearbox operating state is correct based on the maximum confidence level for normal operation and the maximum confidence level for fault, the system can further ensure the high accuracy of the diagnostic result generation by determining whether the diagnostic result for the gearbox operating state is correct based on the signal consistency.

[0034] Please continue reading. Figure 3 As shown, the analysis unit of the present invention is also used to determine, based on the signal consistency, whether the diagnostic result indicating that the gearbox is in a normal operating state is erroneous, and, based on the signal consistency, whether the diagnostic result indicating that the gearbox is in a faulty operating state is erroneous. If the signal consistency is consistent, the analysis unit determines that the diagnostic result for the normal operation status of the gearbox is correct, the gearbox is in normal operation status, completes the periodic determination of the gearbox's operation status, and determines the gearbox's operation status for the next cycle. If the signal consistency is inconsistent, the analysis unit determines to adjust the preset maximum confidence level based on the historical false alarm cases. If the signal consistency is consistent, the analysis unit determines that the diagnosis result for the gearbox operating status as faulty is correct and issues the corresponding fault notification. If the signal consistency is inconsistent, the analysis unit determines to adjust the passband based on the spectrum analysis. Specifically, when determining whether the signal consistency is consistent, the signal used is whether several signals among the vibration signal, the temperature signal and the pressure signal that correspond to the normal state or a certain fault state point to the normal state or the fault state. If the signal consistency is consistent, it indicates that several signals corresponding to the normal state all point to the normal state, or it indicates that several signals corresponding to a certain fault state all point to the fault state. If the signal consistency is inconsistent, it indicates that there is a signal corresponding to the normal state that does not point to the normal state, or that there is a signal corresponding to a certain fault state that does not point to that fault state.

[0035] Please continue reading. Figure 3 As shown, the control unit of the present invention is used to reduce the preset maximum confidence level of normality based on the historical false alarm cases: The confidence level A of the preset percentile in the historical false alarm cases is determined to be the corresponding preset normal maximum confidence level, and the preset normal maximum confidence level is reduced to the corresponding preset maximum confidence level. In this embodiment, the preset percentile A = the 85th percentile.

[0036] Please see Figure 4 The diagram shows a flowchart illustrating how an embodiment of the present invention determines whether a diagnostic result for the gearbox's operating state is erroneous. The analysis unit in this embodiment is further configured to determine, based on the preset maximum confidence level, whether a diagnostic result indicating the gearbox's operating state is normal is erroneous after the preset maximum confidence level has decreased. If the normal maximum confidence level is greater than or equal to the preset normal maximum confidence level W1, the analysis unit determines whether the diagnostic result of determining the gearbox operating state as normal based on the signal consistency is incorrect; If the normal maximum confidence level is less than the preset normal maximum confidence level W1, the analysis unit determines to adjust the cutoff frequency based on the spectrum analysis. If the signal consistency is consistent, the analysis unit determines that the diagnostic result for the normal operation status of the gearbox is correct, the gearbox is in normal operation status, completes the periodic determination of the gearbox's operation status, and determines the gearbox's operation status for the next cycle. If the signal consistency is inconsistent, the analysis unit determines to adjust the cutoff frequency based on the spectrum analysis.

[0037] Please continue reading. Figure 4 As shown, the control unit in this embodiment of the invention is further configured to locate and adjust the cutoff frequency based on the spectrum analysis, and to locate and adjust the passband based on the spectrum analysis: The noise frequency band is located by the spectrum analysis. If high-frequency noise is filtered out, the high-pass cutoff frequency is reduced while retaining the fault characteristic frequency band to the maximum extent. If low-frequency noise is filtered out, the low-pass cutoff frequency is increased while retaining the fault characteristic frequency band to the maximum extent. The fault characteristic frequency band and the noise frequency band are located through the spectrum analysis, and the passband is adjusted to ensure that the fault characteristic frequency band is within the passband and to filter out the high-frequency noise and the low-frequency noise to the greatest extent.

[0038] Please continue reading. Figure 4 As shown, the analysis unit in this embodiment of the invention is further used to determine, based on the maximum confidence level, whether the diagnostic result indicating a normal gearbox operating state is erroneous after the cutoff frequency adjustment is completed, and to determine, based on the maximum confidence level, whether the diagnostic result indicating a fault in the gearbox operating state is erroneous after the passband adjustment is completed: If the normal maximum confidence level is greater than or equal to the preset normal maximum confidence level W1, the analysis unit determines whether the diagnostic result of determining the gearbox operating state as normal based on the signal consistency is incorrect; If the normal maximum confidence level is less than the preset normal maximum confidence level W1, the analysis unit determines to issue a notification to add the current one-dimensional time series data to the model training. If the maximum confidence level of the fault is greater than or equal to the preset maximum confidence level of the fault W2, the analysis unit determines whether the diagnostic result of the gearbox operating state being faulty based on the signal consistency is incorrect. If the maximum confidence level of the fault is less than the preset maximum confidence level W2 of the fault set in the analysis unit, the analysis unit determines to adjust the feature activation threshold based on the confidence level difference; If the signal consistency is consistent, the analysis unit determines that the diagnostic result for the normal operation status of the gearbox is correct, the gearbox is in normal operation status, completes the periodic determination of the gearbox's operation status, and determines the gearbox's operation status for the next cycle. If the signal consistency is inconsistent, the analysis unit determines to issue a notification to add the current one-dimensional time series data to the model training. If the signal consistency is consistent, the analysis unit determines that the diagnosis result for the gearbox operating status as faulty is correct and issues the corresponding fault notification. If the signal consistency is inconsistent, the analysis unit determines to adjust the feature activation threshold based on the confidence difference. Specifically, the confidence difference is the difference between the preset maximum confidence level of the fault and the maximum confidence level of the fault.

[0039] Please continue reading. Figure 4 As shown, the control unit in this embodiment of the invention is further configured to increase the feature activation threshold based on the confidence difference: If the confidence difference is greater than the second preset confidence difference △F2 set in the analysis unit, the analysis unit determines to increase the initial feature activation threshold by a second magnitude E2. In this embodiment, the second preset confidence difference △F2 = 72% and the second magnitude E2 = 0.1. If the confidence difference is less than or equal to the second preset confidence difference △F2 and greater than the first preset confidence difference △F1 set in the analysis unit, the analysis unit determines to increase the initial feature activation threshold by a first amplitude E1. In this embodiment, the first preset confidence difference △F1 = 31% and the first amplitude E1 = 0.05. If the confidence difference is less than or equal to the first preset confidence difference △F1, the analysis unit determines to maintain the current initial feature activation threshold; Specifically, the preset maximum confidence level of the fault is 95%, the range of the confidence level difference is 0% to 95%, the initial feature activation threshold is 0.2, and the larger the confidence level difference, the greater the increase in the feature activation threshold. The values ​​of the confidence level difference and the feature activation threshold are derived based on the actual debugging results.

[0040] Please continue reading. Figure 4 As shown, the analysis unit described in this embodiment of the invention is further used to determine, based on the maximum confidence level of the fault, whether the diagnostic result for a fault in the gearbox operating state is incorrect after the feature activation threshold has been increased: If the maximum confidence level of the fault is greater than or equal to the preset maximum confidence level of the fault W2, the analysis unit determines whether the diagnostic result of the gearbox operating state being faulty based on the signal consistency is incorrect. If the maximum confidence level of the fault is less than the preset maximum confidence level W2 of the fault set in the analysis unit, the analysis unit determines to issue a notification to add the current one-dimensional time series data to the model training. If the signal consistency is consistent, the analysis unit determines that the diagnosis result for the gearbox operating status as faulty is correct and issues the corresponding fault notification. If the signal consistency is inconsistent, the analysis unit determines to issue a notification to add the current one-dimensional time series data to the model training.

[0041] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

[0042] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A gearbox fault diagnosis system based on one-dimensional convolutional neural networks and algorithm optimization, characterized in that, include: The signal unit is used to periodically acquire one-dimensional time-series data of the gearbox and perform noise reduction, filtering and normalization processing on the one-dimensional time-series data. The one-dimensional time-series data includes vibration signals, temperature signals and pressure signals. An extraction unit, connected to the signal unit, is used to extract fault features from the processed one-dimensional time-series data to generate a fault feature map. The fault feature map is then filtered and optimized using a feature activation threshold, and the corresponding probability vector is output. An analysis unit, connected to the extraction unit, is used to determine the operating state of the gearbox based on the probability vector. The analysis unit is also used to determine, based on the normal maximum confidence level, whether the diagnostic result indicating that the gearbox operating state is normal is incorrect. Alternatively, based on the maximum confidence level of the fault, determine whether the diagnostic results regarding the fault in the gearbox's operating state are incorrect. Alternatively, based on signal consistency, determine whether the diagnostic result indicating that the gearbox is operating normally is incorrect. Alternatively, based on signal consistency, determine whether the diagnostic result indicating a fault in the gearbox's operating status is incorrect. Alternatively, generate corresponding processing instructions. Alternatively, it can complete the periodic determination of the gearbox's operating status and determine the gearbox's operating status for the next cycle. The control unit, which is connected to the signal unit, the extraction unit and the analysis unit respectively, is used to adjust the preset maximum confidence level based on historical false alarm cases, adjust the cutoff frequency based on spectrum analysis, adjust the passband based on spectrum analysis, adjust the feature activation threshold based on the confidence difference, or issue a notification to add the current one-dimensional time series data to the model training.

2. The gearbox fault diagnosis system based on one-dimensional convolutional neural networks and algorithm optimization according to claim 1, characterized in that, The analysis unit is used to determine the operating state of the gearbox based on the probability vector, and to determine whether the diagnosis result of the gearbox operating state being normal is erroneous based on the normal maximum confidence level, or to determine whether the diagnosis result of the gearbox operating state having a fault is erroneous based on the fault maximum confidence level.

3. The gearbox fault diagnosis system based on one-dimensional convolutional neural networks and algorithm optimization according to claim 2, characterized in that, The analysis unit is also used to determine whether the diagnostic result for the normal operating state of the gearbox is wrong based on the normal maximum confidence level, and to determine whether the diagnostic result for the normal operating state of the gearbox is wrong based on the signal consistency according to the determination result, or to adjust the preset normal maximum confidence level based on the historical false alarm cases. The analysis unit is also used to determine whether the diagnostic result for a fault in the gearbox operating state is erroneous based on the maximum confidence of the fault, and to determine whether the diagnostic result for a fault in the gearbox operating state is erroneous based on the signal consistency according to the determination result, or to locate and adjust the passband based on the spectrum analysis.

4. The gearbox fault diagnosis system based on one-dimensional convolutional neural networks and algorithm optimization according to claim 3, characterized in that, The analysis unit is also used to determine whether the diagnostic result for the gearbox operating status being normal is erroneous based on the signal consistency, and to determine that the gearbox operating status is normal based on the determination result, or to adjust the preset maximum confidence level of normal based on the historical false alarm cases. The analysis unit is also used to determine whether the diagnostic result for a faulty gearbox operating state is incorrect based on the signal consistency, and to issue a corresponding fault notification based on the determination result, or to locate and adjust the passband based on the spectrum analysis.

5. The gearbox fault diagnosis system based on one-dimensional convolutional neural networks and algorithm optimization according to claim 4, characterized in that, The control unit is used to reduce the preset maximum confidence level based on the historical false alarm cases, and reduce the preset maximum confidence level to the corresponding preset maximum confidence level.

6. The gearbox fault diagnosis system based on one-dimensional convolutional neural networks and algorithm optimization according to claim 5, characterized in that, The analysis unit is also used to determine whether the diagnostic result of the gearbox operating state being normal is incorrect based on the normal maximum confidence level after the preset normal maximum confidence level is reduced, and to determine the gearbox operating state is normal based on the determination result, or to locate and adjust the cutoff frequency based on the spectrum analysis.

7. The gearbox fault diagnosis system based on one-dimensional convolutional neural networks and algorithm optimization according to claim 6, characterized in that, The control unit is also used to locate and adjust the cutoff frequency based on the spectrum analysis, and to reduce the high-pass cutoff frequency and retain the fault characteristic frequency band to the maximum extent according to the determination result, or to increase the low-pass cutoff frequency and retain the fault characteristic frequency band to the maximum extent. The control unit is also used to locate and adjust the passband based on the spectrum analysis to ensure that the fault characteristic frequency band is within the passband and to filter out high-frequency noise and low-frequency noise to the greatest extent.

8. The gearbox fault diagnosis system based on one-dimensional convolutional neural networks and algorithm optimization according to claim 7, characterized in that, The analysis unit is further configured to determine, based on the maximum confidence level, whether the diagnostic result indicating that the gearbox is operating normally is erroneous after the cutoff frequency adjustment is completed, and to determine that the gearbox is operating normally based on the determination result, or to issue a notification to add the current one-dimensional time series data to the model training; the analysis unit is further configured to determine, based on the maximum confidence level, whether the diagnostic result indicating that the gearbox is operating normally is erroneous after the passband adjustment is completed, and to issue a corresponding fault notification based on the determination result, or to adjust the feature activation threshold based on the confidence level difference.

9. The gearbox fault diagnosis system based on one-dimensional convolutional neural networks and algorithm optimization according to claim 8, characterized in that, The control unit is also used to increase the feature activation threshold based on the confidence difference, and the increase in the feature activation threshold is proportional to the confidence difference.

10. The gearbox fault diagnosis system based on one-dimensional convolutional neural networks and algorithm optimization according to claim 9, characterized in that, The analysis unit is also used to determine whether the diagnostic result for the gearbox operating state is incorrect based on the maximum confidence of the fault after the feature activation threshold is increased, and to issue a corresponding fault notification based on the determination result, or to issue a notification to add the current one-dimensional time series data to the model training.

Images