Fault diagnosis method, device and equipment of water jet propulsion pump, medium and product

Abstract

The invention relates to the technical field of ships, and discloses a fault diagnosis method, device and equipment for a water jet propulsion pump, a medium and a product. The method comprises the steps of obtaining a current vibration signal of the water jet propulsion pump, and performing feature extraction on the current vibration signal through a pre-trained target supervision auto-encoder to obtain a current multi-dimensional feature vector; obtaining a preset question and answer template, and generating a target prompt word according to the preset question and answer template and the current multi-dimensional feature vector; and generating a fault diagnosis result corresponding to the water jet propulsion pump according to the target cue word through a pre-trained target large language model. The multi-dimensional depth features are extracted from the vibration signals by adopting the supervision auto-encoder, and the fault diagnosis is automatically performed according to the extracted features by utilizing the large language model, so that the automatic extraction of the fault features can be realized, the adaptability and generalization ability in a cross-working-condition scene can be improved, and the fault diagnosis efficiency and accuracy can be improved.

Description

Technical Field

[0001] This invention relates to the field of marine technology, and in particular to fault diagnosis methods, devices, equipment, media, and products for waterjet propulsion pumps. Background Technology

[0002] Waterjet propulsion systems, as advanced ship propulsion devices, are widely used in high-speed passenger ships and modern warships due to their high efficiency, low noise, and superior maneuverability. The waterjet pump, as the core power component of this system, has a complex internal structure and operates under harsh conditions of high speed and high pressure for extended periods. Key components such as the impeller and guide vanes are prone to various failures, including cavitation, wear, and cracking. If these failures are not detected and diagnosed in a timely manner, they can range from affecting ship performance to causing catastrophic accidents, seriously threatening navigational safety. Therefore, researching precise and intelligent waterjet pump fault diagnosis technology is of crucial theoretical and practical significance for ensuring the reliability and mission availability of ship equipment.

[0003] Currently, existing fault diagnosis methods for waterjet propulsion pumps are typically based on vibration signal analysis. This involves installing accelerometers on key pump components (such as guide vane casings) to collect vibration signals, and relying on signal processing techniques like Fourier transform and wavelet analysis, along with expert experience, to manually design and extract a series of time-domain and frequency-domain features to characterize fault information. However, these methods have significant limitations: First, the feature extraction process is highly dependent on professional knowledge and diagnostic experience, making it subjective and time-consuming; second, in practical applications, the variable navigation conditions of ships lead to complex and varied pump operating conditions (such as speed and load), often limiting the adaptability and generalization ability of manually designed and extracted features across different operating scenarios. Summary of the Invention

[0004] This invention provides a fault diagnosis method, device, equipment, medium, and product for a water jet propulsion pump, which can realize the automatic extraction of fault characteristics, improve the adaptability and generalization ability across operating conditions, and improve the efficiency and accuracy of fault diagnosis.

[0005] According to one aspect of the present invention, a fault diagnosis method for a waterjet propulsion pump is provided, comprising:

[0006] The current vibration signal of the water jet propulsion pump is acquired, and the current vibration signal is used to extract features from the pre-trained target-supervised autoencoder to obtain the current multidimensional feature vector.

[0007] Obtain a preset question-and-answer template, and generate target prompt words based on the preset question-and-answer template and the current multidimensional feature vector;

[0008] Using a pre-trained target large language model, fault diagnosis results are generated for the water jet propulsion pump based on the target prompt words.

[0009] According to another aspect of the present invention, a fault diagnosis device for a waterjet propulsion pump is provided, comprising:

[0010] The feature extraction module is used to acquire the current vibration signal of the water jet propulsion pump and extract features from the current vibration signal through a pre-trained target-supervised autoencoder to obtain the current multidimensional feature vector.

[0011] The prompt word generation module is used to obtain a preset question and answer template and generate target prompt words based on the preset question and answer template and the current multidimensional feature vector;

[0012] The fault diagnosis result generation module is used to generate the fault diagnosis result corresponding to the water jet propulsion pump based on the target prompt words using a pre-trained target large language model.

[0013] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising:

[0014] At least one processor; and

[0015] A memory communicatively connected to the at least one processor; wherein,

[0016] The memory stores a computer program that can be executed by the at least one processor, which enables the at least one processor to perform the fault diagnosis method for the water jet propulsion pump according to any embodiment of the present invention.

[0017] According to another aspect of the present invention, a computer-readable storage medium is provided, the computer-readable storage medium storing a computer program, the computer program being configured to cause a processor to execute and implement the fault diagnosis method for a water jet propulsion pump according to any embodiment of the present invention.

[0018] According to another aspect of the present invention, a computer program product is provided, comprising a computer program that, when executed by a processor, implements the fault diagnosis method for a water jet propulsion pump as described in any embodiment of the present invention.

[0019] The technical solution of this invention involves acquiring the current vibration signal of a waterjet propulsion pump, extracting features from the current vibration signal using a pre-trained target-supervised autoencoder to obtain a current multidimensional feature vector, acquiring a preset question-and-answer template, and generating target prompt words based on the preset question-and-answer template and the current multidimensional feature vector, and generating a fault diagnosis result corresponding to the waterjet propulsion pump using a pre-trained target large language model based on the target prompt words. By employing a supervised autoencoder to extract multidimensional deep features from the vibration signal and using a large language model to automatically perform fault diagnosis based on the extracted features, automatic extraction of fault features can be achieved, improving adaptability and generalization ability across different working conditions, and enhancing the efficiency and accuracy of fault diagnosis.

[0020] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is a flowchart of a fault diagnosis method for a water jet propulsion pump according to Embodiment 1 of the present invention;

[0023] Figure 2 This is a schematic diagram of a preset question and answer template provided in Embodiment 1 of the present invention;

[0024] Figure 3 This is a flowchart of a fault diagnosis method for a water jet propulsion pump according to Embodiment 2 of the present invention;

[0025] Figure 4 This is a flowchart of another method for diagnosing a water jet propulsion pump according to Embodiment 2 of the present invention;

[0026] Figure 5 This is a schematic diagram of the spray pump failure test device and measuring points provided in Embodiment 2 of the present invention;

[0027] Figure 6 This is a schematic diagram of diagnostic accuracy provided in Embodiment 2 of the present invention;

[0028] Figure 7 This is a schematic diagram of the structure of a fault diagnosis device for a water jet propulsion pump according to Embodiment 3 of the present invention;

[0029] Figure 8 This is a schematic diagram of the structure of an electronic device for implementing the fault diagnosis method of the water jet propulsion pump according to an embodiment of the present invention. Detailed Implementation

[0030] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0031] It should be noted that the terms "first," "second," "modification," etc., used in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such terms can be used interchangeably where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0032] Example 1

[0033] Figure 1 This is a flowchart illustrating a fault diagnosis method for a waterjet propulsion pump according to Embodiment 1 of the present invention. This embodiment is applicable to situations requiring automatic fault diagnosis of waterjet propulsion pumps. The method can be executed by a fault diagnosis device for the waterjet propulsion pump, which can be implemented in hardware and / or software. Typically, this fault diagnosis device can be configured in electronic equipment, such as computer equipment or servers. Figure 1 As shown, the method includes:

[0034] S110. Obtain the current vibration signal of the water jet propulsion pump, and extract features from the current vibration signal using a pre-trained target-supervised autoencoder to obtain the current multidimensional feature vector.

[0035] In this embodiment, a unidirectional acceleration sensor can be installed on the guide vane housing of the water jet propulsion pump to collect vibration signals during pump operation. For example, the sampling frequency can be 2500 Hz.

[0036] Furthermore, to address the problems of existing technologies where manual extraction of time-domain and frequency-domain features relies on expert knowledge, is time-consuming and labor-intensive, and has limited generalization ability, this embodiment proposes a novel method for automated feature learning based on a supervised autoencoder (SAE). The core advantage of this method lies in its ability to leverage label information to guide the feature learning process, automatically extracting optimal low-dimensional and robust feature representations for fault classification tasks from the original vibration signals. This provides input with higher information density and stronger discriminative power for subsequent fault diagnosis.

[0037] In this embodiment, the SAE model can be established and trained in advance, and the encoder part of the trained target SAE can be used to extract features from the current vibration signal, directly mapping the current vibration signal into a multi-dimensional depth feature vector to obtain the current multi-dimensional feature vector.

[0038] In an optional example, the multidimensional feature vector can be a 20-dimensional depth feature vector. Optionally, the multidimensional features can include time-domain features and frequency-domain features. Time-domain features can include mean, standard deviation, square root amplitude, absolute mean, peak value, variance, skewness, kurtosis, peak factor, and / or impulse factor. Frequency-domain features can include frequency-domain mean, frequency-domain variance, frequency-domain bias, spectral kurtosis, centroid frequency, frequency standard deviation, frequency mean square, center frequency, regularity, and / or eighth moment. The time-domain features describe the overall statistical properties of the signal, while the frequency-domain features reveal the periodic components of the signal.

[0039] S120. Obtain a preset question-and-answer template, and generate target prompt words based on the preset question-and-answer template and the current multidimensional feature vector.

[0040] The preset question-and-answer template can be a pre-set input format template for a Large Language Model (LLM), used to guide the LLM model in performing fault diagnosis tasks. In this embodiment, the current multidimensional feature vector can be combined with the preset question-and-answer template to generate target prompt words.

[0041] Optionally, generating target prompt words based on the preset question-and-answer template and the current multidimensional feature vector may include:

[0042] Obtain the filling position corresponding to the current multidimensional feature vector, and fill the current multidimensional feature vector into the preset question-and-answer template according to the filling position to generate the target prompt word.

[0043] In an optional example, the preset question-and-answer template can be as follows: Figure 2As shown in the diagram. The Instruction section describes the task information, Input indicates the input information, and Output indicates the output information. It's understandable that the target prompt may not include the Output section; instead, it can be used as the output format for the fault diagnosis result. Specifically, based on a preset question-and-answer template, the current multidimensional feature vector's filling position in the Input section is obtained. Following this filling position, the current multidimensional feature vector is filled into the Input section, and then combined with the Instruction section to generate the final target prompt.

[0044] S130. Using a pre-trained target large language model, generate the fault diagnosis result corresponding to the water jet propulsion pump based on the target prompt words.

[0045] The target large language model can be a large language model fine-tuned and trained using a sample set of a waterjet propulsion pump fault diagnosis scenario. In this embodiment, the target large language model can be invoked, and the target prompt words can be input into the target large language model to obtain its output fault diagnosis results for the waterjet propulsion pump.

[0046] Optionally, the fault diagnosis results may include normal status, bearing failure, and / or combined faults. A normal status indicates that all components of the spray pump are functioning normally; a bearing failure indicates a fault in one of the pump's support bearings; a combined fault indicates that, in addition to a bearing failure, there are also impeller imbalance and axial movement. It is understood that the fault diagnosis results can be adaptively adjusted according to the actual task scenario.

[0047] The technical solution of this invention involves acquiring the current vibration signal of a waterjet propulsion pump, extracting features from the current vibration signal using a pre-trained target-supervised autoencoder to obtain a current multidimensional feature vector, acquiring a preset question-and-answer template, and generating target prompt words based on the preset question-and-answer template and the current multidimensional feature vector, and generating a fault diagnosis result corresponding to the waterjet propulsion pump using a pre-trained target large language model based on the target prompt words. By employing a supervised autoencoder to extract multidimensional deep features from the vibration signal and using a large language model to automatically perform fault diagnosis based on the extracted features, automatic extraction of fault features can be achieved, improving adaptability and generalization ability across different working conditions, and enhancing the efficiency and accuracy of fault diagnosis.

[0048] Example 2

[0049] Figure 3 This is a flowchart illustrating a fault diagnosis method for a waterjet propulsion pump according to Embodiment 2 of the present invention. This embodiment is a further refinement of the above technical solution, and the technical solution in this embodiment can be combined with one or more of the above implementation methods. Figure 3As shown, the method includes:

[0050] S210, Obtain the current vibration signal of the water jet propulsion pump.

[0051] S220. Obtain the initial supervised autoencoder, and train the model on the initial supervised autoencoder based on the reconstruction loss function and the classification loss function to obtain the trained target supervised autoencoder.

[0052] SAE is an improved model based on the standard autoencoder, and its structure consists of three key parts: an encoder, a decoder, and a classifier. The encoder E is a multi-layer neural network responsible for compressing high-dimensional input data x into a low-dimensional latent variable z = E(x). In this embodiment, the dimension of z is 20. The decoder D has a structure roughly symmetrical to the encoder; its task is to receive the latent variable z and attempt to reconstruct the original input. The classifier C is a standalone multi-layer neural network that receives the latent variable z generated by the encoder as input and predicts the fault category corresponding to the sample. .

[0053] It should be noted that SAE uses a unique loss function, which consists of reconstruction loss and classification loss, as shown in the following formula:

[0054] .

[0055] in, Represents the reconstruction loss function. Represents the classification loss function. It is a hyperparameter used to balance the reconstruction loss and the classification loss. The mean squared error is used to measure the input x and the reconstructed output. The difference between the two factors means that the loss term forces the encoder to learn features z that retain enough information to recover the original signal, thus ensuring the integrity of the feature information. The cross-entropy loss function is used to measure the classifier's predicted category. The difference between the true label y and the actual label y is used to directly supervise the encoder's parameter updates through backpropagation, driving the generated features z to have high separability across different fault categories.

[0056] By optimizing this joint loss function, SAE can learn a feature space that retains the structural information of the original signal while specifically amplifying patterns related to the fault category. After training, the decoder and classifier parts are discarded, and only the trained encoder is retained as an efficient and automated feature extractor.

[0057] Specifically, an initial SAE can be built based on preset model parameters, and labeled raw data can be input into the initial SAE; then, the total loss can be calculated according to the joint loss function. and based on Perform backpropagation optimization and update the parameters of the three modules simultaneously; finally, repeat the above process until the total loss converges. At this point, the parameters of the three modules are fixed, and training is complete.

[0058] S230. The current vibration signal is extracted using a pre-trained target-supervised autoencoder to obtain the current multidimensional feature vector.

[0059] S240. Obtain a preset question-and-answer template, and generate target prompt words based on the preset question-and-answer template and the current multidimensional feature vector.

[0060] S250. Acquire historical vibration signals and extract features from the historical vibration signals using a pre-trained target-supervised autoencoder to obtain historical multidimensional feature vectors.

[0061] The core advantage of LLM lies in its powerful generalization and contextual learning capabilities. Through pre-training on massive amounts of data, LLM can acquire extensive knowledge and patterns, and can quickly adapt to new domains by fine-tuning for specific downstream tasks. In this embodiment, SAE can be used to automatically learn high-quality low-dimensional feature representations from vibration signals, extract task-oriented features, and then feed these "learned" features into the LLM for fine-tuning. This method combines the powerful generalization capabilities of LLM, providing it with highly refined and tailored feature representations, aiming to build a more efficient, accurate, and automated intelligent fault diagnosis framework.

[0062] Specifically, firstly, historical vibration signals are read from a specified database and input into a pre-trained target SAE. The encoder of the target SAE then extracts the corresponding historical multidimensional feature vectors.

[0063] S260. Obtain the label corresponding to the historical multidimensional feature vector, and generate a sample set based on the historical multidimensional feature vector and the corresponding label.

[0064] Then, the health status corresponding to the manually labeled historical multidimensional feature vectors is obtained as the label, and the mapping relationship between the historical multidimensional feature vectors and the label is established as a sample data. Then, a set of all sample data is generated as the sample set.

[0065] S270. Obtain a basic large language model, and use a low-rank adaptation method to fine-tune the basic large language model based on the sample set to obtain the trained target large language model.

[0066] Next, each sample data in the sample set can be transformed into, for example... Figure 2 The LLM shown is an easily processed text sequence; optionally, the text sequence may also include manually extracted time-domain and frequency-domain features. Finally, the open-source base LLM can be fine-tuned based on this text sequence to obtain the trained target LLM. This embodiment does not specifically limit the base LLM.

[0067] To efficiently adapt the LLM to the fault diagnosis task of this invention, this embodiment employs a supervised fine-tuning (SFT) strategy. Considering the large number of parameters in the base LLM, this embodiment uses low-rank adaptation (LoRA) for fine-tuning training. The core idea of ​​LoRA is to freeze the original pre-trained weights of the base LLM. And next to it is a parallel matrix consisting of two low-rank matrices. and The "bypass" is formed. During fine-tuning training, The model is frozen and does not accept gradient updates; it only trains the parameters contained in A and B, while the model's weights are updated... Then it is represented by their product AB, as shown in the following formula:

[0068] .

[0069] Since the rank r = min(d, k), the number of parameters that need to be trained is greatly reduced, thus enabling fast and efficient fine-tuning of large models on consumer-grade hardware.

[0070] S280. Using a pre-trained target large language model, generate the fault diagnosis result corresponding to the water jet propulsion pump based on the target prompt words.

[0071] The technical solution of this invention obtains an initial supervised autoencoder and trains it based on a reconstruction loss function and a classification loss function to obtain a trained target supervised autoencoder, ensuring the integrity of feature information and separability between different fault categories. Secondly, by acquiring historical vibration signals and extracting features from them using a pre-trained target supervised autoencoder, historical multidimensional feature vectors are obtained. The corresponding labels for these historical multidimensional feature vectors are then acquired, and a sample set is generated based on the historical multidimensional feature vectors and their corresponding labels. A basic large language model is obtained, and a low-rank adaptation method is used to fine-tune the basic large language model based on the sample set, resulting in a trained target large language model. By combining supervised fine-tuning strategies and the low-rank adaptation method, the efficiency of fine-tuning training of the large language model can be improved.

[0072] In one specific embodiment of this example, the fault diagnosis method for the waterjet propulsion pump can be described as follows: Figure 4 As shown, in the SAE model training step, an SAE model is designed and trained. This model takes the original vibration signal as input and simultaneously optimizes both reconstruction and classification objectives. Next, in the automated feature extraction step, the trained SAE encoder is used to directly map the new vibration signal into a 20-dimensional deep feature vector. Finally, in the feature textualization and LLM fine-tuning step, the feature vector extracted by the SAE is textualized and used to fine-tune the LLM.

[0073] This invention proposes a novel framework for fault diagnosis of waterjet propulsion pumps based on a combination of supervised autoencoders and large language models. This framework aims to integrate task-oriented automated feature learning with the powerful generalization capabilities of LLM. Its main contributions are as follows: (1) A task-oriented automated feature learning paradigm is proposed, utilizing SAE to directly learn deep features for fault classification tasks from raw vibration signals, thus automating feature extraction and ensuring optimal feature discriminability for fault types. (2) A deeply integrated intelligent diagnostic framework is constructed, combining high-quality features learned by SAE with the powerful generalization and reasoning capabilities of LLM. Through efficient parameter fine-tuning techniques, an end-to-end, high-precision diagnostic model is built. (3) The superiority of the proposed framework is fully verified. Comprehensive experiments are conducted in two key scenarios: direct diagnosis and small-sample transfer diagnosis under limited data conditions. The results show that compared to the benchmark method based on manually extracted time-frequency domain features, this framework significantly improves diagnostic accuracy, generalization capability, and data efficiency.

[0074] To verify the technical effectiveness of the present invention, the following two experiments were designed. The experimental data were collected from a waterjet propulsion pump test bench, and the pump failure test device and measuring points are as follows: Figure 5As shown in Table 1, the experiment included operational data of the jet pump under three different health conditions: normal, bearing failure, and combined failure. To simulate the ship's operation at different speeds, data was collected for each of the aforementioned health conditions at three different engine speeds: 100 r / min (Revolutions Per Minute), 150 r / min, and 200 r / min. The detailed structure of the dataset is shown in Table 1.

[0075] Table 1. Description of the experimental dataset

[0076]

[0077] Secondly, all models were implemented using the PyTorch framework, version 2.6.0+cu124. Experiments were conducted on a server equipped with an NVIDIA A100 GPU. The key hyperparameter settings for the models are shown in Table 2.

[0078] Table 2 Key Hyperparameter Settings of the Model

[0079]

[0080] To evaluate the fault diagnosis accuracy under sufficient data conditions, this invention mixes all data from three operating conditions in the dataset and divides it into a training set (80%) and a test set (20%). Simultaneously, this invention selects three different basic large language models with similar parameter sizes: Llama-3.2-3B-Instruct, Qwen-2.5-3B-Instruct, and Gemma-3-4B-Instruct. Features extracted using SAE are fed into different basic LLMs for fine-tuning. The test set is then fed into the fine-tuned LLMs to evaluate their diagnostic accuracy. The diagnostic accuracy is shown below. Figure 6 As shown.

[0081] Furthermore, considering the high cost and expense of acquiring fault data in practical engineering applications, the amount of fault data is limited. To evaluate the model's generalization and transfer learning capabilities under scenarios with sparse data and changing operating conditions, this invention simulates a typical industrial scenario: the model is primarily trained on operating condition 0 (e.g., 100 r / min) and operating condition 1 (e.g., 150 r / min), but only a very small amount of data is available for the target operating condition 2 (e.g., 200 r / min). This invention sets up two small-sample fault data scenarios with limited data: a) Based on training with all data from operating conditions 0 and 1, a small amount of data (10%) from operating condition 2 is added to fine-tune the basic LLM model, and finally, it is tested on the remaining 90% of the data from operating condition 2; b) Only 10% of the data from operating condition 2 is used as the training set to fine-tune the basic LLM, and then it is tested on the remaining 90% of the data from operating condition 2. The transfer diagnosis accuracy of the technical solution of this invention in the two small-sample scenarios is shown in Table 3. Experimental results show that the technical solution of this invention exhibits significant advantages in both small-sample scenarios. In experimental scenario a, the features learned by the SAE method using source domain data exhibit a certain degree of operational condition invariance, allowing it to quickly adapt to new operating conditions with only a small number of target domain samples. Even in scenario b, with limited data and a small sample size, the SAE method achieves high accuracy with only a very small number of samples. This clearly demonstrates that the feature representations automatically learned by SAE are not only highly discriminative but also more robust and possess a certain degree of generalization. It can capture the essential physical patterns of faults, rather than superficial phenomena under specific operating conditions, thus exhibiting a certain degree of adaptability when facing challenges such as changing operating conditions and data sparsity.

[0082] Table 3. Accuracy of small sample migration diagnosis

[0083]

[0084] This invention successfully constructs and validates a new, automated, and highly accurate diagnostic paradigm. Experimental results show that, whether in data-rich direct diagnostic scenarios or in data-sparse cross-condition migration scenarios, the proposed SAE-LLM framework significantly outperforms benchmark methods based on manual features in terms of diagnostic accuracy. Furthermore, it demonstrates the crucial role of task-oriented feature learning in LLM diagnostic performance. Through the supervised learning mechanism of SAE, the extracted features not only achieve automation but, more importantly, possess inherent fault discriminative power far exceeding that of manual statistical features, thus fully leveraging the classification potential of LLM, especially in distinguishing easily confused complex faults. Finally, it reveals the significant potential of the proposed framework in addressing real-world industrial challenges. Its outstanding performance in small-sample and transfer learning experiments demonstrates the method's strong generalization ability and data efficiency, providing an effective technical approach to solving the common problems of sample scarcity and variable operating conditions in the field of industrial intelligent maintenance.

[0085] Example 3

[0086] Figure 7 This is a schematic diagram of a fault diagnosis device for a waterjet propulsion pump provided in Embodiment 3 of the present invention. Figure 7 As shown, the device includes: a feature extraction module 310, a prompt word generation module 320, and a fault diagnosis result generation module 330; wherein,

[0087] The feature extraction module 310 is used to acquire the current vibration signal of the water jet propulsion pump and extract features from the current vibration signal through a pre-trained target supervised autoencoder to obtain the current multidimensional feature vector.

[0088] The prompt word generation module 320 is used to obtain a preset question and answer template and generate target prompt words based on the preset question and answer template and the current multidimensional feature vector;

[0089] The fault diagnosis result generation module 330 is used to generate the fault diagnosis result corresponding to the water jet propulsion pump based on the target prompt words using a pre-trained target large language model.

[0090] The technical solution of this invention involves acquiring the current vibration signal of a waterjet propulsion pump, extracting features from the current vibration signal using a pre-trained target-supervised autoencoder to obtain a current multidimensional feature vector, acquiring a preset question-and-answer template, and generating target prompt words based on the preset question-and-answer template and the current multidimensional feature vector, and generating a fault diagnosis result corresponding to the waterjet propulsion pump using a pre-trained target large language model based on the target prompt words. By employing a supervised autoencoder to extract multidimensional deep features from the vibration signal and using a large language model to automatically perform fault diagnosis based on the extracted features, automatic extraction of fault features can be achieved, improving adaptability and generalization ability across different working conditions, and enhancing the efficiency and accuracy of fault diagnosis.

[0091] Optionally, the fault diagnosis device for the waterjet propulsion pump also includes:

[0092] The supervised autoencoder training module is used to obtain an initial supervised autoencoder and train the model on the initial supervised autoencoder based on the reconstruction loss function and the classification loss function to obtain the trained target supervised autoencoder.

[0093] Optionally, the fault diagnosis device for the waterjet propulsion pump also includes:

[0094] The historical multidimensional feature vector acquisition module is used to acquire historical vibration signals and extract features from the historical vibration signals through a pre-trained target-supervised autoencoder to obtain historical multidimensional feature vectors.

[0095] The sample set generation module is used to obtain the labels corresponding to the historical multidimensional feature vectors and generate a sample set based on the historical multidimensional feature vectors and the corresponding labels.

[0096] The model fine-tuning training module is used to obtain a basic large language model and use a low-rank adaptation method to fine-tune the basic large language model based on the sample set to obtain the target large language model after training.

[0097] Optionally, the prompt word generation module 320 is specifically used to obtain the filling position corresponding to the current multidimensional feature vector, and fill the current multidimensional feature vector into the preset question and answer template according to the filling position to generate the target prompt word.

[0098] Optionally, the multidimensional features include time-domain features and frequency-domain features. The time-domain features include mean, standard deviation, square root amplitude, absolute mean, peak value, variance, skewness, kurtosis, peak factor and / or impulse factor. The frequency-domain features include frequency-domain mean, frequency-domain variance, frequency-domain bias, spectral kurtosis, centroid frequency, frequency standard deviation, frequency mean square value, center frequency, regularity and / or eighth moment.

[0099] Optionally, the fault diagnosis results include normal condition, bearing failure, and / or combined fault.

[0100] The fault diagnosis device for the water jet propulsion pump provided in the embodiments of the present invention can execute the fault diagnosis method for the water jet propulsion pump provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the method.

[0101] The collection, storage, use, processing, transmission, provision, and disclosure of user personal information involved in the technical solution disclosed herein comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0102] Example 4

[0103] Figure 8 A schematic diagram of an electronic device 40 that can be used to implement embodiments of the present invention is shown. The electronic device 40 is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device 40 can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices (such as helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0104] like Figure 8As shown, the electronic device 40 includes at least one processor 41 and a memory, such as a read-only memory (ROM) 42 or a random access memory (RAM) 43, communicatively connected to the at least one processor 41. The memory stores computer programs executable by the at least one processor. The processor 41 can perform various appropriate actions and processes based on the computer program stored in the read-only memory 42 or loaded from the storage unit 48 into the random access memory 43. The RAM 43 can also store various programs and data required for the operation of the electronic device 40. The processor 41, ROM 42, and RAM 43 are interconnected via a bus 44. An input / output (I / O) interface 45 is also connected to the bus 44.

[0105] Multiple components in electronic device 40 are connected to I / O interface 45, including: input unit 46, such as keyboard, mouse, etc.; output unit 47, such as various types of monitors, speakers, etc.; storage unit 48, such as disk, optical disk, etc.; and communication unit 49, such as network card, modem, wireless transceiver, etc. Communication unit 49 allows electronic device 40 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0106] Processor 41 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 41 include, but are not limited to, central processing units, graphics processing units, various special-purpose artificial intelligence computing chips, various processors running machine learning model algorithms, digital signal processors, and any suitable processor, controller, microcontroller, etc. Processor 41 performs the various methods and processes described above, such as the fault diagnosis method for a water jet propulsion pump.

[0107] In some embodiments, the fault diagnosis method for the waterjet propulsion pump can be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 48. In some embodiments, part or all of the computer program can be loaded and / or installed on electronic device 40 via ROM 42 and / or communication unit 49. When the computer program is loaded into RAM 43 and executed by processor 41, one or more steps of the fault diagnosis method for the waterjet propulsion pump described above can be performed. Alternatively, in other embodiments, processor 41 can be configured to perform the fault diagnosis method for the waterjet propulsion pump by any other suitable means (e.g., by means of firmware).

[0108] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays, application-specific integrated circuits (ASICs), application-specific standard products (ASICs), system-on-a-chip (SoCs), complex programmable logic devices, computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0109] Computer programs used to implement the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0110] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory, read-only memory, erasable programmable read-only memory, optical fibers, portable compact disk read-only memory, optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0111] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device 40, which includes: a display device (e.g., a cathode ray tube or liquid crystal display) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device 40. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0112] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or computing systems that include middleware components (e.g., application servers), or computing systems that include frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.

[0113] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact via a communication network. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server.

[0114] This embodiment may also include a computer program product, which includes a computer program that, when executed by a processor, implements the fault diagnosis method for the water jet propulsion pump provided in any embodiment of the present invention.

[0115] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.

[0116] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A fault diagnosis method for a waterjet propulsion pump, characterized in that, include: The current vibration signal of the water jet propulsion pump is acquired, and the current vibration signal is used to extract features from the pre-trained target-supervised autoencoder to obtain the current multidimensional feature vector. Obtain a preset question-and-answer template, and generate target prompt words based on the preset question-and-answer template and the current multidimensional feature vector; Using a pre-trained target large language model, fault diagnosis results are generated for the water jet propulsion pump based on the target prompt words.

2. The method according to claim 1, characterized in that, Before extracting features from the current vibration signal using a pre-trained target-supervised autoencoder to obtain the current multidimensional feature vector, the process also includes: An initial supervised autoencoder is obtained, and a model is trained on the initial supervised autoencoder based on the reconstruction loss function and the classification loss function to obtain the trained target supervised autoencoder.

3. The method according to claim 1, characterized in that, Before generating the fault diagnosis result corresponding to the waterjet propulsion pump based on the target prompt words using a pre-trained target large language model, the process also includes: Historical vibration signals are acquired, and features are extracted from the historical vibration signals using a pre-trained target-supervised autoencoder to obtain historical multidimensional feature vectors. Obtain the labels corresponding to the historical multidimensional feature vectors, and generate a sample set based on the historical multidimensional feature vectors and their corresponding labels; A basic large language model is obtained, and a low-rank adaptation method is used to fine-tune the basic large language model based on the sample set to obtain the target large language model after training.

4. The method according to claim 1, characterized in that, Based on the preset question-and-answer template and the current multidimensional feature vector, target prompt words are generated, including: Obtain the filling position corresponding to the current multidimensional feature vector, and fill the current multidimensional feature vector into the preset question-and-answer template according to the filling position to generate the target prompt word.

5. The method according to any one of claims 1-4, characterized in that, Multidimensional features include time-domain features and frequency-domain features. Time-domain features include mean, standard deviation, square root amplitude, absolute mean, peak value, variance, skewness, kurtosis, peak factor and / or impulse factor. Frequency-domain features include frequency-domain mean, frequency-domain variance, frequency-domain bias, spectral kurtosis, centroid frequency, frequency standard deviation, frequency mean square value, center frequency, regularity and / or eighth moment.

6. The method according to claim 1, characterized in that, Fault diagnosis results include normal conditions, bearing failures, and / or combined faults.

7. A fault diagnosis device for a waterjet propulsion pump, characterized in that, include: The feature extraction module is used to acquire the current vibration signal of the water jet propulsion pump and extract features from the current vibration signal through a pre-trained target-supervised autoencoder to obtain the current multidimensional feature vector. The prompt word generation module is used to obtain a preset question and answer template and generate target prompt words based on the preset question and answer template and the current multidimensional feature vector; The fault diagnosis result generation module is used to generate the fault diagnosis result corresponding to the water jet propulsion pump based on the target prompt words using a pre-trained target large language model.

8. An electronic device, characterized in that, The electronic device includes: At least one processor, and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the fault diagnosis method for the water jet propulsion pump according to any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the fault diagnosis method for the water jet propulsion pump according to any one of claims 1-6.

10. A computer program product, characterized in that, The system includes a computer program that, when executed by a processor, implements the fault diagnosis method for the water jet propulsion pump according to any one of claims 1-6.

Images