A compressor service life prediction method and an electronic starting controller

Abstract

The application discloses a compressor service life prediction method and an electronic starting controller. The method comprises the following steps: obtaining working condition parameters of a compressor, performing time sequence unification and preprocessing on the working condition parameters, and generating quantitative indexes, wherein the quantitative indexes comprise long-time sequence sensitive indexes and short-time sequence fluctuation indexes; extracting physical characteristics and model automatic characteristics of the long-time sequence sensitive indexes and the short-time sequence fluctuation indexes, and obtaining a time sequence sample set; training and testing a pre-constructed fusion model by using the time sequence sample set, obtaining an evaluation result of the fusion model, and deploying based on the evaluation result. The long-period aging characteristics and high-frequency fluctuation characteristics in the compressor are processed in parallel by using the fusion model. Compared with the existing service life prediction mode, the overall precision is higher, the comprehensive consideration factors are more comprehensive, and the double paths can also adjust the hierarchical parameters to adjust the overall precision to retain stable characteristics and adapt to the dynamic working conditions of the compressor.

Description

Technical Field

[0001] This invention relates to the field of compressor technology, specifically to a method for predicting the service life of a compressor and an electronic starter. Background Technology

[0002] The core function of a compressor is to compress gas, increase gas pressure or density, and realize energy transfer. It is widely used in compressed air supply for industrial pneumatic systems, natural gas transportation, gas compression in chemical reactions, and in the fields of air conditioning and heat pumps for new energy vehicles. It is also the refrigerant circulation power source for refrigeration equipment (air conditioners, refrigerators).

[0003] Currently, when predicting the service life of compressors, the average life and remaining life of the compressor are usually estimated based on historical operating data of similar products, such as cumulative operating time, number of start-stop cycles, or fault records. Alternatively, extreme working conditions are simulated in the laboratory to accelerate compressor aging, and the life under actual use scenarios is extrapolated through test data.

[0004] However, the operating environment of compressors varies in different scenarios, and there are differences between simulated operating conditions and actual operating scenarios, such as the intermittency and load fluctuations of actual operating conditions. This leads to a large error in the predicted results, making it impossible to predict based on the current operating conditions of the compressor, and thus failing to fully reflect the true lifespan. Summary of the Invention

[0005] Based on this, the purpose of this invention is to provide a compressor lifespan prediction method and an electronic starter, which aims to solve the problem that the compressor's operating environment varies in different scenarios, and there are differences between simulated operating conditions and actual operating scenarios, making it impossible to predict the compressor's lifespan based on the current operating conditions.

[0006] To achieve the above objectives, this invention proposes a method for predicting the service life of a compressor, characterized in that the method includes:

[0007] The compressor's operating parameters are obtained, and the operating parameters are subjected to time series unification and preprocessing to generate quantitative indicators, wherein the quantitative indicators include long-time series sensitive indicators and short-time series fluctuation indicators.

[0008] Extract the physical features and model-automated features of the long-term time-series sensitive indicators and the short-term time-series fluctuation indicators to obtain a time-series sample set;

[0009] The pre-built fusion model is trained and tested using a time-series sample set to obtain the evaluation results of the fusion model, and then deployed based on the evaluation results.

[0010] According to one aspect of the above technical solution, in the step of obtaining the quantitative indicators of the compressor and performing time-series unification and preprocessing on the quantitative indicators:

[0011] By collecting compressor operating parameters at different frequencies using different types of sensors with a unified time sequence, the operating parameters are cleaned, time-series aligned and standardized to generate quantitative indicators.

[0012] Based on the time-series nature of the quantitative indicators, the quantitative indicators include long-term time-series sensitive indicators and short-term time-series fluctuation indicators. The long-term time-series sensitive indicators include at least the vibration signals and lubrication parameters of the compressor components, and the short-term time-series fluctuation indicators include at least the operating parameters and environmental parameters of the compressor.

[0013] According to one aspect of the above technical solution, the step of extracting the physical characteristics and model-automated characteristics of the long-term sensitive index and the short-term fluctuation index includes:

[0014] Based on the vibration signal, a time-series vibration curve is plotted. The time-series vibration curve is sampled using a first window. Linear regression fitting is performed on the sampling interval to obtain the time-series trend characteristics of the time-series vibration curve. At the same time, the vibration kurtosis and waveform factor of the curve segment in each sliding sampling unit within the first window are calculated. The circumferential growth rate of the lubrication parameters and the oil film thickness are also calculated.

[0015] The time-series curves of operating parameters and environmental parameters are sampled using the second and third windows respectively. Based on the segment data in the sliding sampling units of the second and third windows, the coefficient of variation of operating parameters and the power load rate correlation and corrected power of environmental parameters are calculated.

[0016] In this system, the length of the sliding sampling unit in any window is equal, and the physical characteristics of the long-term time-series sensitive indicators and short-term time-series fluctuation indicators obtained through the sliding sampling unit are summarized.

[0017] According to one aspect of the above technical solution, after obtaining physical features, automatic features of the model are extracted, and a sliding window with a preset length and a preset step size is used. Based on the time axis, the model is cut according to the preset window length and preset step size continuous quantization index to obtain several time segment samples. The dimensionality attribute of any time segment sample matches the window length and step size of the sliding window.

[0018] Based on the fault thresholds of each component of the compressor and the fault downtime timestamps, the remaining lifetime labels of the time-series segment samples are calculated and obtained:

[0019]

[0020] in, For remaining lifetime label, This is the timestamp for the downtime due to the fault. This is the end timestamp of the time sequence segment sample;

[0021] Based on the extracted physical features and the model's automatic features, the time-series sample set obtained from training the fusion model is output.

[0022] According to one aspect of the above technical solution, the constructed fusion model includes an input layer, a feature extraction layer, and an output layer, wherein:

[0023] The feature extraction layer includes parallel connected LSTM and GRU paths, which are connected to the output layer through a feature fusion layer. The output layer includes a fully connected layer and a final output layer.

[0024] The LSTM path includes a first long-time extraction layer and a second long-time extraction layer arranged in series. The first long-time extraction layer has 256 neurons, and the second long-time extraction layer has 128 neurons.

[0025] The GRU path includes a first short-time extraction layer and a second short-time extraction layer arranged in series. The first short-time extraction layer has 128 neurons, and the second short-time extraction layer has 64 neurons.

[0026] The LSTM path and GRU path are concatenated using a feature fusion layer to output a fused feature vector.

[0027] According to one aspect of the above technical solution, a fully connected layer is used to perform a nonlinear mapping on the fused feature vector to obtain intermediate features, and the intermediate features are then linearly mapped through the final output layer to obtain the predicted lifetime value.

[0028]

[0029] in, As an intermediate feature, It is a fusion of feature vectors. It is the weight matrix of the fully connected layer. This is the bias vector of the fully connected layer. For activation functions;

[0030]

[0031] in, This is a predicted service life value. This is the weight matrix of the final output layer. The final output layer's bias vector.

[0032] According to one aspect of the above technical solution, the step of training and testing the pre-built fusion model using a time-series sample set, obtaining the evaluation result of the fusion model, and deploying it based on the evaluation result includes:

[0033] The time series sample set is divided into a training set, a validation set, and a test set according to a preset ratio, and the fusion model is trained. The distribution ratio of the lifetime prediction values ​​of the training set, the validation set, and the test set is the same.

[0034] Obtain the overall accuracy of the test set. If the overall accuracy does not reach the expected threshold, adjust the hierarchical parameters in the LSTM and GRU paths to optimize the fusion model.

[0035] The optimized fusion model is deployed on an edge server, and the parameters of the feature fusion layer are adjusted based on real-time data.

[0036] This invention also proposes a compressor lifespan prediction system, which is used to implement the above-mentioned compressor lifespan prediction method. The system includes:

[0037] The index quantification module is used to acquire the compressor's operating parameters, perform time series unification and preprocessing on the operating parameters, and generate quantitative indicators, wherein the quantitative indicators include long-time series sensitive indicators and short-time series fluctuation indicators.

[0038] The feature extraction module is used to extract the physical features and model-automated features of the long-term sensitive index and the short-term fluctuation index to obtain a time-series sample set;

[0039] The evaluation and deployment module is used to train and test the pre-built fusion model using a time-series sample set, obtain the evaluation results of the fusion model, and deploy it based on the evaluation results.

[0040] The present invention also proposes a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the compressor lifespan prediction method described above.

[0041] The present invention also proposes an electronic starter, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the compressor life prediction method as described above.

[0042] In summary, the compressor lifespan prediction method provided by this invention filters out parameters such as compressor vibration type and operating type, preprocesses them to generate quantitative indicators, extracts physical features and model-automated features from the quantitative indicators, generates a time-series sample set, and uses the LSTM path and GRU path set in parallel in the fusion model to capture the long-cycle aging features of the physical type in the time-series sample set, as well as efficiently process high-frequency fluctuation features. This allows the fusion model to simultaneously learn the effects of slow-changing aging trends and fast-changing operating conditions, improving robustness. This invention, by using a fusion model to process long-cycle aging features and high-frequency fluctuation features in the compressor in parallel, achieves higher overall accuracy and more comprehensive consideration of factors compared to existing lifespan prediction methods. Furthermore, the dual-path approach allows for adjustment of hierarchical parameters to adjust the overall accuracy, preserving stable features and adapting to the dynamic operating conditions of the compressor.

[0043] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0044] Figure 1 This is a flowchart of the compressor lifespan prediction method in Embodiment 1 of the present invention;

[0045] Figure 2 This is a schematic diagram of the compressor lifespan prediction system in Embodiment 2 of the present invention;

[0046] Figure 3 This is a structural block diagram of the electronic starter in Embodiment 4 of the present invention. Detailed Implementation

[0047] To make the objectives, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Several embodiments of the present invention are shown in the drawings. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided so that the disclosure of the present invention will be more thorough and complete.

[0048] It should be noted that when an element is referred to as being "fixed to" another element, it can be directly on the other element or there may be an intervening element. When an element is considered to be "connected" to another element, it can be directly connected to the other element or there may be an intervening element. The terms "vertical," "horizontal," "left," "right," "upper," "lower," and similar expressions used herein are for illustrative purposes only and are not intended to indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as limiting the invention.

[0049] In this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," "linking," and "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal communication between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances. The term "and / or" as used herein includes any and all combinations of one or more of the related listed items.

[0050] Example 1

[0051] Please see Figure 1 The diagram shows a flowchart of a compressor lifespan prediction method according to Embodiment 1 of the present invention. The compressor lifespan prediction method includes the following steps S01-S03, wherein:

[0052] S01. Obtain the operating parameters of the compressor, perform time series unification and preprocessing on the operating parameters, and generate quantitative indicators, wherein the quantitative indicators include long-time series sensitive indicators and short-time series fluctuation indicators.

[0053] Different types of sensors are used to acquire vibration, lubrication, operational, and environmental parameters during compressor operation. Vibration and lubrication parameters are long-term sensitive indicators, while operational and environmental parameters are short-term fluctuation indicators. Vibration parameters include physical characteristics such as bearing housing acceleration and cylinder wall vibration velocity; lubrication parameters include oil contamination level and oil film thickness; operational parameters include exhaust pressure, intake temperature, and instantaneous power; and environmental parameters include ambient temperature and load rate. It is important to note that the timing of all sensors must be standardized during parameter acquisition, and different parameters should be sampled at different frequencies. Vibration parameters can be sampled at 10kHz or 5kHz. For example, bearing housing acceleration can be sampled at 10kHz, taking 1024 time-domain statistical values ​​every 10 minutes, such as peak value, kurtosis, and root mean square (RMS). Cylinder wall vibration velocity can be sampled at 5kHz. After sampling in this manner, the data can be stored in the format of component name-timestamp-quantized value.

[0054] For data cleaning after storage, box plots can be used to identify abrupt changes in pressure or temperature values, and the moving average of the preceding several acquisition cycles can be used to replace them. For missing values ​​of vibration parameters, spline differences can be used to replace them, and for missing values ​​of power, the average of the acquisition cycles symmetrical before and after the missing value can be used to replace them.

[0055] In addition, the time granularity of all cleaned data is unified to ensure that the standardized data within a unit of time includes all long-term time-sensitive indicators and short-term time-fluctuation indicators. For long-term time-series indicators, the average value is taken according to the unit of time, and for short-term time-fluctuation indicators, the sampling is directly performed in units of time, which can be one minute or two minutes.

[0056] S02. Extract the physical features and automatic model features of the long-term time-series sensitive indicators and the short-term time-series fluctuation indicators to obtain a time-series sample set.

[0057] The physical characteristics need to be calculated by classifying them into long-term sensitive indicators and short-term fluctuation indicators. Each physical characteristic corresponds to the specific operating state or deterioration pattern of the compressor. This step is based on the standardized data per unit time in step S01 for extraction.

[0058] Based on the parameters of the vibration type, a time-series vibration curve is plotted. The time-series vibration curve is then sampled using a first window, and linear regression fitting is performed on the sampling interval to obtain the time-series trend characteristics of the vibration curve. At the same time, the vibration kurtosis and waveform factor of the curve segment in each sliding sampling unit within the first window are calculated. The vibration kurtosis reflects the concentration of signal peaks, and the kurtosis will increase significantly in the early stage of the fault. The waveform factor reflects the impact characteristics of the signal. By sampling the time-series vibration curves of components such as the compressor's bearing housing or cylinder wall, the vibration intensity and wear condition can be quantified.

[0059] For lubrication parameters, the time-series data can also be slip-sampled, and the week-on-week growth rate and oil film thickness of the lubrication parameters can be calculated. The week-on-week growth rate represents the degree of oil contamination. At the same time, the oil film thickness is used to calculate the sliding standard deviation, which represents the stability of the oil film. By observing the growth trend of the sliding standard deviation, the stability of the oil film can be determined, and thus it can be used to predict whether there is a possibility of lubrication failure.

[0060] The time-series curves of operating parameters and environmental parameters are sampled using the second and third windows respectively. Based on the segment data in the sliding sampling units of the second and third windows, the coefficient of variation of operating parameters and the power load rate correlation and corrected power of environmental parameters are calculated.

[0061] In this step, the length of the sliding sampling unit in the first window, the second window, and the third window is equal. The length of the sliding sampling unit in this step is the unit time in step S01. The first window, the second window, and the third window can be set with different time lengths. The physical characteristics of the long-term time-series sensitive indicators and the short-term time-series fluctuation indicators obtained through the sliding sampling unit are summarized.

[0062] After obtaining physical features, automatic features of the model are extracted. Using a sliding window with a preset length and a preset step size, the model is divided according to the preset window length and preset step size continuous quantization index based on the time axis to obtain several time segment samples. The dimensionality attribute of any time segment sample matches the window length and step size of the sliding window.

[0063] Based on the fault thresholds of each component of the compressor and the fault downtime timestamps, the remaining lifetime labels of the time-series segment samples are calculated and obtained:

[0064]

[0065] in, For remaining lifetime label, This is the timestamp for the downtime due to the fault. This is the end timestamp of the time sequence segment sample.

[0066] It should be noted that the remaining life tag in this embodiment focuses on compressor failure shutdown and sets the failure judgment criteria. Failure shutdown is defined as the parameters in the long-term sensitive index and the short-term fluctuation index exceeding the set threshold.

[0067] S03. Use the time-series sample set to train and test the pre-built fusion model, obtain the evaluation results of the fusion model, and deploy it based on the evaluation results.

[0068] The fusion model in this embodiment is an LSTM-GRU fusion model, which includes an input layer, a feature extraction layer, and an output layer. The feature extraction layer includes parallel-connected LSTM and GRU paths. The LSTM path focuses on extracting long-term aging features, while the GRU path focuses on extracting short-term fluctuation features. The LSTM and GRU paths are connected to the output layer through the feature fusion layer. The output layer includes a fully connected layer and a final output layer.

[0069] The LSTM path includes a first long-term extraction layer and a second long-term extraction layer arranged in series. The first long-term extraction layer has 256 neurons, which is sufficient to capture long-term correlations of slowly varying indicators such as vibration and lubricating oil (e.g., the cumulative trend of bearing wear). This allows the output of the first long-term extraction layer to be used as the input of the next LSTM layer, further deepening the long-term feature mining. The second long-term extraction layer has 128 neurons. Reducing the number of neurons from 256 to 128 reduces the number of model parameters and computation, avoids overfitting, and compresses the features within a sliding window into a global vector (128 dimensions), facilitating subsequent feature fusion with the GRU path. Furthermore, a Dropout layer can be added after the second long-term extraction layer, with a dropout rate of 0.3 (randomly discarding 30% of the neurons) to suppress overfitting in the long-term path and prevent the model from excessively memorizing long-period noise in the training set, such as abnormal vibrations that are not fault signals over a certain period.

[0070] The GRU path comprises a first short-time extraction layer and a second short-time extraction layer arranged in series. The first short-time extraction layer has 128 neurons. Compared to the LSTM path, the GRU path is simpler and more computationally efficient. 128 neurons are sufficient to capture short-term correlations of rapidly changing indicators such as pressure and power (e.g., power fluctuations caused by load changes). The output of the first short-time extraction layer can be adapted to the input requirements of the second short-time extraction layer. The second short-time extraction layer has 64 neurons, matching the lightweight and efficient localization of the GRU path and avoiding excessive computational resources. The second short-time extraction layer compresses the features within a sliding window into a global vector (64 dimensions). Furthermore, a Dropout layer can be connected after the second short-time extraction layer, with a dropout rate of 0.2, which randomly discards 20% of the neurons. Compared to the LSTM path, the noise in short-term fluctuation features is relatively less, and the lower dropout rate can retain more effective features (e.g., load fluctuation signals under normal operating conditions) while avoiding overfitting.

[0071] In the feature fusion layer, a concatenation function can be used to concatenate the LSTM path and the GRU path, outputting a 192-dimensional fused feature vector. This fused feature vector contains two core types of information: long-term aging trend features and short-term operating condition fluctuation features. This concatenation-style fusion does not lose the feature weights of either path and retains the advantages of the two paths better than additive fusion (addition would weaken the specificity of each feature), allowing subsequent fully connected layers to learn the association between the two types of features and RUL simultaneously.

[0072] A fully connected layer is used to perform a non-linear mapping on the fused feature vector to obtain intermediate features, and then the intermediate features are linearly mapped through the final output layer to obtain the lifespan prediction value.

[0073]

[0074] in, As an intermediate feature, It is a fusion of feature vectors. It is the weight matrix of the fully connected layer. This is the bias vector of the fully connected layer. For activation functions;

[0075]

[0076] in, This is a predicted service life value. This is the weight matrix of the final output layer. The final output layer's bias vector.

[0077] The time series sample set is divided into a training set, a validation set, and a test set according to a preset ratio, and the fusion model is trained. The distribution ratio of the lifetime prediction values ​​of the training set, the validation set, and the test set is the same.

[0078] Obtain the overall accuracy of the test set. If the overall accuracy does not reach the expected threshold, adjust the hierarchical parameters in the LSTM and GRU paths to optimize the fusion model.

[0079] The optimized fusion model is deployed on an edge server, and the parameters of the feature fusion layer are adjusted based on real-time data.

[0080] In summary, the compressor lifespan prediction method provided by this invention filters out parameters such as compressor vibration type and operating type, preprocesses them to generate quantitative indicators, extracts physical features and model-automated features from the quantitative indicators, generates a time-series sample set, and uses the LSTM path and GRU path set in parallel in the fusion model to capture the long-cycle aging features of the physical type in the time-series sample set, as well as efficiently process high-frequency fluctuation features. This allows the fusion model to simultaneously learn the effects of slow-changing aging trends and fast-changing operating conditions, improving robustness. This invention, by using a fusion model to process long-cycle aging features and high-frequency fluctuation features in the compressor in parallel, achieves higher overall accuracy and more comprehensive consideration of factors compared to existing lifespan prediction methods. Furthermore, the dual-path approach allows for adjustment of hierarchical parameters to adjust the overall accuracy, preserving stable features and adapting to the dynamic operating conditions of the compressor.

[0081] Example 2

[0082] In another aspect, the present invention also provides a compressor lifespan prediction system, please refer to [link / reference needed]. Figure 2The diagram shows a schematic of the compressor lifespan prediction system in Embodiment 2 of the present invention. The compressor lifespan prediction system includes:

[0083] The index quantification module 11 is used to acquire the operating parameters of the compressor, perform time series unification and preprocessing on the operating parameters, and generate quantitative indicators, wherein the quantitative indicators include long time series sensitive indicators and short time series fluctuation indicators.

[0084] Feature extraction module 12 is used to extract the physical features and model-automated features of the long-term sensitive index and the short-term fluctuation index to obtain a time-series sample set;

[0085] The evaluation and deployment module 13 is used to train and test the pre-built fusion model using a time-series sample set, obtain the evaluation results of the fusion model, and deploy it based on the evaluation results.

[0086] Example 3

[0087] In another aspect, the present invention provides a computer-readable storage medium having stored thereon one or more computer programs that, when executed by a processor, implement the above-described compressor lifespan prediction method.

[0088] Those skilled in the art will understand that the logic or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable storage medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable storage medium" can mean any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0089] More specific examples (a non-exhaustive list) of computer-readable storage media include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable storage media can even be paper or other suitable media on which the program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0090] Example 4

[0091] Figure 3 This is a structural block diagram of an electronic starter provided in Embodiment 4. The electronic starter includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the compressor lifespan prediction method in the above embodiments. Figure 3 The electronic starter 30 shown is merely an example and should not be construed as limiting the functionality and scope of use of the embodiments of the present invention.

[0092] like Figure 3 As shown, the electronic starter 30 can be manifested in the form of a general-purpose computing device, such as a server device. The components of the electronic starter 30 may include, but are not limited to: at least one processor 31, at least one memory 32, and a bus 33 connecting different system components (including memory 32 and processor 31).

[0093] Bus 33 includes a data bus, an address bus, and a control bus.

[0094] The memory 32 may include volatile memory, such as RAM 321 (random access memory), and / or cache memory 322, and may further include ROM 323 (read-only memory).

[0095] The memory 32 may also include a program tool 325 having a set (at least one) of program modules 324, including but not limited to: an operating system, one or more application programs, other program modules, and program data, each or some combination of these examples may include an implementation of a network environment.

[0096] The processor 31 executes various functional applications and data processing by running computer programs stored in the memory 32, such as the compressor life prediction method of the present invention as described above.

[0097] The electronic starter 30 can also communicate with one or more external devices 34 (e.g., keyboard, pointing device, etc.). This communication can be performed via I / O interface 35 (input / output interface). Furthermore, the electronic starter 30 can also communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN), and / or public networks, such as the Internet) via network adapter 36. Figure 3As shown, network adapter 36 communicates with other modules of the model-generated electronic starter 30 via bus 33. It should be understood that, although not shown in the figure, other hardware and / or software modules can be used in conjunction with the model-generated electronic starter 30, including but not limited to: microcode, device drivers, redundant processors, disk drive arrays, RAID (disk array) systems, tape drives, and data backup storage systems.

[0098] It should be noted that although several units / modules or sub-units / modules of the electronic starter have been mentioned in the detailed description above, this division is merely exemplary and not mandatory. In fact, according to embodiments of the invention, the features and functions of two or more units / modules described above can be embodied in one unit / module. Conversely, the features and functions of one unit / module described above can be further divided and embodied by multiple units / modules.

[0099] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0100] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this patent should be determined by the appended claims.

Claims

1. A method for predicting the service life of a compressor, characterized in that, The compressor lifespan prediction method includes: By collecting compressor operating parameters at different frequencies using different types of sensors with a unified time sequence, the operating parameters are cleaned, time-series aligned, and standardized to generate quantitative indicators. These quantitative indicators include long-time-series sensitive indicators and short-time-series fluctuation indicators. Based on the time-series nature of the quantitative indicators, the long-time-series sensitive indicators include at least the vibration signals and lubrication parameters of compressor components, while the short-time-series fluctuation indicators include at least the compressor's operating parameters and environmental parameters. The physical characteristics and model-automated characteristics of the long-term sensitive index and the short-term fluctuation index are extracted to obtain a time-series sample set. A time-series vibration curve is plotted based on the vibration signal. The time-series vibration curve is sampled using a first window, and linear regression fitting is performed on the sampling interval to obtain the time-series trend characteristics of the time-series vibration curve. At the same time, the vibration kurtosis and waveform factor of the curve segment in each sliding sampling unit in the first window are calculated, and the week-on-week growth rate and oil film thickness of the lubrication parameters are calculated. The time-series curves of the operating parameters and environmental parameters are sampled using a second window and a third window, respectively. Based on the segment data in the sliding sampling units of the second window and the third window, the coefficient of variation of the operating parameters and the power load rate correlation and corrected power of the environmental parameters are calculated. The length of the sliding sampling unit in any window is equal. The physical characteristics of the long-term sensitive index and the short-term fluctuation index obtained through the sliding sampling unit are summarized. The pre-built fusion model is trained and tested using a time-series sample set to obtain the evaluation results of the fusion model, and then deployed based on the evaluation results.

2. The compressor lifespan prediction method according to claim 1, characterized in that, After obtaining physical features, automatic features of the model are extracted. Using a sliding window with a preset length and a preset step size, the model is divided according to the preset window length and preset step size continuous quantization index based on the time axis to obtain several time segment samples. The dimensionality attribute of any time segment sample matches the window length and step size of the sliding window. Based on the fault thresholds of each component of the compressor and the fault downtime timestamps, the remaining lifetime labels of the time-series segment samples are calculated and obtained: in, For remaining lifetime label, This is the timestamp for the downtime due to the fault. This is the end timestamp of the time sequence segment sample; Based on the extracted physical features and the model's automatic features, the time-series sample set obtained from training the fusion model is output.

3. The compressor lifespan prediction method according to claim 1, characterized in that, The constructed fusion model includes an input layer, a feature extraction layer, and an output layer, where: The feature extraction layer includes parallel connected LSTM and GRU paths, which are connected to the output layer through a feature fusion layer. The output layer includes a fully connected layer and a final output layer. The LSTM path includes a first long-time extraction layer and a second long-time extraction layer arranged in series. The first long-time extraction layer has 256 neurons, and the second long-time extraction layer has 128 neurons. The GRU path includes a first short-time extraction layer and a second short-time extraction layer arranged in series. The first short-time extraction layer has 128 neurons, and the second short-time extraction layer has 64 neurons. The LSTM path and GRU path are concatenated using a feature fusion layer to output a fused feature vector.

4. The compressor lifespan prediction method according to claim 3, characterized in that, A fully connected layer is used to perform a non-linear mapping on the fused feature vector to obtain intermediate features, and then the intermediate features are linearly mapped through the final output layer to obtain the lifespan prediction value. in, As an intermediate feature, It is a fusion of feature vectors. It is the weight matrix of the fully connected layer. This is the bias vector of the fully connected layer. For activation functions; in, This is a predicted service life value. This is the weight matrix of the final output layer. The final output layer's bias vector.

5. The compressor lifespan prediction method according to claim 1, characterized in that, The steps of training and testing the pre-built fusion model using a time-series sample set, obtaining the evaluation results of the fusion model, and deploying it based on the evaluation results include: The time series sample set is divided into a training set, a validation set, and a test set according to a preset ratio, and the fusion model is trained. The distribution ratio of the lifetime prediction values ​​of the training set, the validation set, and the test set is the same. Obtain the overall accuracy of the test set. If the overall accuracy does not reach the expected threshold, adjust the hierarchical parameters in the LSTM and GRU paths to optimize the fusion model. The optimized fusion model is deployed on an edge server, and the parameters of the feature fusion layer are adjusted based on real-time data.

6. A compressor lifespan prediction system, characterized in that, The compressor lifespan prediction system is used to implement the compressor lifespan prediction method according to any one of claims 1-5, and the system includes: The index quantification module is used to collect compressor operating parameters at different frequencies using different types of sensors with a unified time series, perform data cleaning, time series alignment and standardization on the operating parameters, and generate quantitative indices. The quantitative indices include long-time series sensitive indices and short-time series fluctuation indices. Based on the time series nature of the quantitative indices, the quantitative indices include long-time series sensitive indices and short-time series fluctuation indices. The long-time series sensitive indices include at least the vibration signals and lubrication parameters of compressor components, and the short-time series fluctuation indices include at least the compressor's operating parameters and environmental parameters. The feature extraction module is used to extract the physical features and model-automated features of the long-term sensitive index and the short-term fluctuation index, obtain a time-series sample set, draw a time-series vibration curve based on the vibration signal, sample the time-series vibration curve using a first window, perform linear regression fitting on the sampling interval, obtain the time-series trend features of the time-series vibration curve, and simultaneously calculate the vibration kurtosis and waveform factor of the curve segment in each sliding sampling unit within the first window, as well as the week-on-week growth rate and oil film thickness of the lubrication parameters. The second and third windows are used to sample the time-series curves of the operating parameters and environmental parameters, respectively, and the coefficient of variation of the operating parameters and the power load rate correlation and corrected power of the environmental parameters are calculated based on the segment data in the sliding sampling units of the second and third windows. The length of the sliding sampling unit in any window is equal, and the physical features of the long-term sensitive index and the short-term fluctuation index obtained through the sliding sampling unit are summarized. The evaluation and deployment module is used to train and test the pre-built fusion model using a time-series sample set, obtain the evaluation results of the fusion model, and deploy it based on the evaluation results.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by the processor, the program implements the compressor life prediction method as described in any one of claims 1-5.

8. An electronic starter, 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 a computer program, it implements the compressor lifespan prediction method as described in any one of claims 1-5.

Images