Fault early warning method and system for photovoltaic support connecting piece processing equipment

Abstract

The invention discloses a fault early warning method and system for photovoltaic support connecting piece machining equipment, and relates to the technical field of photovoltaic equipment manufacturing, the method utilizes the characteristic that assembly line machining of a connecting piece is repeated and regular mechanical reciprocating motion, the machining period from raw material machining to finished machining of the connecting piece is a machining period, and the machining efficiency of the connecting piece is improved. Even if a semi-finished product has a plurality of different stages in a processing period, the characteristics are also met, specifically, audio superposition in different stages is embodied, the fault of processing equipment mostly refers to gradual excessive wear of machinery in a large reciprocating process, the excessive wear refers to linear action which is slowly superposed along with a time sequence, and the processing efficiency is greatly improved. Through the audio data of the plurality of processing periods and the predicted audio data, the audio data and the predicted audio data are compared with each other, so that the excessive wear can be found at the initial stage, or the excessive wear can be predicted when the excessive wear does not actually occur through the prediction function provided by the invention, and the vacancy of fault detection of workpiece processing equipment in the mountain photovoltaic field is made up.

Description

Technical Field

[0001] This application relates to the technical field of photovoltaic equipment manufacturing, and particularly relates to a fault warning method and system for a processing device of photovoltaic bracket connectors. Background Art

[0002] Mountain photovoltaic refers to a power generation method that utilizes the mountain terrain and solar energy resources to convert solar energy into electrical energy by installing photovoltaic panels. The biggest challenges faced by mountain photovoltaic projects lie in the complex and changeable terrain and the uncertainty of geological conditions, such as steep slopes, variable rock structures, and possible geological disasters, which pose relatively high requirements for the design of photovoltaic brackets.

[0003] The brackets of mountain photovoltaic arrays are usually built using C-shaped steel purlins made of weathering steel. For easy transportation, the C-shaped steel purlins need to be segmented and cut. During construction and installation, special connectors are used to connect the segmented C-shaped steel purlins. To meet the overall structural strength and stability of the photovoltaic array brackets, the material of the connectors is usually selected as steel with a higher strength than weathering steel. To meet the assembly requirements, the steel plate raw materials need to be processed into the required specifications through methods such as sheet metal bending, punching, and cutting. However, high-yield-strength steels such as Q690 are difficult to be processed into the desired shape in one step. Therefore, it is necessary to process through an assembly line or an integrated processing device, but essentially, it is still about quickly passing the steel raw materials through each processing link.

[0004] Currently, for the monitoring of the processing equipment of C-shaped steel purlin connectors, it is usually to observe whether the quality of the finished products meets the requirements or to monitor abnormal noises of the processing equipment. The aforementioned monitoring actions are usually achieved by installing monitoring equipment on the equipment. In actual applications, it often takes until a certain processing link of the assembly line or processing equipment completely fails to operate, resulting in the shutdown of the assembly line, and then the overall shutdown and maintenance are carried out. This makes the fault judgment of the processing equipment from the early stage have prior experience and subjective errors, and to the subsequent situation where the processing equipment breaks down and the assembly line stops production before maintenance, it has the characteristics of subjective error judgment and lag, resulting in low efficiency of the assembly line processing of the connectors. Summary of the Invention

[0005] The main purpose of this application is to provide a fault warning method and system for a processing device of photovoltaic bracket connectors, so as to solve the problem in the prior art that the fault judgment of the processing equipment from the early stage has prior experience and subjective errors, and to the subsequent situation where the processing equipment breaks down and the assembly line stops production before maintenance, it has the characteristics of subjective error judgment and lag, resulting in low efficiency of the assembly line processing of the connectors.

[0006] To achieve the above purpose, this application provides the following technical solutions:

[0007] A fault warning method for a photovoltaic support connection part processing device, the photovoltaic support connection part processing device includes a punching end, at least one bending end, and a shearing end that are connected in a production line or integrated linearly. The steel raw material forms a connection part after passing through the punching end, all the bending ends, and the shearing end in sequence. The fault warning method includes:

[0008] Step S1, define the duration required to process a steel raw material into a connection part as a data acquisition cycle;

[0009] Step S2, repeatedly acquire the processing audio data of several data acquisition cycles through external sound pickup ends with at least two different listening angles;

[0010] Step S3, perform noise reduction and interpolation processing on each processing audio data respectively to obtain the processed audio data;

[0011] Step S4, perform amplitude averaging processing on all the processed audio data of the same data acquisition cycle, and obtain a mean audio data based on one data acquisition cycle;

[0012] Step S5, respectively obtain the extreme value coordinate points of all extreme points in each mean audio data based on the time process, and integrate all the extreme value coordinate points of one mean audio data into a coordinate point data set;

[0013] Step S6, use a machine learning algorithm to learn all the coordinate point data sets and predict at least one future coordinate point data set based on the data acquisition cycle. Each future coordinate point data set includes several future extreme value coordinate points;

[0014] Step S7, obtain the Euclidean distance between the extreme value coordinate points and the future extreme value coordinate points at the same time process;

[0015] Step S8, retain the future extreme value coordinate points with the Euclidean distance greater than the preset distance threshold and mark them as abnormal extreme value coordinate points;

[0016] Step S9, obtain the processing links corresponding to all the abnormal extreme value coordinate points based on the time process respectively, and define the processing links with the corresponding quantity exceeding the preset quantity threshold as processing fault risk links.

[0017] As a further improvement of the present application, in step S9, obtain the processing links corresponding to all the abnormal extreme value coordinate points based on the time process respectively, and define the processing links with the corresponding quantity exceeding the preset quantity threshold as processing fault risk links. After that, it includes:

[0018] Step S10, respectively obtain the number of processing fault risk links in each data acquisition cycle;

[0019] Step S20: Generate a rectangular coordinate system with the time process of the data acquisition period as the horizontal axis and the number of processing fault risk links as the vertical axis.

[0020] Step S30: Convert the current data acquisition period and the corresponding number of processing fault risk links into fault number coordinate points based on the rectangular coordinate system.

[0021] Step S40: Linearly fit all the fault number coordinate points to obtain a fault fitting function.

[0022] Step S50: Determine whether the fault fitting function is monotonically increasing. If so, execute Step S60.

[0023] Step S60: Determine that the processing equipment for the photovoltaic bracket connector has a fatal fault risk.

[0024] As a further improvement of the present application, after Step S60, which determines that the processing equipment for the photovoltaic bracket connector has a fatal fault risk, it includes:

[0025] Step S100: Send all the processing fault risk links and the corresponding future timestamps to an external monitoring terminal.

[0026] Step S200: Send the fatal fault risk and the corresponding future timestamp to an external monitoring terminal.

[0027] As a further improvement of the present application, in Step S3, perform denoising and interpolation processing on each processed audio data to obtain processed audio data, including:

[0028] Step S31: Perform denoising processing on each processed audio data respectively by improved spectral subtraction, and obtain a denoised frequency data based on one processed audio data.

[0029] Step S32: Obtain all the data coordinate points and the positions of all missing coordinate points of the current denoised frequency data based on the sound wave spectrum with a preset scale interval.

[0030] Step S33: Calculate the coordinate values of the positions of all missing coordinate points by data interpolation method with all the data coordinate points as known quantities.

[0031] Step S34: Input the coordinate values of all missing coordinate points into the sound wave spectrum to obtain interpolation coordinate points.

[0032] Step S35: Connect all the coordinate points and all the interpolation coordinate points linearly in sequence to obtain the processed audio data.

[0033] As a further improvement of the present application, in step S31, denoising processing is performed on each processed audio data by improving spectral subtraction, and a denoised frequency data is obtained based on one processed audio data, including:

[0034] Step S311, dividing the current processed audio data into voice segments and silent segments;

[0035] Step S312, extracting pure noise samples in the silent segments, and calculating the average spectral value of the pure noise samples through multi-frame averaging;

[0036] Step S313, equally dividing the voice segments into several mixed short-time frames;

[0037] Step S314, performing Fourier transform on each mixed short-time frame respectively, and obtaining a mixed spectrum based on one mixed short-time frame;

[0038] Step S315, calculating a denoised spectrum through improved spectral subtraction based on the mixed short-time frame and the mixed spectrum;

[0039] Step S316, inversely transforming all denoised spectra into several denoised short-time frames through inverse Fourier transform;

[0040] Step S317, splicing all denoised short-time frames along the time series into the denoised frequency data corresponding to the current processed audio data.

[0041] As a further improvement of the present application, in step S33, the coordinate values at the positions of all missing coordinate points are calculated by data interpolation method with all data coordinate points as known quantities, including:

[0042] Step S331, obtaining the Euclidean distance and the semi-variogram value between two adjacent data coordinate points respectively;

[0043] Step S332, obtaining the fitting curve of all Euclidean distances and all semi-variograms, so that the fitting curve calculates the corresponding semi-variogram according to any Euclidean distance;

[0044] Step S333, solving the semi-variogram value between any two data coordinate points according to the fitting curve;

[0045] Step S334, defining the optimal coefficient matrix of all semi-variograms according to all solved semi-variogram values, and the optimal coefficient matrix includes the semi-variogram values to be solved at the positions of the missing coordinate points;

[0046] Step S335, performing weighted summation on all semi-variogram values according to all optimal coefficients in the optimal coefficient matrix to obtain the solutions of all semi-variogram values to be solved;

[0047] Step S336: Substitute the solution of each semi-variogram value to be solved into the fitting curve respectively to obtain the Euclidean distance between each missing coordinate point position and other data coordinate points.

[0048] Step S337: Convert the Euclidean distance between the current missing coordinate point position and other data coordinate points into the coordinate value of the current missing coordinate point position based on multi-point positioning.

[0049] As a further improvement of the present application, in step S6, all coordinate point data sets are learned by a machine learning machine and at least one future coordinate point data set is predicted based on the data acquisition period. Each future coordinate point data set includes several future extreme value coordinate points, including:

[0050] Step S61: Perform normalization processing on each coordinate point data set respectively to obtain a normalized data set based on one coordinate point data set.

[0051] Step S62: Divide a normalized data set into a training set, a validation set, and a test set according to a preset ratio.

[0052] Step S63: Train all training sets through a training model, and update the weights and biases of the training model by the backpropagation algorithm according to the training results and the loss function of the validation set.

[0053] Step S64: Repeat step S63 several times until the loss function reaches the minimum value.

[0054] Step S65: Obtain the training model corresponding to the minimum value of the loss function and define it as a prediction model.

[0055] Step S66: Input all test sets into the prediction model to obtain a set of future extreme value coordinate points based on one data acquisition period.

[0056] To achieve the above object, the present application also provides the following technical solutions:

[0057] A fault warning system for a photovoltaic bracket connector processing device. The fault warning system is applied to the fault warning method as described above. The fault warning system includes:

[0058] A data acquisition period definition module, which is used to define the duration required to process a steel raw material into a connector as a data acquisition period.

[0059] A processing audio data acquisition module, which is used to repeatedly acquire the processing audio data of several data acquisition periods through external pick-up ends with at least two different listening angles.

[0060] The processed audio data processing module is used to perform denoising and interpolation processing on each processed audio data respectively to obtain the processed audio data;

[0061] The mean audio data acquisition module is used to perform amplitude averaging on all the processed audio data in the same data acquisition period, and obtain a mean audio data based on one data acquisition period;

[0062] The extreme value coordinate point set acquisition module is used to respectively obtain the extreme value coordinate points of all extreme points in each mean audio data based on the time process, and integrate all the extreme value coordinate points of one mean audio data into a coordinate point data set;

[0063] The coordinate point data set prediction module is used to learn all the coordinate point data sets through machine learning and predict at least one future coordinate point data set based on the data acquisition period. Each future coordinate point data set includes several future extreme value coordinate points;

[0064] The Euclidean distance acquisition module is used to obtain the Euclidean distance between the extreme value coordinate points and the future extreme value coordinate points at the same time process;

[0065] The abnormal extreme value coordinate marking module is used to retain the future extreme value coordinate points whose Euclidean distance is greater than the preset distance threshold and mark them as abnormal extreme value coordinate points;

[0066] The processing failure risk definition module is used to obtain the respective processing links corresponding to all the abnormal extreme value coordinate points based on the time process, and define the processing links with the corresponding quantity exceeding the preset quantity threshold as the processing failure risk links.

[0067] To achieve the above purpose, the present application also provides the following technical solutions:

[0068] An electronic device includes a processor and a memory coupled to the processor. The memory stores program instructions executable by the processor. When the processor executes the program instructions stored in the memory, the above-mentioned fault warning method is implemented.

[0069] To achieve the above purpose, the present application also provides the following technical solutions:

[0070] A storage medium stores program instructions, and when the program instructions are executed by a processor, the above-mentioned fault warning method can be implemented.

[0071] Beneficial effects:

[0072] In this application, the time required to process a steel raw material into a connecting piece is defined as a data acquisition cycle; the processed audio data for several data acquisition cycles is repeatedly acquired through external sound pick-up ends with at least two different listening angles; noise reduction and interpolation processing are respectively performed on each processed audio data to obtain the processed audio data; amplitude averaging processing is performed on all the processed audio data for the same data acquisition cycle, and a mean audio data is obtained based on one data acquisition cycle; the extreme value coordinate points of all the extreme points in each mean audio data based on the time process are respectively obtained, and all the extreme value coordinate points of one mean audio data are integrated into a coordinate point data set; all the coordinate point data sets are learned by a machine learning machine and at least one future coordinate point data set is predicted based on the data acquisition cycle, and each future coordinate point data set includes several future extreme value coordinate points; the Euclidean distance between the extreme value coordinate points and the future extreme value coordinate points at the same time process is obtained; the future extreme value coordinate points with the Euclidean distance greater than the preset distance threshold are retained and marked as abnormal extreme value coordinate points; the processing links corresponding to all the abnormal extreme value coordinate points based on the time process are obtained, and the processing link with the corresponding quantity exceeding the preset quantity threshold is defined as the processing failure risk link. This application makes use of the characteristic that the pipeline processing of the connecting piece is a repetitive and regular mechanical reciprocating motion. By taking one connecting piece from the raw material to the completion of processing as one processing cycle, even if there are semi-finished products in multiple different stages within one processing cycle, it still conforms to the above characteristics, which is specifically reflected in the audio superposition of different stages. Moreover, most of the failures of the processing equipment are the gradual excessive wear of the machine during a large number of reciprocating processes, and the excessive wear is a linear action that slowly accumulates over time. Therefore, this application compares the audio data of several processing cycles and the predicted audio data with each other, so that the excessive wear can be detected in the initial stage, or can be predicted through the prediction function provided by this application before the excessive wear actually occurs, thus making up for the gap in the failure detection of the workpiece processing equipment in the mountain photovoltaic field. In addition, this application makes up for the audio missing during the sound pick-up process and then performs subsequent judgment operations, so that the failure judgment and failure warning of this application have good accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] Figure 1 FIG. is a schematic flow chart of the steps of an embodiment of the method for fault warning of the processing equipment of the photovoltaic support connecting piece of this application;

[0074] Figure 2 FIG. is a schematic diagram of the functional modules of an embodiment of the fault warning system of the processing equipment of the photovoltaic support connecting piece of this application;

[0075] Figure 3 FIG. is a schematic structural diagram of an embodiment of the electronic device of this application;

[0076] Figure 4Schematic structural diagram of an embodiment of the storage medium of the present application. Detailed implementation manners

[0077] The following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

[0078] The terms "first", "second", and "third" in the present application are only used for descriptive purposes, and cannot be construed as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first", "second", and "third" may explicitly or implicitly include at least one of such features. In the description of the present application, the meaning of "a plurality" is at least two, such as two, three, etc., unless otherwise specifically defined. All directional indications (such as up, down, left, right, front, back...) in the embodiments of the present application are only used to explain the relative positional relationship and movement conditions between components in a specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indications will also change accordingly. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units is not limited to the listed steps or units, but optionally further includes steps or units not listed, or optionally further includes other steps or units inherent to these processes, methods, products, or devices.

[0079] Referring to "embodiments" herein means that the specific features, structures, or characteristics described in connection with the embodiments can be included in at least one embodiment of the present application. The phrase appears in various places in the specification and does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment mutually exclusive with other embodiments. Those skilled in the art will explicitly and implicitly understand that the embodiments described herein can be combined with other embodiments.

[0080] As Figure 1 shown, this embodiment provides a fault warning method for a photovoltaic bracket connector processing device. The photovoltaic bracket connector processing device includes a punching end, at least one bending end, and a shearing end that are connected in a pipeline or linearly integrated. The steel raw material forms a connector after passing through the punching end, all the bending ends, and the shearing end in sequence.

[0081] Preferably, for the detailed structure of the processing equipment for photovoltaic bracket connectors, reference can be made to the patent "CN118832054A - A processing device for high - efficiency, high - strength and weather - resistant steel C - shaped purlin connectors". This patent describes the processing equipment for photovoltaic bracket connectors in this embodiment and the necessary structures and processes for manufacturing the connectors. This embodiment can perform fault warning on the processing device in the above - mentioned patent.

[0082] Specifically, the fault warning method includes the following steps:

[0083] Step S1: Define the time required to process one steel raw material into one connector as one data acquisition cycle.

[0084] Preferably, according to the processing process of the above - mentioned reference patent, one connector goes through four stages: punching, then two bends, and finally shearing, namely punching - first bend - second bend - shearing in sequence. If the processing intensity is high, when the previous semi - finished product reaches the second - bend stage, new raw materials will start to enter the punching stage, and the audio will show a superimposed state, but this does not affect subsequent analysis and detection, and the audio at different times can be mutually verified. However, under normal circumstances, since one connector only requires four processes and generally a new raw material is not put in until one connector is processed, there will be no audio superposition.

[0085] Step S2: Repeatedly acquire the processing audio data of several data acquisition cycles through external sound - picking ends with at least two different listening angles.

[0086] Preferably, if the processing equipment with the above - mentioned structure is adopted, an external sound - picking end can be set in each of the six normal directions of front, back, left, right, up, and down. The sound - picking end can be any pick - up device in the prior art that can meet the requirements of this embodiment, and the detailed structure of the pick - up device will not be elaborated in this embodiment.

[0087] Step S3: Perform denoising and interpolation processing on each processing audio data respectively to obtain the processed audio data.

[0088] Preferably, this embodiment uses the improved spectral subtraction method as the denoising means and the semi - variogram function as the interpolation basis.

[0089] Step S4: Perform amplitude averaging processing on all the processed audio data of the same data acquisition cycle to obtain a mean audio data based on one data acquisition cycle.

[0090] Preferably, since there are multiple external sound - picking ends picking up sounds in the same data acquisition cycle, the design intention of this embodiment is to eliminate the audio differences at different positions caused by the self - occlusion of the equipment by taking the average value. Therefore, the audio acquired by all external sound - picking ends in the same data acquisition cycle is averaged.

[0091] Step S5: Obtain the extreme value coordinate points based on the time process for all extreme points in each average audio data, and integrate all the extreme value coordinate points of one average audio data into a coordinate point data set.

[0092] Preferably, the coordinate points can be obtained in the form of a sound wave spectrum or a frequency spectrum.

[0093] Step S6: Use machine learning to learn all the coordinate point data sets and predict at least one future coordinate point data set based on the data acquisition period. Each future coordinate point data set includes several future extreme value coordinate points.

[0094] Preferably, in this embodiment, the prediction function can be implemented through a neural network.

[0095] Step S7: Obtain the Euclidean distance between the extreme value coordinate points and the future extreme value coordinate points at the same time process.

[0096] Preferably, for example, a data acquisition period is 30s, that is, it takes 30s for a raw material to be processed into a connector through four steps. Then the same time process is the same position within these 30s, for example, both are at the 10th s, which is equivalent to comparing different ordinates with the same abscissa.

[0097] Step S8: Retain the future extreme value coordinate points with the Euclidean distance greater than the preset distance threshold and mark them as abnormal extreme value coordinate points.

[0098] Preferably, the preset distance threshold can be set to 10% of the amplitude of the extreme value coordinate points, that is, the Euclidean distance between the future extreme value coordinate points and the extreme value coordinate points at the same time process is greater than 10% of the amplitude of the extreme value coordinate points.

[0099] Step S9: Obtain the processing links corresponding to all the abnormal extreme value coordinate points based on the time process, and define the processing links with the corresponding quantity exceeding the preset quantity threshold as the processing failure risk links.

[0100] For example, if the Euclidean distance between the future extreme value coordinate points and the extreme value coordinate points at the 10th s is greater than 10% of the amplitude of the extreme value coordinate points, then the processing link at the 10th s is defined as the processing failure risk link; according to the fact that it takes 30s for a raw material to be processed into a connector through four steps, assuming that the punching link takes 8s, the first bending takes 8s, the second bending takes 8s, and the shearing takes 6s, then the 10th s is located in the first bending link, so it is determined that the first bending is the processing failure risk link.

[0101] It should be noted that the above-mentioned link durations are only for illustration and are not used to limit the duration of each link.

[0102] Further, in step S9, obtain the processing links corresponding to all abnormal extreme value coordinate points based on the time process, and define the processing links with the corresponding quantity exceeding the preset quantity threshold as the processing fault risk links. After that, the following steps are further included:

[0103] Step S10, respectively obtain the number of processing fault risk links in each data acquisition cycle.

[0104] For example, using the above four processing links, and one of them has been marked as a processing fault risk link. At this time, the number of processing fault risk links is 1. If an additional processing link is marked subsequently, it will be accumulated.

[0105] Step S20, generate a rectangular coordinate system with the time process of the data acquisition cycle as the horizontal axis and the value of the number of processing fault risk links as the vertical axis.

[0106] Step S30, convert the current data acquisition cycle and the corresponding number of processing fault risk links into fault number coordinate points based on the rectangular coordinate system.

[0107] Step S40, linearly fit all the fault number coordinate points to obtain a fault fitting function.

[0108] Preferably, the four links in this embodiment are fewer. For processing equipment with multiple links, the above-mentioned fault fitting function has more obvious fluctuations.

[0109] Step S50, determine whether the fault fitting function is monotonically increasing. If so, execute step S60.

[0110] Preferably, a monotonically increasing trend indicates that the number of fault links is increasing over time, that is, current faults breed more faults.

[0111] Step S60, determine that the processing equipment of the photovoltaic support connector has a fatal fault risk.

[0112] Further, in step S60, after determining that the processing equipment of the photovoltaic support connector has a fatal fault risk, the following steps are further included:

[0113] Step S100, send all the processing fault risk links and the corresponding future timestamps to an external monitoring terminal.

[0114] Step S200, send the fatal fault risk and the corresponding future timestamp to an external monitoring terminal.

[0115] Further, in step S3, perform denoising and interpolation processing on each processed audio data to obtain the processed audio data, including:

[0116] Step S31, perform denoising processing on each processed audio data through improved spectral subtraction, and obtain a denoised frequency data based on one processed audio data.

[0117] Preferably, the improved spectral subtraction introduces an over-subtraction factor greater than 1, subtracts additional energy based on the estimated noise spectrum value to reduce residual noise, and compensates for possible speech distortion caused by over-subtraction through a gain compensation factor, and retains more sound components by adjusting the gain factor.

[0118] Preferably, the improved spectral subtraction generally divides the spectrum into multiple sub-bands, and dynamically adjusts the over-subtraction factor and compensation factor according to the signal-to-noise ratio of each sub-band. There is usually more noise in the high-frequency band, and a larger over-subtraction factor is used; the speech energy is concentrated in the low-frequency band, and the reduction amplitude is reduced to retain the sound quality; the noise spectrum is dynamically updated through adaptive noise estimation, using recursive averaging or minimum tracking techniques to update the noise spectrum in real time to adapt to the non-stationary noise environment; the minimum statistic (such as minimum tracking) or a noise model based on probability distribution (such as Gaussian mixture model) is adopted to improve the robustness of noise estimation; then the spectral subtraction function is dynamically adjusted according to the signal-to-noise ratio, for example, a more aggressive subtraction is adopted in the low signal-to-noise ratio region, and more original spectra are retained in the high signal-to-noise ratio region to balance denoising and distortion; finally, a minimum spectral value (such as a certain proportion of the noise spectrum) is set to avoid speech breakage or musical noise caused by excessive reduction, and the processed spectrum is smoothed in the time domain and frequency domain to suppress the random fluctuation of the residual noise.

[0119] Step S32, obtain all data coordinate points and all missing coordinate point positions of the current denoised frequency data based on the acoustic wave spectrum with a preset scale interval.

[0120] Preferably, if the above-mentioned 30s time process is adopted, a large scale interval can be set to 3s, a large scale interval includes ten small scale intervals, and each small scale interval is 0.3s.

[0121] Step S33, calculate the coordinate values of all missing coordinate point positions by data interpolation method with all data coordinate points as known quantities.

[0122] Step S34, input the coordinate values of all missing coordinate points in the acoustic wave spectrum to obtain interpolated coordinate points.

[0123] Step S35, linearly connect all coordinate points and all interpolated coordinate points in sequence to obtain the processed audio data.

[0124] Furthermore, Step S31, perform denoising processing on each processed audio data through improved spectral subtraction, and obtain a denoised frequency data based on one processed audio data, including:

[0125] Step S311, divide the current processed audio data into voiced segments and silent segments.

[0126] Step S312, extract pure noise samples in the silent segment, and calculate the average spectral value of the pure noise samples through multi-frame averaging.

[0127] Step S313, equally divide the voiced segment into several mixed short-time frames.

[0128] Step S314, perform Fourier transform on each mixed short-time frame respectively, and obtain a mixed spectrum based on one mixed short-time frame.

[0129] Step S315, calculate the denoised spectrum through improved spectral subtraction based on the mixed short-time frame and the mixed spectrum.

[0130] Step S316, inverse-transform all the denoised spectra into several denoised short-time frames through inverse Fourier transform.

[0131] Step S317, splice all the denoised short-time frames along the time series into the denoised frequency data corresponding to the current processed audio data.

[0132] Preferably, the above short-time frame can be set to one of 20 ms to 40 ms.

[0133] Furthermore, in step S33, calculate the coordinate values at the positions of all missing coordinate points through data interpolation method with all data coordinate points as known quantities, including:

[0134] Step S331, obtain the Euclidean distance and the semi-variogram value between two adjacent data coordinate points respectively.

[0135] Step S332, obtain the fitting curve of all Euclidean distances and all semi-variograms, so that the fitting curve can calculate the corresponding semi-variogram according to any Euclidean distance.

[0136] Step S333, solve the semi-variogram value between any two data coordinate points according to the fitting curve.

[0137] Step S334, define the optimal coefficient matrix of all semi-variograms according to all the solved semi-variogram values, and the optimal coefficient matrix includes the semi-variogram values to be solved at the positions of the missing coordinate points.

[0138] Preferably, the optimal coefficient matrix is expressed as follows:

[0139]

[0140] where r ij is the semi-variogram between the i-th coordinate point and the j-th coordinate point, and λ i is the optimal coefficient of the i-th coordinate point and other coordinate points, and r iois the semivariogram of the position of the i-th missing coordinate point to all coordinate points, and φ is the Lagrange multiplier.

[0141] Step S335: Weighted sum all semivariogram values according to all optimal coefficients in the optimal coefficient matrix to obtain the solutions of all semivariogram values to be solved.

[0142] Step S336: Substitute the solution of each semivariogram value to be solved into the fitting curve respectively to obtain the Euclidean distance between each missing coordinate point position and other data coordinate points.

[0143] Step S337: Convert the Euclidean distance between the current missing coordinate point position and other data coordinate points into the coordinate value of the current missing coordinate point position based on multi-point positioning.

[0144] Preferably, according to the Euclidean distance between the current missing coordinate point position and other data coordinate points, use other data coordinate points as the centers of circles and the corresponding Euclidean distances as the radii. The intersection points of all the obtained circles are the coordinate values of the current missing coordinate point position.

[0145] It should be noted that the formulas in the above additional content are for principle explanations, and the symbol meanings of the formulas are not interoperable with other formulas.

[0146] Furthermore, in step S6, learn all coordinate point data sets through machine learning and predict at least one future coordinate point data set based on the data acquisition period. Each future coordinate point data set includes several future extreme value coordinate points, including:

[0147] Step S61: Normalize each coordinate point data set respectively to obtain a normalized data set based on one coordinate point data set.

[0148] Step S62: Divide a normalized data set into a training set, a validation set, and a test set according to a preset ratio.

[0149] Preferably, the preset ratio is 70%:15%:15%, that is, 70% training set, 15% validation set, and 15% test set.

[0150] Step S63: Train all training sets through the training model, and update the weights and biases of the training model through the backpropagation algorithm according to the training results and the loss function of the validation set.

[0151] Step S64: Repeat step S63 several times until the loss function reaches the minimum value.

[0152] Step S65: Obtain the training model corresponding to the minimum value of the loss function and define it as the prediction model.

[0153] Step S66: Input all test sets into the prediction model to obtain a set of future extreme value coordinate points based on one data acquisition cycle.

[0154] In this embodiment, the time required to process a steel raw material into a connector is defined as one data acquisition cycle; external sound pick-up ends at at least two different listening angles are used to repeatedly acquire the processing audio data of several data acquisition cycles; each piece of processing audio data is denoised and interpolated to obtain the processed audio data; amplitude averaging is performed on all the processed audio data of the same data acquisition cycle to obtain a mean audio data based on one data acquisition cycle; the extreme value coordinate points of all extreme points in each mean audio data based on the time process are obtained respectively, and all the extreme value coordinate points of one mean audio data are integrated into a coordinate point data set; a machine learning algorithm is used to learn all the coordinate point data sets and predict at least one future coordinate point data set based on the data acquisition cycle, and each future coordinate point data set includes several future extreme value coordinate points; the Euclidean distance between the extreme value coordinate points and the future extreme value coordinate points at the same time process is obtained; the future extreme value coordinate points with Euclidean distance greater than the preset distance threshold are retained and marked as abnormal extreme value coordinate points; the processing links corresponding to all the abnormal extreme value coordinate points based on the time process are obtained, and the processing links with the corresponding quantity exceeding the preset quantity threshold are defined as the processing fault risk links. This embodiment utilizes the characteristic that the pipeline processing of connectors is a repetitive and regular mechanical reciprocating motion. By taking the processing of a connector from raw material to completion as one processing cycle, even if there are semi-finished products in multiple different stages within one processing cycle, it still conforms to the above characteristics, specifically reflected in the audio superposition of different stages. Moreover, most of the faults of processing equipment are caused by the gradual excessive wear of the machine during a large number of reciprocations, and excessive wear is a linear action that slowly accumulates over time. Therefore, in this embodiment, the audio data of several processing cycles and the predicted audio data are compared with each other, so that excessive wear can be detected in the initial stage, or predicted through the prediction function provided in this embodiment before excessive wear actually occurs, thus filling the gap in the fault detection of workpiece processing equipment in the mountain photovoltaic field. In addition, this embodiment makes up for the audio missing during the sound pick-up process and then performs subsequent judgment operations, so that the fault judgment and fault warning of this embodiment have good accuracy.

[0155] As Figure 2 shown, this embodiment provides an embodiment of the fault warning system for the photovoltaic support connector processing equipment. In this embodiment, the fault warning system is applied to the fault warning method in the above-mentioned embodiment.

[0156] Specifically, the fault warning system includes a data acquisition cycle definition module 1, a processing audio data acquisition module 2, a processing audio data processing module 3, a mean audio data acquisition module 4, an extreme value coordinate point set acquisition module 5, a coordinate point data set prediction module 6, a Euclidean distance acquisition module 7, an abnormal extreme value coordinate marking module 8, and a processing fault risk definition module 9 that are electrically connected in sequence.

[0157] Among them, the data acquisition cycle definition module 1 is used to define the duration required to process a steel raw material into a connector as a data acquisition cycle; the processing audio data acquisition module 2 is used to repeatedly acquire the processing audio data of several data acquisition cycles through external sound pick-up ends at at least two different listening angles; the processing audio data processing module 3 is used to perform denoising and interpolation processing on each processing audio data respectively to obtain the processed audio data; the mean audio data acquisition module 4 is used to perform amplitude averaging on all the processed audio data of the same data acquisition cycle, and obtain a mean audio data based on one data acquisition cycle; the extreme value coordinate point set acquisition module 5 is used to respectively obtain the extreme value coordinate points of all extreme points in each mean audio data based on the time process, and integrate all the extreme value coordinate points of one mean audio data into a coordinate point data set; the coordinate point data set prediction module 6 is used to learn all the coordinate point data sets through machine learning and predict at least one future coordinate point data set based on the data acquisition cycle, and each future coordinate point data set includes several future extreme value coordinate points; the Euclidean distance acquisition module 7 is used to obtain the Euclidean distance between the extreme value coordinate points and the future extreme value coordinate points of the same time process; the abnormal extreme value coordinate marking module 8 is used to retain the future extreme value coordinate points with Euclidean distance greater than the preset distance threshold and mark them as abnormal extreme value coordinate points; the processing fault risk definition module 9 is used to obtain the processing links corresponding to each of all the abnormal extreme value coordinate points based on the time process, and define the processing links with the corresponding quantity exceeding the preset quantity threshold as the processing fault risk links.

[0158] Furthermore, the fault warning system further includes a processing fault risk link number acquisition module, a rectangular coordinate system generation module, a fault number coordinate point conversion module, a fault fitting function acquisition module, a fault fitting function judgment module, and a fatal fault risk determination module that are electrically connected in sequence; the processing fault risk link number acquisition module is electrically connected to the processing fault risk definition module 9.

[0159] Among them, the processing fault risk link number acquisition module is used to acquire the number of processing fault risk links in each data acquisition period respectively; the rectangular coordinate system generation module is used to generate a rectangular coordinate system with the time process of the data acquisition period as the horizontal axis and the value of the number of processing fault risk links as the vertical axis; the fault number coordinate point conversion module is used to convert the current data acquisition period and the corresponding number of processing fault risk links into fault number coordinate points based on the rectangular coordinate system; the fault fitting function acquisition module is used to linearly fit all the fault number coordinate points to obtain a fault fitting function; the fault fitting function judgment module is used to judge whether the fault fitting function is monotonically increasing; the fatal fault risk determination module is used to determine that if so, the photovoltaic support connection part processing equipment generates a fatal fault risk.

[0160] Furthermore, the fault warning system further includes a fault risk sending module and a fatal fault risk sending module that are electrically connected in sequence; the fault risk sending module is electrically connected to the fatal fault risk determination module.

[0161] Among them, the fault risk sending module is used to send all the processing fault risk links and the corresponding future timestamps to an external monitoring terminal; the fatal fault risk sending module is used to send the fatal fault risk and the corresponding future timestamps to the external monitoring terminal.

[0162] Furthermore, the processing audio data processing module 3 specifically includes a first processing audio data processing sub-module, a second processing audio data processing sub-module, a third processing audio data processing sub-module, a fourth processing audio data processing sub-module, and a fifth processing audio data processing sub-module that are electrically connected in sequence; the first processing audio data processing sub-module is electrically connected to the processing audio data acquisition module 2, and the fifth processing audio data processing sub-module is electrically connected to the average audio data acquisition module 4.

[0163] Among them, the first processing audio data processing sub-module is used to perform denoising processing on each processing audio data respectively by improving the spectral subtraction method, and obtain a denoised frequency data based on one processing audio data; the second processing audio data processing sub-module is used to obtain all the data coordinate points and the positions of all missing coordinate points of the current denoised frequency data based on the sound wave spectrum with a preset scale interval; the third processing audio data processing sub-module is used to calculate the coordinate values of the positions of all missing coordinate points by data interpolation method with all the data coordinate points as known quantities; the fourth processing audio data processing sub-module is used to input the coordinate values of all missing coordinate points into the sound wave spectrum to obtain interpolated coordinate points; the fifth processing audio data processing sub-module is used to linearly connect all the coordinate points and all the interpolated coordinate points in sequence to obtain the processed audio data.

[0164] Further, the first processed audio data processing sub-module specifically includes a first processed audio data processing unit, a second processed audio data processing unit, a third processed audio data processing unit, a fourth processed audio data processing unit, a fifth processed audio data processing unit, a sixth processed audio data processing unit, and a seventh processed audio data processing unit that are electrically connected in sequence; the first processed audio data processing unit is electrically connected to the processed audio data acquisition module 2, and the seventh processed audio data processing unit is electrically connected to the second processed audio data processing sub-module.

[0165] Among them, the first processed audio data processing unit is used to divide the current processed audio data into voice segments and silent segments; the second processed audio data processing unit is used to extract pure noise samples in the silent segments and calculate the average spectral value of the pure noise samples through multi-frame averaging; the third processed audio data processing unit is used to equally divide the voice segments into several mixed short-time frames; the fourth processed audio data processing unit is used to perform Fourier transform on each mixed short-time frame respectively, and obtain a mixed spectrum based on one mixed short-time frame; the fifth processed audio data processing unit is used to calculate the denoised spectrum through improved spectral subtraction based on the mixed short-time frame and the mixed spectrum; the sixth processed audio data processing unit is used to inverse-transform all denoised spectra into several denoised short-time frames through inverse Fourier transform; the seventh processed audio data processing unit is used to splice all denoised short-time frames into the denoised audio data corresponding to the current processed audio data along the time series.

[0166] Further, the third processed audio data processing sub-module specifically includes an eighth processed audio data processing unit, a ninth processed audio data processing unit, a tenth processed audio data processing unit, an eleventh processed audio data processing unit, a twelfth processed audio data processing unit, a thirteenth processed audio data processing unit, and a fourteenth processed audio data processing unit that are electrically connected in sequence; the eighth processed audio data processing unit is electrically connected to the second processed audio data processing sub-module, and the fourteenth processed audio data processing unit is electrically connected to the fourth processed audio data processing sub-module.

[0167] Among them, the eighth processed audio data processing unit is used to obtain the Euclidean distance and the semivariogram value between two adjacent data coordinate points respectively; the ninth processed audio data processing unit is used to obtain the fitting curve of all Euclidean distances and all semivariograms, so that the fitting curve can calculate the corresponding semivariogram according to any Euclidean distance; the tenth processed audio data processing unit is used to solve the semivariogram value between any two data coordinate points according to the fitting curve; the eleventh processed audio data processing unit is used to define the optimal coefficient matrix of all semivariograms according to all the solved semivariogram values, and the optimal coefficient matrix includes the semivariogram values to be solved at the positions of the missing coordinate points; the twelfth processed audio data processing unit is used to perform weighted summation on all the semivariogram values according to all the optimal coefficients in the optimal coefficient matrix to obtain the solutions of all the semivariogram values to be solved; the thirteenth processed audio data processing unit is used to substitute the solution of each semivariogram value to be solved into the fitting curve respectively to obtain the Euclidean distance between each missing coordinate point position and other data coordinate points; the fourteenth processed audio data processing unit is used to convert the Euclidean distance between the current missing coordinate point position and other data coordinate points into the coordinate value of the current missing coordinate point position based on multi-point positioning.

[0168] Further, the coordinate point dataset prediction module 6 specifically includes a first coordinate point dataset prediction sub-module, a second coordinate point dataset prediction sub-module, a third coordinate point dataset prediction sub-module, a fourth coordinate point dataset prediction sub-module, a fifth coordinate point dataset prediction sub-module, and a sixth coordinate point dataset prediction sub-module that are electrically connected in sequence; the first coordinate point dataset prediction sub-module is electrically connected to the extreme value coordinate point set acquisition module 5, and the sixth coordinate point dataset prediction sub-module is electrically connected to the Euclidean distance acquisition module 7.

[0169] Among them, the first coordinate point dataset prediction sub-module is used to perform normalization processing on each coordinate point dataset respectively to obtain a normalized dataset based on one coordinate point dataset; the second coordinate point dataset prediction sub-module is used to divide a normalized dataset into a training set, a validation set, and a test set according to a preset ratio; the third coordinate point dataset prediction sub-module is used to train all the training sets through a training model, and update the weights and biases of the training model through the backpropagation algorithm according to the training results and the loss function of the validation set; the fourth coordinate point dataset prediction sub-module is used to repeat the third coordinate point dataset prediction sub-module several times until the loss function reaches the minimum value; the fifth coordinate point dataset prediction sub-module is used to obtain the training model corresponding to the minimum value of the loss function and define it as the prediction model; the sixth coordinate point dataset prediction sub-module is used to input all the test sets into the prediction model to obtain a set of future extreme value coordinate points based on one data acquisition cycle.

[0170] It should be noted that this embodiment is a functional module embodiment based on the above method embodiment. For the preferred, extended, limited, exemplified, and principle description parts of this embodiment, please refer to the above embodiments, and this embodiment will not be elaborated here.

[0171] In this embodiment, the duration required to process a steel raw material into a connector is defined as a data acquisition cycle; the processing audio data of several data acquisition cycles is repeatedly acquired through external sound pickup terminals at at least two different listening angles; each piece of processing audio data is respectively denoised and interpolated to obtain processed audio data; amplitude averaging is performed on all the processed audio data of the same data acquisition cycle, and a mean audio data is obtained based on one data acquisition cycle; the extreme value coordinate points of all extreme points in each mean audio data based on the time process are respectively obtained, and all the extreme value coordinate points of one mean audio data are integrated into a coordinate point data set; all the coordinate point data sets are learned through a machine learning machine and at least one future coordinate point data set is predicted based on the data acquisition cycle, and each future coordinate point data set includes several future extreme value coordinate points; the Euclidean distance between the extreme value coordinate points and the future extreme value coordinate points at the same time process is obtained; the future extreme value coordinate points with the Euclidean distance greater than the preset distance threshold are retained and marked as abnormal extreme value coordinate points; the processing links corresponding to all the abnormal extreme value coordinate points based on the time process are obtained, and the processing links with the corresponding quantity exceeding the preset quantity threshold are defined as processing fault risk links. This embodiment utilizes the characteristic that the pipeline processing of the connector is a repetitive and regular mechanical reciprocating motion. By taking one connector from raw material to completed processing as a processing cycle, even if there are semi-finished products in multiple different stages within one processing cycle, it conforms to the above characteristics, specifically reflected in the audio superposition of different stages. Moreover, most of the faults of the processing equipment are the gradual excessive wear of the machine during a large number of reciprocating processes, and the excessive wear is a linear action that slowly superimposes over time. Therefore, in this embodiment, the audio data of several processing cycles and the predicted audio data are compared with each other, so that the excessive wear can be detected in the initial stage, or can be predicted through the prediction function provided by this embodiment before the excessive wear actually occurs, thus making up for the gap in the fault detection of workpiece processing equipment in the mountain photovoltaic field. And this embodiment makes up for the audio missing during the sound pickup process and then performs subsequent judgment operations, making the fault judgment and fault warning of this embodiment have good accuracy.

[0172] As Figure 3 shown, this embodiment provides an embodiment of an electronic device. In this embodiment, the electronic device 10 includes a processor 101 and a memory 102 coupled to the processor 101.

[0173] The memory 102 stores program instructions for implementing the fault warning method of the photovoltaic support connector processing equipment in any of the above embodiments.

[0174] The processor 101 is used to execute the program instructions stored in the memory 102 for fault warning of the photovoltaic support connecting piece processing equipment.

[0175] Among them, the processor 101 can also be called a CPU (Central Processing Unit). The processor 101 may be an integrated circuit chip with data processing capabilities. The processor 101 can also be a general-purpose processor, a digital data processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc.

[0176] Furthermore, Figure 4 FIG. is a schematic structural diagram of a storage medium according to an embodiment of the present application. The storage medium 11 of the embodiment of the present application stores program instructions 111 that can implement all the above methods. Among them, the program instructions 111 can be stored in the above storage medium in the form of a software product, including several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor to execute all or part of the steps of the methods described in various embodiments of the present application. The foregoing storage medium includes: various media that can store program codes such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disc, or a terminal device such as a computer, a server, a mobile phone, or a tablet.

[0177] In several embodiments provided by the present application, it should be understood that the disclosed system, system, and method can be implemented in other ways. For example, the system embodiments described above are merely illustrative. For example, the division of units is only a logical function division, and there may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed mutual coupling or direct coupling or communication connection may be through some interfaces, and the indirect coupling or communication connection of the system or unit may be in an electrical, mechanical, or other form.

[0178] In addition, in each embodiment of the present application, each functional unit can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of a software functional unit. The above is only the implementation mode of the present application, and does not limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made by using the content of the specification and drawings of the present application, or directly or indirectly applied in other related technical fields, shall be included in the patent protection scope of the present application by the same token.

[0179] The specific implementation manners of the present application have been described in detail above, but they are only examples, and the present application is not limited to the specific implementation manners described above. For those skilled in the art, any equivalent modification or substitution to the present application is also within the scope of the present application. Therefore, all equal transformations, modifications, improvements, etc. made without departing from the spirit and principle scope of the present application should be covered within the scope of the present application.

Claims

1. A fault warning method for a processing device of a photovoltaic support connector. The photovoltaic support connector processing device includes a punching end, at least one bending end, and a shearing end that are connected in a production line or integrated linearly. Steel raw materials pass through the punching end, all the bending ends, and the shearing end in sequence to form connectors. It is characterized in that, The described fault warning method includes: Step S1, defining the duration required to process a steel raw material into a connector as a data acquisition cycle; Step S2, repeatedly acquiring the processing audio data of several data acquisition cycles through external sound pick-up ends with at least two different listening angles; Step S3, respectively performing denoising and interpolation processing on each processing audio data to obtain processed audio data; Step S4, performing amplitude averaging processing on all the processed audio data of the same data acquisition cycle, and obtaining a mean audio data based on one data acquisition cycle; Step S5, respectively obtaining the extreme value coordinate points of all extreme points in each mean audio data based on the time process, and integrating all the extreme value coordinate points of one mean audio data into a coordinate point data set; Step S6, learning all the coordinate point data sets through a machine learning machine and predicting at least one future coordinate point data set based on the data acquisition cycle, and each future coordinate point data set includes several future extreme value coordinate points; Step S7, obtaining the Euclidean distance between the extreme value coordinate points and the future extreme value coordinate points at the same time process; Step S8, retaining the future extreme value coordinate points with the Euclidean distance greater than the preset distance threshold and marking them as abnormal extreme value coordinate points; Step S9, obtaining the processing links corresponding to all the abnormal extreme value coordinate points based on the time process respectively, and defining the processing links with the corresponding quantity exceeding the preset quantity threshold as the processing fault risk links.

2. The fault warning method according to claim 1, characterized in that Step S9, obtaining the processing links corresponding to all the abnormal extreme value coordinate points based on the time process respectively, and defining the processing links with the corresponding quantity exceeding the preset quantity threshold as the processing fault risk links. After that, it includes: Step S10, respectively obtaining the number of processing fault risk links in each data acquisition cycle; Step S20, generating a rectangular coordinate system with the time process of the data acquisition cycle as the horizontal axis and the value of the number of processing fault risk links as the vertical axis; Step S30, converting the current data acquisition cycle and the corresponding number of processing fault risk links into fault number coordinate points based on the rectangular coordinate system; Step S40, linearly fitting all the fault number coordinate points to obtain a fault fitting function; Step S50, judging whether the fault fitting function is monotonically increasing. If so, execute Step S60; Step S60, determining that the processing equipment of the photovoltaic bracket connector has a fatal fault risk.

3. The fault warning method according to claim 2, wherein, Step S60, determining that the processing equipment of the photovoltaic bracket connector has a fatal fault risk. After that, it includes: Step S100, sending all the processing fault risk links and the corresponding future timestamps to an external monitoring end; Step S200, sending the fatal fault risk and the corresponding future timestamps to an external monitoring end.

4. The fault warning method according to claim 1, wherein Step S3, respectively performing denoising and interpolation processing on each processing audio data to obtain processed audio data, including: Step S31, respectively performing denoising processing on each processing audio data through an improved spectral subtraction method, and obtaining a denoised audio data based on one processing audio data; Step S32, obtaining all the data coordinate points and the positions of all the missing coordinate points of the current denoised audio data based on the sound wave spectrum with a preset scale interval; Step S33: Calculate the coordinate values of all missing coordinate point positions by data interpolation method with all data coordinate points as known quantities; Step S34: Input the coordinate values of all missing coordinate points into the acoustic wave spectrum to obtain interpolated coordinate points; Step S35: Sequentially and linearly connect all coordinate points and all interpolated coordinate points to obtain the processed audio data.

5. The fault warning method according to claim 4, wherein Step S31: Perform denoising processing on each processed audio data respectively by improved spectral subtraction, and obtain a denoised frequency data based on one processed audio data, including: Step S311: Divide the current processed audio data into voice segments and silent segments; Step S312: Extract pure noise samples in the silent segments, and calculate the average spectral value of the pure noise samples by multi-frame averaging; Step S313: Divide the voice segments into several mixed short-time frames equally; Step S314: Perform Fourier transform on each mixed short-time frame respectively, and obtain a mixed frequency spectrum based on one mixed short-time frame; Step S315: Calculate the denoised frequency spectrum by improved spectral subtraction based on the mixed short-time frame and the mixed frequency spectrum; Step S316: Inversely transform all denoised frequency spectra into several denoised short-time frames by inverse Fourier transform; Step S317: Concatenate all denoised short-time frames along the time series to obtain the denoised frequency data corresponding to the current processed audio data.

6. The fault warning method according to claim 4, characterized in that Step S33: Calculate the coordinate values of all missing coordinate point positions by data interpolation method with all data coordinate points as known quantities, including: Step S331: Obtain the Euclidean distance and semi-variogram value between two adjacent data coordinate points respectively; Step S332: Obtain the fitting curve of all Euclidean distances and all semi-variograms, so that the fitting curve calculates the corresponding semi-variogram according to any Euclidean distance; Step S333: Solve the semi-variogram value between any two data coordinate points according to the fitting curve; Step S334: Define the optimal coefficient matrix of all semi-variograms according to all solved semi-variogram values, and the optimal coefficient matrix includes the semi-variogram values to be solved at the positions of missing coordinate points; Step S335: Perform weighted summation on all semi-variogram values according to all optimal coefficients in the optimal coefficient matrix to obtain the solutions of all semi-variogram values to be solved; Step S336: Substitute the solution of each semi-variogram value to be solved into the fitting curve respectively to obtain the Euclidean distance between each missing coordinate point position and other data coordinate points; Step S337: Convert the Euclidean distance between the current missing coordinate point position and other data coordinate points into the coordinate value of the current missing coordinate point position based on multi-point positioning.

7. The fault warning method according to claim 1, wherein Step S6: Learn all coordinate point data sets through machine learning and predict at least one future coordinate point data set based on the data acquisition period. Each future coordinate point data set includes several future extreme value coordinate points, including: Step S61: Perform normalization processing on each coordinate point data set respectively, and obtain a normalized data set based on one coordinate point data set; Step S62: Divide a normalized data set into a training set, a validation set, and a test set according to a preset ratio. Step S63: Train all training sets through a training model, and update the weights and biases of the training model through the backpropagation algorithm according to the training results and the loss function of the validation set. Step S64: Repeat Step S63 several times until the loss function reaches the minimum value. Step S65: Obtain the training model corresponding to the minimum value of the loss function and define it as a prediction model. Step S66: Input all test sets into the prediction model to obtain a set of future extreme value coordinate points based on one data acquisition cycle.

8. A fault warning system for a processing device of a photovoltaic bracket connector, the fault warning system being applied to the fault warning method according to any one of claims 1 to 7, characterized in that, The fault warning system includes: A data acquisition cycle definition module for defining the duration required to process a steel raw material into a connector as one data acquisition cycle. A processed audio data acquisition module for repeatedly acquiring processed audio data for several data acquisition cycles through external sound pick-up ends at at least two different listening angles. A processed audio data processing module for denoising and interpolating each processed audio data to obtain processed audio data. A mean audio data acquisition module for performing amplitude averaging on all processed audio data in the same data acquisition cycle to obtain a mean audio data based on one data acquisition cycle. An extreme value coordinate point set acquisition module for respectively obtaining the extreme value coordinates of all extreme points in each mean audio data based on the time process, and integrating all extreme value coordinates of one mean audio data into a coordinate point data set. A coordinate point data set prediction module for learning all coordinate point data sets through machine learning and predicting at least one future coordinate point data set based on the data acquisition cycle, where each future coordinate point data set includes several future extreme value coordinate points. An Euclidean distance acquisition module for obtaining the Euclidean distance between the extreme value coordinates and the future extreme value coordinates at the same time process. An abnormal extreme value coordinate marking module for retaining the future extreme value coordinates with the Euclidean distance greater than the preset distance threshold and marking them as abnormal extreme value coordinates. A processing fault risk definition module for obtaining the processing links corresponding to all abnormal extreme value coordinates based on the time process, and defining the processing links with the corresponding quantity exceeding the preset quantity threshold as processing fault risk links.

9. An electronic device, characterized in that, It includes a processor and a memory coupled to the processor, and the memory stores program instructions executable by the processor; when the processor executes the program instructions stored in the memory, it implements the fault warning method according to any one of claims 1 to 7.

10. A storage medium, characterized in that, The storage medium stores program instructions, and when the program instructions are executed by the processor, they can implement the fault warning method according to any one of claims 1 to 7.

Images