A port equipment operation and maintenance fault sound monitoring method based on a cooperative neural network algorithm

Abstract

The application provides a port equipment operation and maintenance fault sound monitoring method based on a cooperative neural network algorithm, comprising: pre-acquiring sound data, wherein the sound data comprises sound data of normal equipment operation and fault sound data; extracting audio features in the pre-acquired sound data; creating a sound recognition model based on the extracted audio features; and identifying fault sound by using the sound recognition model. The port equipment operation and maintenance fault sound monitoring method based on the cooperative neural network algorithm proposes a port equipment operation fault detection scheme, classifies and identifies sound faults, and further proposes a hierarchical processing method.

Description

Technical Field

[0001] This invention belongs to the field of port equipment operation and maintenance technology, and in particular relates to a method for monitoring port equipment operation and maintenance fault sounds based on a collaborative neural network algorithm. Background Technology

[0002] Currently, with the rapid development of new information technologies, port enterprises both domestically and internationally are actively applying these technologies to actual port operations. As the automation level of large-scale equipment continues to improve, the challenges of maintaining and operating such equipment are becoming increasingly prominent. To improve equipment maintenance efficiency, ensure the reliable, safe, and long-term stable operation of large-scale equipment, and break through the limitations of manual inspections and simplified mechanical maintenance processes, sensor-based IoT intelligent maintenance systems have gradually emerged. Among these, sound sensors are the most widely used in maintenance applications. During the operation of motors, belts, rollers, wire ropes, trolleys, and other parts of large equipment such as belt conveyors, stacker-reclaimers, gantry cranes, and ship unloaders, sound can be used to determine whether a fault has occurred and its severity, guiding maintenance operations accordingly.

[0003] Currently, in the sound monitoring of equipment operation and maintenance, sensors are mainly installed on a single part of the equipment for monitoring, and then transmitted to the Internet of Things system for detection. Due to the insufficient data processing capabilities of the system's backend, there are problems such as: (1) the sound is noisy and it is impossible to accurately identify the fault of the equipment part; (2) it is impossible to classify and monitor different levels of fault of the same part; (3) the sound data processing algorithm of the system is simple, with low processing efficiency and poor accuracy. Summary of the Invention

[0004] In view of this, the present invention proposes a method for monitoring port equipment operation and maintenance fault sounds based on a collaborative neural network algorithm, in order to partially solve one of the above-mentioned technical problems.

[0005] To achieve the above objectives, the technical solution of the present invention is implemented as follows:

[0006] On the one hand, this invention provides a method for monitoring port equipment operation and maintenance fault sounds based on a collaborative neural network algorithm, including:

[0007] Pre-collected audio data, which includes audio data of normal equipment operation and audio data of faults;

[0008] Extract audio features from pre-acquired audio data;

[0009] A sound recognition model is created based on the extracted audio features, and the sound recognition model is used to identify fault sounds.

[0010] Furthermore, the fault sound data includes fault sound data of different levels.

[0011] Furthermore, before extracting audio features from the pre-acquired audio data, preprocessing of the pre-acquired audio data is required, as follows:

[0012] The pre-acquired audio data is processed by denoising, segmenting into frames, pre-emphasizing, and windowing.

[0013] Furthermore, audio features are extracted from the pre-acquired audio data, including:

[0014] After preprocessing, the collected sound data is converted into spectral parameters using a short-time Fourier transform. Then, a convolution operation is performed using a Mel frequency filter, and logarithmic operations are applied to the data results for each frequency band to obtain a set of logarithmic energy X(t), t = [1, 30]. To obtain the Mel cepstral coefficients, a discrete cosine transform (DCT) is performed on the obtained parameters. The Mel cepstral coefficients are characteristic parameters of the port equipment's operating sound, and the formula is as follows:

[0015]

[0016] Where n is the number of coefficients, q i (n) is the nth Mel coefficient of the i-th frame, X(t) is the output of the logarithmic energy module, and M is the number of Mel filters.

[0017] Furthermore, a sound recognition model is created based on the extracted audio data, and this model is used to identify fault sounds, including:

[0018] The pre-collected fault sound data of different levels, after preprocessing and audio feature extraction, are used as the prototype vector v for sound fault level classification. k ;

[0019] The collected real-time sound data from the device is preprocessed and audio features are extracted, which is used as the quantity to be identified, q.

[0020] The variable to be identified, q, is moved from the intermediate state q(t) to the fault level prototype variable v. k In this way, the identification of the quantity to be identified, q, can be achieved.

[0021] Furthermore, the following steps are included:

[0022] 1. Using the variable to be identified q and the prototype variable v k A certain number of order parameters are formed and used as parameters of the algorithm model;

[0023] The dynamic method that satisfies sound recognition is as follows:

[0024]

[0025] In the formula, q is the state vector; γ k >0;

[0026] 2. Design a normalized and zero-mean sound prototype vector v k And calculate the adjoint vector. Where v k =(v k1 ,v k2 ,...,v kN )′, where N is the dimension of the prototype vector;

[0027] When designing normalization and mean conditions, the following requirements must be met: and Adjoint vector satisfy:

[0028]

[0029] 3. The state vector q is derived from v k The remaining vector w represents:

[0030]

[0031] Where, the adjoint vector of q is:

[0032]

[0033] For order parameters;

[0034] 4. Based on the feature vector q(0) of the original value of the selected sound to be identified, and using the formula... Right now get

[0035] 5. Using order parameters The dynamic equations describe the evolution of the system;

[0036] Assuming equation (2) represents the extreme value of the potential function, and disregarding F(t) and transient quantities in equation (2), the cooperative potential function is obtained as follows:

[0037]

[0038] Substituting the order parameter into equation (5) yields two equations related to the cooperative potential function: the order parameter dynamic equation and the potential function.

[0039]

[0040]

[0041] (6) Based on the order parameter If the actual effect is stable, proceed to the next step; otherwise, go back to the previous step.

[0042] When the system's dynamic energy drops to 0 At this point, the system is in a stable state, that is:

[0043]

[0044] in

[0045]

[0046] At this point, after the system discretization process, the dynamic equation of the order parameter becomes:

[0047]

[0048] Where r is the iteration step size. The stability of a collaborative neural network algorithm system is determined by the magnitude of r;

[0049] (7) Fault identification and classification determination --- The stable sequence parameters of the collaborative neural network system are presented according to formula (11) to finally complete the sound recognition process;

[0050]

[0051] On the other hand, the present invention provides a port equipment maintenance fault sound monitoring device based on a collaborative neural network algorithm, comprising:

[0052] The sound acquisition module is used to acquire and pre-acquire sound data in real time, including sound data of normal equipment operation and fault sound data.

[0053] The preprocessing module is used to preprocess the real-time and pre-acquired audio data;

[0054] The extraction module is used to extract audio features from real-time and pre-acquired sound data;

[0055] A module is created to generate a sound recognition model based on the audio features extracted from the pre-acquired sound data, and the sound recognition model is used to identify fault sounds.

[0056] Compared with existing technologies, the port equipment operation and maintenance fault sound monitoring method based on collaborative neural network algorithm described in this invention has the following advantages:

[0057] (1) The port equipment operation and maintenance fault sound monitoring method based on collaborative neural network algorithm described in this invention proposes a port equipment operation fault detection scheme, classifies and identifies sound faults, and then proposes processing methods in different levels.

[0058] (2) The port equipment operation and maintenance fault sound monitoring method based on the collaborative neural network algorithm of the present invention proposes a fast identification method based on the sound feature parameters to be measured by the collaborative neural network algorithm, which has strong robustness and improves the operation and maintenance efficiency of port equipment.

[0059] (3) The port equipment operation and maintenance fault sound monitoring method based on the collaborative neural network algorithm described in this invention uses the Mel-Cepstral Coefficient method to quickly obtain the feature values ​​of equipment sound in the frequency domain, thereby improving the accuracy of system diagnosis. Attached Figure Description

[0060] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings:

[0061] Figure 1 This is a flowchart of the port equipment operation and maintenance fault sound monitoring method based on the collaborative neural network algorithm described in this invention;

[0062] Figure 2 This is a flowchart of the audio data preprocessing process described in this invention. Detailed Implementation

[0063] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other.

[0064] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0065] This embodiment provides a method for monitoring the sound of port equipment operation and maintenance faults based on a collaborative neural network algorithm. According to the port equipment operation and maintenance management method, it manages the sound fault quantity during equipment operation in a hierarchical manner, including: normal operation sounds corresponding to no action, first-level fault sounds corresponding to continued monitoring, second-level fault sounds corresponding to post-operation maintenance, and third-level fault sounds corresponding to immediate shutdown and maintenance, etc. Figure 1 As shown, it includes the following steps:

[0066] 1. Pre-collection of port equipment operating sound data

[0067] The use of, but not limited to, sound sensors is employed to collect normal operating sound data and fault sound data of port equipment.

[0068] 2. Audio preprocessing – noise reduction, unsupervised segmentation

[0069] In addition to the noise generated by the port equipment itself, there is also noise from surrounding equipment during actual operation. The collected sound data undergoes denoising processing, followed by frame segmentation, with each frame being 20ms long. This process includes pre-emphasis and windowing. For details, please refer to [link to documentation]. Figure 2 .

[0070] 3. Audio Feature Extraction – Mel Frequency Cepstral Coefficient Method

[0071] (1) The pre-collected sound states of different levels are processed and used as the prototype vector v for sound fault level classification. k Feature values ​​are extracted from the real-time operating sound status of the equipment and used as the target quantity q. The target quantity q is then transferred from the intermediate state q(t) to the fault level prototype quantity v. k In this way, the recognition of q can be achieved.

[0072] (2) In this method, the number of Mel frequency cepstral series is used to extract sound features under different conditions.

[0073] When processing the sound signal of port equipment during operation, it is unstable in the time domain. Furthermore, directly extracting feature parameters from the time domain would be time-consuming and fail to effectively approximate the time-domain signal, resulting in an overly smooth signal. Therefore, the signal is converted from the time domain to the frequency domain using a short-time Fourier transform for extraction. After processing in steps 1 and 2, the sound signal is transformed into spectral parameters using a short-time Fourier transform. Then, a convolution operation is performed using a Mel frequency filter, and logarithmic operations are performed on the data results for each frequency band to obtain a set of logarithmic energy X(t), t = [1, 30]. To obtain the Mel cepstral coefficients, a discrete cosine transform (DCT) is performed on the obtained parameters. The Mel cepstral coefficients are the sound feature parameters of the port equipment operation, as shown in the following equation:

[0074]

[0075] Where n is the number of coefficients, q i (n) is the nth Mel coefficient of the i-th frame, X(t) is the output of the logarithmic energy module, and M is the number of Mel filters.

[0076] 4. Training based on collaborative neural network algorithm model --- Establish a collaborative neural network algorithm model and use feature data extracted from fault classification under different states to train the algorithm model.

[0077] (1) The system evolves according to the order parameter dynamics equation. In the algorithm model, the variable to be identified, q, and the prototype variable, v, are used. k A certain number of order parameters are formed, serving as important parameters for the algorithm model. The dynamics of sound recognition are as follows:

[0078]

[0079] Where: q is the state vector; γ k >0.

[0080] (2) Design the normalized and zero-mean sound prototype vector v k And calculate the adjoint vector. Where v k =(v k1 ,v k2 ,...,v kN )′, where N is the dimension of the prototype vector.

[0081] When designing normalization and mean conditions, the following requirements must be met: and Adjoint vector satisfy:

[0082]

[0083] (3) The state vector q can be derived from v k And w (residual vector) means:

[0084]

[0085] Where, the adjoint vector of q is

[0086]

[0087] The order parameter.

[0088] (4) Based on the feature vector q(0) of the original value of the selected sound to be identified, and using the formula Right now You can get

[0089] (5) Using sequence parameters The dynamic equations describe the evolution of the system.

[0090] Equation (2) can be considered as the extreme value of the potential function. If F(t) and transient quantities in equation (2) are not considered, the cooperative potential function can be obtained:

[0091]

[0092] Substituting the order parameter into equation (5) yields two equations related to the cooperative potential function: the order parameter dynamic equation and the potential function.

[0093]

[0094]

[0095] (6) Based on the order parameter If the actual effect is stable, proceed to the next step; otherwise, jump to the previous step.

[0096] When the system's dynamic energy drops to 0 At this point, the system is in a stable state, that is:

[0097]

[0098] in

[0099]

[0100] At this point, after the system discretization process, the dynamic equation of the order parameter becomes:

[0101]

[0102] Where r is the iteration step size. The stability of a collaborative neural network algorithm system is determined by the size of r.

[0103] (7) Fault identification and classification determination --- The stable sequence parameters of the collaborative neural network system are displayed according to formula (11) to finally complete the sound recognition process.

[0104]

[0105] 5. Real-time sound acquisition of port equipment

[0106] Sound sensors are installed in port equipment to collect sound signal data under different operating conditions.

[0107] 6. Real-time sound preprocessing for port equipment

[0108] The sound signals collected in real time during the operation of port equipment are preprocessed using the method described in step 2.

[0109] 7. Real-time sound feature extraction of port equipment

[0110] The preprocessed port equipment operating sound data were used to extract feature values ​​using the Mel frequency cepstral coefficient method, which is consistent with the sample data feature extraction method in step 3.

[0111] 8. Real-time sound of port equipment incorporated into the model

[0112] Real-time data, after preprocessing and feature extraction, is fed into the trained dynamic equations, while simultaneously setting... The values ​​of B, C, and r are used for identification.

[0113] 9. Real-time sound fault diagnosis of port equipment

[0114] The collaborative neural network algorithm can classify the collected sounds into different levels of warning information, thereby making corresponding fault diagnoses, realizing fault classification and proposing solutions.

[0115] On the other hand, the present invention provides a port equipment maintenance fault sound monitoring device based on a collaborative neural network algorithm, comprising:

[0116] The sound acquisition module is used to acquire and pre-acquire sound data in real time, including sound data of normal equipment operation and fault sound data.

[0117] The preprocessing module is used to preprocess the real-time and pre-acquired audio data;

[0118] The extraction module is used to extract audio features from real-time and pre-acquired sound data;

[0119] A module is created to generate a sound recognition model based on the audio features extracted from the pre-acquired sound data, and the sound recognition model is used to identify fault sounds.

[0120] It should be noted that each module in this embodiment can be hardware or software, or a combination of both. The execution method of each module can be referred to the above embodiment. The main purpose of this device is to realize the above-mentioned method for monitoring port equipment operation and maintenance fault sound based on a collaborative neural network algorithm.

[0121] This invention installs sound sensors in port equipment components, using the equipment's operational sound status as a basis. It employs normal operation, level 1 fault operation, level 2 fault operation, and level 3 fault operation as initial reference standards. A collaborative neural network algorithm is used to train sound recognition models for different states, enabling robust monitoring of the real-time operational sound status of port equipment. This overcomes the inability to accurately identify equipment faults and proposes a method for classifying and monitoring different fault levels. The sound recognition algorithm exhibits fast convergence, strong stability, and high applicability and continuity. This research on a port equipment operation and maintenance fault sound monitoring method based on a collaborative neural network algorithm, which integrates and identifies data using advanced data processing algorithms on top of sensor-monitored equipment operation status, has significant positive implications for improving the intelligent operation and maintenance of large port equipment.

[0122] Those skilled in the art will recognize that the units and method steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0123] In the several embodiments provided in this application, it should be understood that the disclosed methods and systems can be implemented in other ways. For example, the division of units described above is merely a logical functional division, and in actual implementation, there may be other division methods. For instance, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. The aforementioned units may or may not be physically separated. The components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of the embodiments of the present invention according to actual needs.

[0124] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be covered within the scope of the claims and specification of the present invention.

[0125] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A port equipment operation and maintenance fault sound monitoring method based on a cooperative neural network algorithm, characterized in that, include: Pre-collected sound data includes sound data of normal equipment operation and fault sound data, wherein the fault sound data includes fault sound data of different levels; Extract audio features from pre-acquired audio data; A sound recognition model is created based on the extracted audio features. This model is then used to identify fault sounds, including: The pre-acquired different levels of fault sound data are pre-processed and audio feature extracted to be the prototype vector of sound fault level classification ; The collected device real-time running sound data is preprocessed and audio feature extraction is performed, as a to-be-identified quantity q ; The to-be-identified quantity q From the intermediate state q(t) Into the fault level prototype quantity In, and further realize the identification of the to-be-identified quantity q ; ①Using the to-be-identified quantity q and the prototype quantity constitute a certain amount of order parameter as the parameter of the algorithm model; The dynamic method that satisfies sound recognition is as follows: In the formula, q is a state vector; ; ii. Designing the normalized and zero-mean conditioned sound prototype vector and computing the accompanying vector where , N is the dimension of the prototype vector; When designing the normalization and mean conditions, one requires that and ; the accompanying vector satisfies: iii) state vector q By and residual vector w denotes: wherein q The adjoint vector is: is an order parameter; IV. According to the selected sound to be identified quantity original value of the eigenvector , and using the formula , get ; • use of order parameters The dynamics equations evolve the system; It is considered that formula (2) is an extreme value of potential function, if not considering and transient quantity, then the cooperative potential function is obtained: Substituting the order parameter into equation (5) yields two equations related to the cooperative potential function: the order parameter dynamic equation and the potential function. (6) According to the stability of the evolution process, if the actual effect is stable, the next step is executed; otherwise, jump to the previous step; As the system dynamics energy drops to 0, ; at which time the system is in a stable state, i.e.: wherein At this point, after the system discretization process, the dynamic equation of the order parameter becomes: wherein, is the iteration step size, ; the stability of the collaborative neural network algorithm system is determined by the size of . ⑦ Fault identification and classification; The stable sequence parameters of the collaborative neural network system are presented according to formula (11) to finally complete the sound recognition process; 。 2. The port equipment operation and maintenance fault sound monitoring method based on the synergic neural network algorithm according to claim 1, characterized in that, Before extracting audio features from the pre-acquired audio data, the pre-acquired audio data needs to be pre-processed, as follows: The pre-acquired audio data is processed by denoising, segmenting into frames, pre-emphasizing, and windowing.

3. The port equipment operation and maintenance fault sound monitoring method based on the synergic neural network algorithm according to claim 1, characterized in that: Extracting audio features from the pre-acquired audio data, including: After preprocessing, the collected sound data is converted into spectral parameters using a short-time Fourier transform. Then, a convolution operation is performed using a Mel frequency filter, and logarithmic operations are applied to the data results for each frequency band to obtain a set of logarithmic energy X(t), t=[1,30]. To obtain the Mel cepstral coefficients, a discrete cosine transform (DCT) is performed on the obtained parameters. The Mel cepstral coefficients are characteristic parameters of the port equipment's operating sound, and the formula is as follows: where n is the number of coefficients, is the n-th mel coefficient of the i-th frame, X(t) is the output of the log energy module, and M is the number of mel filters.

4. A port equipment operation and maintenance fault sound monitoring device based on a cooperative neural network algorithm, characterized in that, include: The sound acquisition module is used to acquire and pre-acquire sound data in real time. The pre-acquired sound data includes sound data of normal operation of the equipment and fault sound data. The fault sound data includes fault sound data of different levels. The preprocessing module is used to preprocess the real-time and pre-acquired audio data; The extraction module is used to extract audio features from real-time and pre-acquired sound data; The module creates a sound recognition model based on the audio features extracted from the pre-acquired sound data. This model is then used to identify fault sounds, including: The pre-acquired different level of fault sound data is pre-processed and audio feature extracted to be a prototype vector for sound fault level classification ; The collected device real-time running sound data is preprocessed and audio feature extraction is performed, as a to-be-identified quantity q ; The to-be-identified quantity q From the intermediate state q(t) Into the fault level prototype quantity In, and further realize the identification of the to-be-identified quantity q ; ①Using the to-be-identified quantity q and the prototype quantity constitute a certain amount of order parameter as the parameter of the algorithm model; The dynamic method that satisfies sound recognition is as follows: In the formula, q is a state vector; ; ii. Designing the normalized and zero-mean conditioned sound prototype vector and computing the accompanying vector where , N is the dimension of the prototype vector; When designing the normalization and mean conditions, one requires that and ; the accompanying vector satisfies: iii) state vector q By and the residual vector w denotes: wherein q The accompanying vector of is: is an order parameter; IV. According to the selected sound to be identified quantity original value of the eigenvector , and using the formula , get ; • use of order parameters The dynamics equations evolve the system; It is considered that formula (2) is an extreme value of potential function, if not considering the and transient quantity, then the cooperative potential function is obtained: Substituting the order parameter into equation (5) yields two equations related to the cooperative potential function: the order parameter dynamic equation and the potential function. (6) According to the stability of the evolution process, if the actual effect is stable, the next step is executed; otherwise, jump to the previous step; As the system dynamics energy drops to 0, ; at which time the system is in a stable state, i.e.: wherein At this point, after the system discretization process, the dynamic equation of the order parameter becomes: wherein, is the iteration step size, The stability of the collaborative neural network algorithm system is determined by the size of . ⑦ Fault identification and classification; The stable sequence parameters of the collaborative neural network system are presented according to formula (11) to finally complete the sound recognition process; 。

Images