CN112101301A - A kind of good sound stability early warning method, device and storage medium of screw water-cooling unit - Google Patents

Abstract

The invention provides a good sound stability early warning method, device and storage medium for a screw water-cooled unit. The method includes: importing a plurality of original audio data, and performing data cleaning on the plurality of original audio data, and the remaining original audio data after data cleaning The audio data is used as the target audio data to obtain a plurality of target audio data, and the original audio data is obtained by the screw water-cooling unit equipment; the feature extraction is performed on each of the target audio data respectively to obtain the corresponding feature points, and the collection is extracted. All the feature points get the feature point dataset. The present invention can further improve the accuracy of audio recognition, overcome the defects of large workload, low efficiency and insufficient accuracy of traditional good-sound stability early warning, and can automatically perform intelligent detection and recognition on a large amount of audio data, and real-time detected audio It has the characteristics of high efficiency, strong stability and high accuracy.

Description

A kind of good sound stability early warning method, device and storage medium of screw water-cooling unit

technical field

The invention mainly relates to the technical field of audio recognition, and in particular relates to a good sound stability early warning method, device and storage medium for a screw water cooling unit.

Background technique

Good sound stability warning is essentially a pattern recognition technology. In real life applications, people's intuitive sensory recognition of audio is based on the semantic layer of the highest level of audio, and audio is easily affected by the environment and transmission conversion equipment.

Due to the risk of dehumanization, remote control, low accuracy and complexity of audio data, it is not very reliable. There are many other factors that affect its accuracy, such as the quality of the sound sample, mood, background noise, and changes in sound over time. At present, it is still difficult to rely on the automatic identification and identification of instruments, and it is impossible to guarantee a high identification accuracy.

SUMMARY OF THE INVENTION

The technical problem to be solved by the present invention is to aim at the deficiencies of the prior art, and to provide a good sound stability early warning method, device and storage medium for a screw water cooling unit.

The technical solution of the present invention to solve the above-mentioned technical problems is as follows: a good sound stability early warning method for a screw water-cooled unit, comprising the following steps:

Import a plurality of original audio data, and perform data cleaning on a plurality of the original audio data. After the data cleaning, the remaining original audio data is used as the target audio data to obtain a plurality of target audio data. The original audio data is cooled by screw water. obtained by the unit equipment;

Perform feature extraction on each of the target audio data respectively to obtain corresponding feature points, and collect all the extracted feature points to obtain a feature point data set;

constructing a training model, and training the training model according to the feature point data set to obtain an audio recognition model;

Optimizing the audio recognition model to obtain an audio recognition optimization model;

The audio data to be recognized is recognized and processed according to the audio recognition optimization model to obtain a recognition result of the audio data.

Another technical solution of the present invention to solve the above-mentioned technical problems is as follows: a good sound stability early warning device for a screw water-cooled unit, comprising:

The data cleaning module is used to import a plurality of original audio data, and perform data cleaning on the plurality of said original audio data, and the remaining original audio data after data cleaning is used as target audio data to obtain a plurality of target audio data. Audio data is obtained through screw water cooling unit equipment;

a feature extraction module, which is used to perform feature extraction on each of the target audio data respectively to obtain corresponding feature points, and collect all the extracted feature points to obtain a feature point data set;

a model training module for constructing a training model, and training the training model according to the feature point data set to obtain an audio recognition model;

an optimization processing module for performing optimization processing on the audio recognition model to obtain an audio recognition optimization model;

A good sound stability early warning module is used for identifying and processing the audio data to be recognized according to the audio recognition optimization model to obtain a recognition result of the audio data.

Another technical solution of the present invention to solve the above-mentioned technical problems is as follows: a good sound stability early warning device for a screw water cooling unit, comprising a memory, a processor and a computer program stored in the memory and running on the processor, When the processor executes the computer program, the above-mentioned method for early warning of good sound stability of the screw water cooling unit is realized.

Another technical solution of the present invention to solve the above technical problem is as follows: a computer-readable storage medium, the computer-readable storage medium stores a computer program, when the computer program is executed by a processor, the above-mentioned screw is realized Good sound stability early warning method for water-cooled units.

The beneficial effects of the present invention are: by cleaning the data of a plurality of original audio data to obtain a plurality of target audio data, the data with missing values can be filtered out, and the data with more useful information can be analyzed and the impact on identification and identification can be analyzed. With larger information, it is convenient to make a data set according to the obtained target audio data, so as to obtain a recognition model with higher recognition and recognition accuracy, and extract the features of multiple target audio data respectively to obtain a feature point data set and build a training model. And the audio recognition model is obtained by training the training model according to the feature point data set, which can effectively improve the reliability and stability of the good sound stability early warning. The audio recognition optimization model is obtained by optimizing the audio recognition model according to the preset adjustment parameters. The audio recognition optimization model is to process the recognition audio data to obtain the recognition result of the audio data, which can further improve the audio recognition accuracy, and overcome the defects of large workload, low efficiency and insufficient accuracy of the traditional good sound stability early warning. Intelligent detection and identification of a large amount of audio data, real-time detection of audio data, timely intervention for good sound stability early warning, with high efficiency, strong stability and high accuracy.

Description of drawings

Fig. 1 is the schematic flow chart of the good sound stability early warning method of the screw water-cooled unit provided by the embodiment of the present invention;

FIG. 2 is a block diagram of a module of a good sound stability early warning device of a screw water cooling unit provided by an embodiment of the present invention.

Detailed ways

The principles and features of the present invention will be described below with reference to the accompanying drawings. The examples are only used to explain the present invention, but not to limit the scope of the present invention.

FIG. 1 is a schematic flowchart of a good sound stability early warning method for a screw water-cooled unit provided by an embodiment of the present invention.

As shown in Figure 1, a good sound stability early warning method for a screw water-cooled unit includes the following steps:

Import a plurality of original audio data, and perform data cleaning on a plurality of the original audio data. After the data cleaning, the remaining original audio data is used as the target audio data to obtain a plurality of target audio data. The original audio data is cooled by screw water. obtained by the unit equipment;

Perform feature extraction on each of the target audio data respectively to obtain corresponding feature points, and collect all the extracted feature points to obtain a feature point data set;

constructing a training model, and training the training model according to the feature point data set to obtain an audio recognition model;

Optimizing the audio recognition model to obtain an audio recognition optimization model;

The audio data to be recognized is recognized and processed according to the audio recognition optimization model to obtain a recognition result of the audio data.

It should be understood that data cleaning refers to the final process of finding and correcting identifiable errors in data files, including checking data consistency, handling invalid and missing values, etc.

It should be understood that the audio recognition model is optimized according to the preset adjustment parameters to obtain an audio recognition optimization model.

Specifically, using the manual manual parameter tuning method to perform parameter tuning can ensure that the optimal parameters corresponding to the audio recognition model can be obtained, further ensure the audio recognition accuracy of the audio recognition optimization model, and detect audio samples in real time. The audio data of the discharge can be used for clinical intervention in a timely manner.

In the above-mentioned embodiment, by cleaning the data of a plurality of original audio data to obtain a plurality of target audio data, the data with missing values can be filtered out, and the data with more useful information can also be analyzed, which has a greater impact on the identification and identification. It is convenient to make a data set according to the obtained target audio data, so as to obtain a recognition model with higher recognition and recognition accuracy, extract the features of multiple target audio data respectively, obtain a feature point data set, build a training model and according to The feature point dataset is used to train the training model to obtain an audio recognition model, which can effectively improve the reliability and stability of the stable early warning of good sounds. The optimized model processes the audio data to be recognized to obtain the audio data recognition results, which can further improve the audio recognition accuracy, overcome the defects of large workload, low efficiency and insufficient accuracy of the traditional good sound stability early warning, and can automatically detect a large number of The audio data is intelligently detected and recognized, and the audio data detected in real time can be intervened in time for good sound stability early warning, which has the characteristics of high efficiency, strong stability and high accuracy.

Optionally, as an embodiment of the present invention, the process of performing feature extraction on each of the target audio data to obtain corresponding feature points includes:

Respectively perform dimension reduction processing on a plurality of the target audio data to obtain a plurality of dimension-reduced audio data;

Feature extraction is performed on a plurality of the dimensionality-reduced audio data by using a preset spectrogram to obtain a plurality of feature points.

It should be understood that in the field of machine learning and statistics, dimensionality reduction refers to the process of reducing the number of random variables to obtain a set of "uncorrelated" main variables under certain limited conditions. On the one hand, dimensionality reduction of data can save the storage space of the computer, on the other hand, it can eliminate the noise in the data and improve the performance of the machine learning algorithm; represents the original data.

It should be understood that the LDA linear discriminant analysis algorithm is used to perform dimensionality reduction processing on a plurality of the target audio data respectively to obtain a plurality of dimensionally reduced audio data.

Specifically, a plurality of the target audio data are projected on low dimensions, so that the audio text is reduced from thousands of dimensions to k dimensions, so that the intra-class variance of the projected target audio data is minimized and the inter-class variance is maximized ; compress the target audio data reduced in dimension to a new feature space, and retain information as much as possible to obtain a plurality of audio data after the dimension reduction.

It should be understood that the spectrogram is a display image of time-series-related Fourier analysis, which can reflect the transformation of the spectrum of the music signal with time. The abscissa of the spectrogram is time, the ordinate is frequency, and the coordinate point value. is the voice data energy. Since the two-dimensional plane is used to express three-dimensional information, the energy value is represented by color. The darker the color, the stronger the speech energy of the point.

The spectrogram shows a lot of information related to the characteristics of the music signal, such as the variation of frequency domain parameters such as formants and energy over time. It has the characteristics of time domain waveform and spectrogram at the same time. That is to say, the spectrogram itself contains all the spectral information of the music signal without any processing, so the information about the music of the spectrogram is lossless.

The patterns in the spectrogram include horizontal lines, random patterns and vertical bars. The horizontal lines are black stripes parallel to the time axis. They are formants. The corresponding formant frequency can be determined from the frequency and bandwidth corresponding to the horizontal lines. and bandwidth, whether there is a horizontal line in the spectrogram of a piece of audio is an important sign to judge whether it is a voiced sound; the vertical bar is a narrow black bar perpendicular to the time axis, each vertical bar is equivalent to a pitch, and the stripe The starting point is equivalent to the starting point of the voiceprint pulse, the distance between the stripes represents the pitch, and the denser the stripes, the higher the pitch frequency.

In the above embodiment, the dimensionality reduction processing of multiple target audio data is performed to obtain multiple dimensionally reduced audio data, and the preset spectrogram is used to extract the features of the multiple dimensionally reduced audio data to obtain multiple features. It is convenient to obtain the main features that have a greater impact on the sound stability and early warning of the screw water-cooled unit, reducing the amount of computation in the subsequent steps, and the subsequent support vector machine only needs less training data to obtain higher accuracy.

Optionally, as an embodiment of the present invention, the process of training the training model according to the feature point data set to obtain an audio recognition model includes:

S1: randomly divide the feature point data set into a feature point training set and a feature point test set;

S2: build a model based on the support vector machine algorithm, and obtain the support vector machine structure;

S3: Input the feature point training set and the feature point test set together into the support vector machine structure to search for the optimal recognition hyperplane, obtain the optimal recognition hyperplane, and according to the optimal recognition hyperplane Get support vector set and VC credibility;

S4: carry out discriminant processing on the support vector set according to the VC credibility, obtain a discriminant function, and generate a training model according to the discriminant function;

S5: Perform model screening processing on the training model according to the feature point training set and the feature point test set to obtain an audio recognition model.

It should be understood that since the feature point data set is randomly divided into the feature point training set and the feature point test set each time, the random proportions are different each time, and the train_test_split function can be called to perform random division.

Specifically, Support Vector Machine (SVM for short) is a supervised learning algorithm. SVM theory provides a way to avoid the complexity of high-dimensional space and directly use the inner product function of this space (that is, a kernel function) , and then use the solution method in the case of linear separability to directly solve the decision problem in the corresponding high-dimensional space. When the kernel function is known, the difficulty of solving the high-dimensional space problem can be simplified. At the same time, SVM is based on the basis of small sample statistical theory Above, this is in line with the purpose of machine learning. And support vector machine has better generalization ability than neural network, good robustness to noisy data and data with missing values, and has a faster learning depth , which combines the advantages of a variety of machine recognition algorithms, with high performance and strong stability.

It should be understood that the VC credibility is the VC dimension confidence or confidence risk. The so-called VC dimension is a measure of the function class, which can be simply understood as the complexity of the problem. The higher the VC dimension, the more complex a problem. . For example, many classification functions can easily achieve 100% accuracy on the sample set, but they are messed up in the real classification (that is, the so-called poor generalization ability, or poor generalization ability). The situation at this time is that a sufficiently complex classification function is selected (that is, its VC dimension is high), which can accurately remember each sample, but the data outside the sample is always classified incorrectly.

Specifically, both the feature point training set and the feature point test set are input into the support vector machine structure, and the sample feature space of the support vector machine is used to find out the optimal recognition performance of each category of feature samples and other feature samples. plane, obtain the support vector set representing each category of samples and the corresponding VC reliability, form the discriminant function for discriminating individual features, and obtain the training model.

In the above embodiment, the feature point data set is randomly divided into the feature point training set and the feature point test set, which can ensure the objectivity of the data, reduce human factors, effectively improve the accuracy of the subsequent identification model, and divide the feature point training set. Enter the optimal recognition hyperplane of the support vector machine structure together with the feature point test set to find the optimal recognition hyperplane, and obtain the support vector set and VC reliability according to the optimal recognition hyperplane, and according to the VC reliability, support The training model is obtained by the discriminative processing of the vector set; the audio recognition model is obtained by the model screening of the training model according to the feature point training set and the feature point test set, which can keep the accuracy of the audio data to be recognized at a high level and improve the audio frequency. Stability and reliability of identification.

Optionally, as an embodiment of the present invention, the process of step S5 includes:

S51: Input the feature point training set into the training model for iterative training according to a preset number of iterative training times to obtain a first detection model;

S52: Input the feature point test set into the first detection model for detection to obtain a first accuracy rate, and determine whether the first accuracy rate reaches a preset expected value, and if so, apply the first detection rate The model is used as an audio recognition model, and then the audio recognition model is optimized, and if not, step S53 is performed;

S53: Input the feature point test set into the training model for iterative training according to the preset iterative training times to obtain a second detection model;

S54: Input the feature point training set into the second detection model for detection to obtain a second accuracy rate;

S55: Determine whether the second accuracy rate reaches the preset expected value, if so, use the second detection model as the audio recognition model, and then perform optimization processing on the audio recognition model, if not, then Return to the step S1.

Preferably, the preset expected value is 0.90.

In the above-mentioned embodiment, the obtained first detection model and the second detection model can ensure a high recognition accuracy rate, and obtain an audio recognition model that meets the expectations, when the first accuracy rate of the first detection model does not reach the expected value. When , the second detection model is obtained by training the test set of feature points, and the second accuracy rate is obtained by using the training set of feature points, which is equivalent to exchanging the training set and the test set, which can further ensure that the expected recognition model can be obtained. The accuracy of detecting the audio data to be recognized by the audio recognition model corresponding to the expected value is kept at a high level, and the stability and reliability of audio recognition are improved.

FIG. 2 is a block diagram of a module of a good sound stability early warning device of a screw water cooling unit provided by an embodiment of the present invention.

Optionally, as another embodiment of the present invention, as shown in Figure 2, a good sound stability early warning device for a screw water-cooled unit includes:

The data cleaning module is used to import a plurality of original audio data, and perform data cleaning on the plurality of said original audio data, and the remaining original audio data after data cleaning is used as target audio data to obtain a plurality of target audio data. Audio data is obtained through screw water cooling unit equipment;

a feature extraction module, which is used to perform feature extraction on each of the target audio data respectively to obtain corresponding feature points, and collect all the extracted feature points to obtain a feature point data set;

a model training module for constructing a training model, and training the training model according to the feature point data set to obtain an audio recognition model;

an optimization processing module for performing optimization processing on the audio recognition model to obtain an audio recognition optimization model;

A good sound stability early warning module is used for identifying and processing the audio data to be recognized according to the audio recognition optimization model to obtain a recognition result of the audio data.

Optionally, as an embodiment of the present invention, the feature extraction module is specifically used for:

Respectively perform dimension reduction processing on a plurality of the target audio data to obtain a plurality of dimension-reduced audio data;

Feature extraction is performed on a plurality of the dimensionality-reduced audio data by using a preset spectrogram to obtain a plurality of feature points.

Optionally, as an embodiment of the present invention, the model training module is specifically used for:

Randomly dividing the feature point data set into a feature point training set and a feature point test set;

Build a model based on the support vector machine algorithm, and get the support vector machine structure;

Input the feature point training set and the feature point test set together into the SVM structure to search for the optimal recognition hyperplane, obtain the optimal recognition hyperplane, and obtain support according to the optimal recognition hyperplane vector set and VC credibility;

According to the VC credibility, the support vector set is discriminated to obtain a discriminant function, and a training model is generated according to the discriminant function;

Perform model screening processing on the training model according to the feature point training set and the feature point test set to obtain an audio recognition model.

Optionally, as an embodiment of the present invention, the model training module is specifically used for:

Inputting the feature point training set into the training model for iterative training according to a preset number of iterative training times to obtain a first detection model;

Input the feature point test set into the first detection model for detection, obtain a first accuracy rate, and determine whether the first accuracy rate reaches a preset expected value, and if so, use the first detection model as audio recognition model, and then perform optimization processing on the audio recognition model, if not, input the feature point test set into the training model for iterative training according to the preset iterative training times to obtain a second detection model;

Inputting the feature point training set into the second detection model for detection to obtain a second accuracy rate;

Determine whether the second accuracy rate reaches the preset expected value, if so, use the second detection model as the audio recognition model, and then perform optimization processing on the audio recognition model, if not, use the audio recognition model again. The feature point data set is randomly divided into a feature point training set and a feature point test set.

Optionally, another embodiment of the present invention provides a good sound stability early warning device for a screw water cooling unit, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, when When the processor executes the computer program, the above-mentioned good-sound stability early warning method of the screw water-cooling unit is realized. The device can be a computer or the like.

Optionally, another embodiment of the present invention provides a computer-readable storage medium, the computer-readable storage medium stores a computer program, when the computer program is executed by the processor, the above-mentioned screw water cooling is realized Good sound stability early warning method for the crew.

Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the above-described devices and units may refer to the corresponding processes in the foregoing method embodiments, which will not be repeated here.

In the several embodiments provided in this application, it should be understood that the disclosed apparatus and method may be implemented in other manners. For example, the apparatus embodiments described above are only illustrative. For example, the division of units is only a logical function division. In actual implementation, there may be other division methods, for example, multiple units or components may be combined or integrated. to another system, or some features can be ignored, or not implemented.

Units described as separate components may or may not be physically separated, and components shown as units may or may not be physical units, that is, may be located in one place, or may be distributed to multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solutions in the embodiments of the present invention.

In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically alone, or two or more units may be integrated into one unit. The above-mentioned integrated units may be implemented in the form of hardware, or may be implemented in the form of software functional units.

The integrated unit, if implemented as a software functional unit and sold or used as a stand-alone product, may be stored in a computer-readable storage medium. For such understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage device. The medium includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods of various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, Read-Only Memory (ROM, Read-Only Memory), Random Access Memory (RAM, Random Access Memory), magnetic disk or optical disk and other media that can store program codes .

The above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of various equivalent modifications or modifications within the technical scope disclosed by the present invention. Replacement, these modifications or replacements should all be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

Claims (10)

1. a good-sound stability early-warning method for screw water-cooling unit, is characterized in that, comprises the steps: Import a plurality of original audio data, and perform data cleaning on a plurality of the original audio data. After the data cleaning, the remaining original audio data is used as the target audio data to obtain a plurality of target audio data. The original audio data is cooled by screw water. obtained by the unit equipment; Perform feature extraction on each of the target audio data respectively to obtain corresponding feature points, and collect all the extracted feature points to obtain a feature point data set; constructing a training model, and training the training model according to the feature point data set to obtain an audio recognition model; Optimizing the audio recognition model to obtain an audio recognition optimization model; The audio data to be recognized is recognized and processed according to the audio recognition optimization model to obtain a recognition result of the audio data. 2. The good-sound stability early-warning method of the screw water-cooled unit according to claim 1, wherein the feature extraction is carried out to each of the target audio data respectively, and the process of obtaining the corresponding feature points comprises: Respectively perform dimension reduction processing on a plurality of the target audio data to obtain a plurality of dimension-reduced audio data; Feature extraction is performed on a plurality of the dimensionality-reduced audio data by using a preset spectrogram to obtain a plurality of feature points. 3. the good sound stability early warning method of screw water-cooling unit according to claim 1, is characterized in that, described training model is trained according to described feature point data set, the process that obtains audio frequency recognition model comprises: S1: randomly divide the feature point data set into a feature point training set and a feature point test set; S2: build a model based on the support vector machine algorithm, and obtain the support vector machine structure; S3: Input the feature point training set and the feature point test set together into the support vector machine structure to search for the optimal recognition hyperplane, obtain the optimal recognition hyperplane, and according to the optimal recognition hyperplane Get support vector set and VC credibility; S4: carry out discriminant processing on the support vector set according to the VC credibility, obtain a discriminant function, and generate a training model according to the discriminant function; S5: Perform model screening processing on the training model according to the feature point training set and the feature point test set to obtain an audio recognition model. 4. The good sound stability early warning method of the screw water-cooled unit according to claim 3, is characterized in that, the process of described step S5 comprises: S51: Input the feature point training set into the training model for iterative training according to a preset number of iterative training times to obtain a first detection model; S52: Input the feature point test set into the first detection model for detection to obtain a first accuracy rate, and determine whether the first accuracy rate reaches a preset expected value, and if so, apply the first detection rate The model is used as an audio recognition model, and then the audio recognition model is optimized, and if not, step S53 is performed; S53: Input the feature point test set into the training model for iterative training according to the preset iterative training times to obtain a second detection model; S54: Input the feature point training set into the second detection model for detection to obtain a second accuracy rate; S55: Determine whether the second accuracy rate reaches the preset expected value, if so, use the second detection model as the audio recognition model, and then perform optimization processing on the audio recognition model, if not, then Return to the step S1. 5. A good-sound stability early-warning device for a screw water-cooled unit is characterized in that, comprising: The data cleaning module is used to import a plurality of original audio data, and perform data cleaning on the plurality of said original audio data, and the remaining original audio data after data cleaning is used as target audio data to obtain a plurality of target audio data. Audio data is obtained through screw water cooling unit equipment; a feature extraction module, configured to perform feature extraction on each of the target audio data respectively to obtain corresponding feature points, and collect all the extracted feature points to obtain a feature point data set; a model training module for constructing a training model, and training the training model according to the feature point data set to obtain an audio recognition model; an optimization processing module for performing optimization processing on the audio recognition model to obtain an audio recognition optimization model; A good sound stability early warning module is used for identifying and processing the audio data to be recognized according to the audio recognition optimization model to obtain a recognition result of the audio data. 6. The good sound stability early warning device of the screw water-cooled unit according to claim 5, is characterized in that, described feature extraction module is specifically used for: Respectively perform dimension reduction processing on a plurality of the target audio data to obtain a plurality of dimension-reduced audio data; Feature extraction is performed on a plurality of the dimensionality-reduced audio data by using a preset spectrogram to obtain a plurality of feature points. 7. The good sound stability early warning device of the screw water-cooled unit according to claim 5, is characterized in that, described model training module is specifically used for: Randomly dividing the feature point data set into a feature point training set and a feature point test set; Build a model based on the support vector machine algorithm, and get the support vector machine structure; Input the feature point training set and the feature point test set together into the SVM structure to search for the optimal recognition hyperplane, obtain the optimal recognition hyperplane, and obtain support according to the optimal recognition hyperplane vector set and VC credibility; According to the VC credibility, the support vector set is discriminated to obtain a discriminant function, and a training model is generated according to the discriminant function; Perform model screening processing on the training model according to the feature point training set and the feature point test set to obtain an audio recognition model. 8. The good-sound stability early-warning device of the screw water-cooled unit according to claim 7, is characterized in that, described model training module is specifically used for: Inputting the feature point training set into the training model for iterative training according to a preset number of iterative training times to obtain a first detection model; Input the feature point test set into the first detection model for detection, obtain a first accuracy rate, and determine whether the first accuracy rate reaches a preset expected value, and if so, use the first detection model as the audio recognition model, and then perform optimization processing on the audio recognition model, if not, input the feature point test set into the training model for iterative training according to the preset iterative training times to obtain a second detection model; Inputting the feature point training set into the second detection model for detection to obtain a second accuracy rate; Determine whether the second accuracy rate reaches the preset expected value, and if so, use the second detection model as the audio recognition model, and then perform optimization processing on the audio recognition model, if not, use the audio recognition model again. The feature point data set is randomly divided into a feature point training set and a feature point test set. 9. A good-sound stability early-warning device for a screw water-cooled unit, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that when the processor executes all When the computer program is executed, the good-sound stability early-warning method of the screw water-cooling unit according to any one of claims 1 to 4 is realized. 10. A computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the computer program according to any one of claims 1 to 4 is implemented. Good sound stability early warning method for screw water-cooled units.

Images