M-BaseFre clustering method for assisting conversion of complex signals into texts in multi-modal large model

Abstract

The invention relates to the technical field of fault diagnosis, and discloses an M-BaseFre clustering method for assisting conversion of a complex signal into a text in a multi-modal large model, and the method comprises the following steps: 1) solving an amplitude spectrum of a component signal; 2) finding out frequency peak values in all component amplitude spectrums based on a built-in findpeaks function of MATLAB, and reserving corresponding frequency values; 3) calculating the greatest common divisor among all peak frequency values of each component and the greatest common divisor among the components, and recombining the component signals with the common fundamental frequency; and 4) calculating a side frequency band value among all peak frequency values of the residual components, and recombining the component signals with the same value. According to the method, basic support can be provided for signal-to-text description, and application of a large language model in fault diagnosis is realized, so that major accidents caused by missed diagnosis and misdiagnosis of equipment faults are prevented.

Description

Technical Field

[0001] The present invention relates to the fields of fault diagnosis and large language models, and particularly to an M-BaseFre clustering method for assisting in converting complex signals into text in a multimodal large model. Background Art

[0002] In the wide applications of industrial equipment, mechanical systems, and complex structures, a large number of complex vibration signals of equipment are collected and analyzed. Traditional fault diagnosis methods mainly rely on expert experience and signal processing techniques such as Fourier transform and wavelet decomposition. These methods can extract certain features of the signals, but there are also obvious limitations. Specifically, traditional methods often rely on experts to manually interpret the features, making it difficult to meet the requirements of the development of intelligence and automation.

[0003] In recent years, large language models have demonstrated powerful capabilities in natural language processing, text generation, and understanding context relationships, making them potential powerful tools for industrial intelligence. By combining large language models, the feature extraction of complex signals can be closely integrated with text generation, improving the intelligence level of signal analysis. However, there are significant problems in directly inputting complex signal data into multimodal large models for training. There are often multiple feature aliases in complex signals, and interference noise is mixed with useful features, making it difficult for multimodal large models to accurately extract meaningful features when learning these data. As a result, the interpretability of the models is poor, affecting the accuracy of diagnosis. Therefore, relying solely on the powerful processing ability of large language models cannot directly solve the problem of complex signal analysis. Especially in fault diagnosis applications, how to ensure the accuracy and effectiveness of signal feature extraction and text description remains a major challenge.

[0004] In view of this, it is necessary to decompose and preprocess complex signals, decompose complex signals into single-component signals, then combine signal components with the same physical characteristics together, and then generate accurate text descriptions to improve the accuracy and effectiveness of the descriptions. This process not only more conforms to the signal diagnosis and analysis process of analog fault diagnosis experts but also can effectively reduce the interference of noise, making the descriptions generated by large language models more interpretable and practical.

[0005] Therefore, the present invention proposes an M-BaseFre clustering method for assisting in converting complex signals into text in a multimodal large model to better provide structured signal feature input for large language models. By decomposing and clustering complex signals, signal components with the same physical characteristics can be recombined, so as to be more accurate in text description, effectively reduce the influence of interference components, improve the efficiency of signal analysis and the application effect of large language models, and meet the high-precision and high-efficiency requirements of industrial intelligence for signal analysis. Summary of the Invention

[0006] The technical problem to be solved by the first aspect of the present invention is to provide an M-BaseFre clustering method for assisting in converting complex signals into text in a multi-modal large model. The technical problem to be solved by the first aspect of the present invention is to provide an M-BaseFre clustering method for assisting in converting complex signals into text in a multi-modal large model, which can provide basic support for device intelligent diagnosis based on large language models, avoid problems of artificial misdiagnosis and missed diagnosis, and thus prevent major accidents caused by faults.

[0007] To solve the above technical problems, the present invention provides an M-BaseFre clustering method for assisting in converting complex signals into text in a multi-modal large model. The method includes the following steps:

[0008] S1 Calculate the amplitude spectrum of all component signals.

[0009] S2 Based on the findpeaks function built in MATLAB, find the frequency peaks in all component amplitude spectra and retain the corresponding frequency values.

[0010] S3 Calculate the greatest common divisor between all peak frequency values of each component and the greatest common divisor between components, and reorganize the component signals with a common fundamental frequency.

[0011] S4 Calculate the sideband values between all peak frequency values of the remaining components and reorganize the component signals with the same values.

[0012] S5 Perform text description on the reorganized signals.

[0013] Preferably, in step S1, the component signals can be obtained by signal decomposition algorithms, including but not limited to matrix singular value decomposition, matrix singular spectrum decomposition, tensor singular value decomposition, and local mean decomposition.

[0014] Further preferably, in step S2, three operations of ceiling, floor, or rounding are performed on the retained peak frequency values, and they are accurately retained to one decimal place to reduce the error influence caused by frequency floating, thereby improving the accuracy of clustering. In MATLAB, the built-in ceil, floor, and round functions can be called to cooperate to achieve this purpose.

[0015] Preferably, the step S3 includes the following specific steps:

[0016] 41) For each component signal, calculate the greatest common divisor between all its frequency peaks. If there is only one peak in the amplitude spectrum of the component signal, the frequency of this peak is used as the greatest common divisor of this component. If no effective greatest common divisor is found in the component signal, mark this component as a "component not yet grouped".

[0017] 42) For the component signals with the greatest common divisor, further iteratively calculate whether there is a common fundamental frequency among the components. If so, recombine the component signals with the common fundamental frequency. For the component signals without a common fundamental frequency, group them separately; if there are finally N component signals and each component signal has an independent fundamental frequency, divide these N component signals into independent groups respectively.

[0018] Further preferably, step S4 includes the following specific steps:

[0019] 51) For the components marked as "components not yet grouped" in step S3, calculate the sideband values between all their frequency peaks, that is, calculate whether the difference between adjacent peak frequencies is a fixed value;

[0020] 52) For the component signals without sideband values, recombine them into a group of signals;

[0021] 53) For the component signals with sideband values, further determine whether the sideband values between the components are the same, and then recombine the component signals with the common sideband values. For the component signals without a common sideband, group them separately.

[0022] Preferably, step S5 includes the following specific steps:

[0023] 61) For the signals grouped in step S3, directly describe the signal characteristics in text from the amplitude spectrum and its phase spectrum characteristics, including the fundamental frequency, fundamental amplitude and phase, 2 times frequency, amplitude and its phase, and so on. If there is only a fundamental frequency without harmonics in the signal, only describe the fundamental frequency, fundamental amplitude and phase of the signal;

[0024] 62) For the signals grouped in step S4, draw their corresponding envelope spectrum diagrams, then find the frequency peaks in the envelope spectrum and describe them in text. If there is a common fundamental frequency between the peaks, the text description includes the fundamental frequency, fundamental amplitude and phase, 2 times frequency, amplitude and its phase, and so on. If there is no harmonic relationship between the peaks, only describe the frequency, amplitude and phase of each peak.

[0025] Through the above technical solutions, the M-BaseFre clustering method of the present invention can accurately and efficiently reorganize and group all component signals, enabling reasonable combination of signal components with the same physical meaning, and then realizing accurate signal-to-text description. This method combines a decomposition algorithm to achieve in-depth analysis and clustering processing of complex signals, effectively eliminating interference noise and feature aliasing problems, making the description results more accurate and physically interpretable. This method not only improves the extraction accuracy of signal features, but also significantly enhances the clarity and consistency of signal-to-text description, thus providing a reliable and high-quality basic support for intelligent fault diagnosis based on large language models.

[0026] Other features and advantages of the present invention will be described in detail in the subsequent specific implementation section. Brief Description of the Drawings

[0027] The present invention will be further described below in conjunction with the drawings and embodiments.

[0028] Figure 1 is the workflow diagram of the present invention;

[0029] Figure 2 is the time domain diagram of the embodiment of the present invention;

[0030] Figure 3 is the first grouping result processed by the method of the present invention in the embodiment of the present invention.

[0031] Figure 4 is the second grouping result processed by the method of the present invention in the embodiment of the present invention. Detailed Description of the Invention

[0032] The following will describe in detail the specific implementation of the present invention in conjunction with the drawings. It should be understood that the specific implementation described here is only for explaining and understanding the present invention, and is not used to limit the present invention.

[0033] An M-BaseFre clustering method for assisting complex signal-to-text conversion in a multi-modal large model, the method comprising the following steps:

[0034] S1 Calculate the amplitude spectrum of all component signals;

[0035] As Figure 1 shown is the workflow diagram of the present invention. Perform tensor singular value decomposition on the Figure 2 shown embodiment to obtain 11 component signals. Use Fourier transform to calculate the amplitude spectra of these 11 component signals.

[0036] S2 Based on the findpeaks function in MATLAB, find the frequency peaks in all component amplitude spectra and retain the corresponding frequency values;

[0037] Call the findpeaks function in MATLAB to find the frequency peaks in the amplitude spectra of the 11 component signals, and retain the frequency values corresponding to the peaks. Perform two operations of rounding up and rounding down on the frequency values to ensure that all frequency values are accurately retained to one decimal place, so as to reduce the error impact caused by frequency floating and improve the accuracy of clustering.

[0038] S3 Calculate the greatest common divisor between all peak frequency values of each component and the greatest common divisor between components, and reorganize the component signals with a common fundamental frequency;

[0039] For the 11 component signals in this embodiment, calculate the greatest common divisor between all frequency peaks within each component. Among them, there is only one frequency peak in the amplitude spectra of some component signals, and the unique frequency value is used as the greatest common divisor of these component signals. Mark the component signals for which no effective greatest common divisor is found as "components not yet grouped"; then, for the component signals with a greatest common divisor, further iteratively calculate whether there is a common fundamental frequency between these components, and reorganize the component signals with a common fundamental frequency. For the component signals without a common fundamental frequency, group them separately.

[0040] S4 Calculate the sideband values between all peak frequency values of the remaining components and reorganize the component signals with the same values.

[0041] For the "components not yet grouped" marked in the previous step, calculate the sideband values between all their frequency peaks, that is, calculate whether the difference between adjacent peak frequencies is a fixed value; for the component signals without sideband values, reorganize them into a group of signals; for the component signals with sideband values, further determine whether the sideband values between the components are the same, and then reorganize the component signals with a common sideband value. For the component signals without a common sideband, group them separately.

[0042] S5 Make a text description of the reorganized signals.

[0043] For the signals grouped in step S3, directly make a text description of the signal characteristics from the amplitude spectrum and its phase spectrum characteristics, including the fundamental frequency, fundamental amplitude and phase, second harmonic frequency, amplitude and its phase, and so on. If the signal has only a fundamental frequency and no harmonics, only describe the fundamental frequency, fundamental amplitude and phase of the signal; for the signals grouped in step S4, draw their corresponding envelope spectra, then find the frequency peaks in the envelope spectra and make a text description of them. If there is a common fundamental frequency between the peaks, the text description includes the fundamental frequency and amplitude, second harmonic frequency and amplitude, and so on. If there is no harmonic relationship between the peaks, only describe the frequency, amplitude and phase of each peak.

[0044] After being processed by step S3 and step S4 in this embodiment, the 11 component signals are recombined into 2 signals. Figure 3 is the signal obtained in step S3. Figure 4 is the signal processed by step S4. For the Figure 3 signal, the signal characteristics are directly described in text from the amplitude spectrum and its phase spectrum characteristics, including the fundamental frequency of 10.98 Hz, the fundamental amplitude of 0.23 m / s 2 and the phase of 0.78 radians, the second harmonic frequency of 22.01 Hz, the amplitude of 0.146 m / s 2 and its phase of 3.02 radians, the third harmonic frequency of 32.99 Hz, the amplitude of 0.037 m / s 2 , the phase of -1.36 radians; the fourth harmonic frequency of 43.98 Hz, the amplitude of 0.21 m / s 2 , the phase of -1.2 radians. For the Figure 4 signal, draw its corresponding envelope spectrum, then find the frequency peaks in the envelope spectrum and describe them in text, and the text description includes the fundamental frequency of 42.99 Hz and the fundamental amplitude of 0.215 m / s 2 , the second harmonic frequency of 85.98 Hz and the amplitude of 0.089 m / s 2 , the third harmonic frequency of 128.975 Hz and the amplitude of 0.041 m / s 2 , the fourth harmonic frequency of 172.00 Hz and the amplitude of 0.0096 m / s 2 , the fifth harmonic frequency of 214.99 Hz and the amplitude of 0.0081 m / s 2 . The text description corresponding to the entire signal is as follows:

[0045]

[0046]

[0047] In the description of the present invention, the descriptions referring to the terms "one embodiment", "some embodiments", "one implementation manner", etc. mean that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In the present invention, the schematic expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures or characteristics described can be combined in a suitable manner in any one or more embodiments or examples.

[0048] The preferred embodiments of the present invention have been described in detail above in conjunction with the accompanying drawings. However, the present invention is not limited thereto. Within the scope of the technical concept of the present invention, various simple modifications can be made to the technical solutions of the present invention, including the combination of each specific technical feature in any suitable manner. To avoid unnecessary repetition, the present invention will not separately describe various possible combination methods. However, these simple modifications and combinations should also be regarded as the content disclosed by the present invention and fall within the protection scope of the present invention.

Claims

1. An M-BaseFre clustering method for assisting complex signal conversion into text in a multimodal large model, characterized in that: The method comprises the following steps: S1 calculates the amplitude spectrum of all component signals; S2 finds the frequency peaks in all component amplitude spectra based on the findpeaks function provided by MATLAB and retains the corresponding frequency values; S3 calculates the greatest common divisor between all peak frequency values ​​of each component and the greatest common divisor between the components, and reorganizes the component signals having a common fundamental frequency; S4 calculates the sideband values ​​between all peak frequency values ​​of the remaining components and reassembles the component signals having the same value. S5 provides a text description of the reorganized signal.

2. According to claim 1, the M-BaseFre clustering method for assisting complex signal conversion into text in a multimodal large model is characterized in that: In step S1, the component signals are obtained by a signal decomposition algorithm, including matrix singular value decomposition, matrix singular spectrum decomposition, tensor singular value decomposition, and local mean decomposition.

3. The M-BaseFre clustering method for assisting complex signal conversion into text in a multimodal large model according to claim 2, characterized in that: In step S2, three operations of rounding up, rounding down or rounding off are performed on the retained peak frequency value to one decimal place to reduce the error impact caused by frequency fluctuation, thereby improving the accuracy of clustering.

4. The M-BaseFre clustering method for assisting complex signal conversion into text in a multimodal large model according to claim 3, characterized in that: The step S3 comprises the following specific steps: 41) For each component signal, calculate the greatest common divisor between all its frequency peaks. If there is only one peak in the amplitude spectrum of the component signal, the frequency of the peak is used as the greatest common divisor of the component. If no valid greatest common divisor is found in the component signal, mark the component as "ungrouped component"; 42) For the component signals with the greatest common divisor, further iteratively calculate whether there is a common fundamental frequency between the components. If so, reorganize the component signals with the common fundamental frequency. For the component signals without a common fundamental frequency, group them separately; if there are N component signals in the end and each component signal has an independent fundamental frequency, then group the N component signals into independent groups.

5. The M-BaseFre clustering method for assisting complex signal conversion into text in a multimodal large model according to claim 4, characterized in that: The step S4 comprises the following specific steps: 51) For the component marked as "not yet grouped" in step S3, calculate the sideband values ​​between all its frequency peaks, that is, calculate whether the difference between adjacent peak frequencies is a fixed value; 52) For component signals without sideband values, regroup them into a group of signals; 53) For component signals with sideband values, further determine whether the sideband values ​​between the components are the same, and then regroup the component signals with common sideband values. For component signals without common sidebands, group them separately.

6. The M-BaseFre clustering method for assisting complex signal conversion into text in a multimodal large model according to claim 5, characterized in that: The step S5 comprises the following specific steps: 61) For the signals grouped in step S3, describe the signal characteristics in text form from the amplitude spectrum and phase spectrum characteristics, including the fundamental frequency, fundamental amplitude and phase, and double frequency, amplitude and phase. If there is only the fundamental frequency without double frequency, only the fundamental frequency, fundamental amplitude and phase of the signal need to be described; 62) For the signal grouped in step S4, draw its corresponding envelope spectrum, then find the frequency peaks in the envelope spectrum and describe them in text. If there is a common fundamental frequency between the peaks, the text description includes the fundamental frequency, fundamental frequency amplitude and phase, 2nd harmonic frequency, amplitude and phase. If there is no harmonic relationship between the peaks, only the frequency, amplitude and phase of each peak are described.

Images