A turbo expander fault warning method, device, equipment and medium

Abstract

The present invention relates to the field of turbine expanders, and discloses a turbine expander fault warning method, device, equipment and medium. Preprocessing is performed on multiple parameter value time series data of the turbine expander to obtain multiple corresponding processed time series data, and an autoencoder is used to reconstruct the processed time series data, and a target residual between the reconstructed data and the processed time series data is calculated. The conditional probability density function of the target residual is determined based on the target residual and the reconstructed data, and an adaptive threshold is determined based on a preset confidence. Whether to perform a fault warning is determined according to the relationship between the adaptive threshold and the target residual. The method can effectively realize real-time intelligent warning of the turbine expander. Compared with the related technology of performing fault warning by using a fixed threshold, the method has good universality and stability, is suitable for abnormal warning of turbine expanders under variable working conditions, and effectively reduces false alarms and missed alarms of fault warnings.

Description

Technical Field

[0001] The present invention relates to the field of turbine expanders, and in particular to a turbine expander fault early warning method, device, equipment and medium. Background Art

[0002] Turbine expanders can convert the pressure of high-pressure gas into mechanical energy and are key equipment in industries such as natural gas processing, chemical production and waste heat power generation.

[0003] The normal operation of the turbo expander is important for industrial production efficiency. Timely identification and early warning of turbo expander failures can effectively avoid production accidents and improve production efficiency. Related technologies generally set fixed thresholds for one or some operating parameters. When the parameter value of the operating parameter exceeds the fixed threshold, it is considered that the turbo expander is abnormal, and an early warning is issued at this time.

[0004] However, the fixed threshold may not be able to adapt to the changes in the operating conditions of the turboexpander and may no longer be applicable when the operating conditions change, resulting in false alarms and missed alarms. Summary of the invention

[0005] The present invention provides a turbine expander fault warning method, device, equipment and medium, which are used to solve the defects of fault false alarm and missed alarm in related technologies that use fixed thresholds for fault warning, and reduce fault false alarm and missed alarm.

[0006] In a first aspect, the present invention provides a turbo expander fault early warning method, comprising: During the operation period of the turboexpander, collecting time series data of parameter values ​​of a plurality of operating parameters; Preprocessing each of the parameter value time series data to obtain processed time series data corresponding to each of the parameter value time series data; Input each of the processed time series data into a trained autoencoder for data reconstruction to obtain reconstructed data corresponding to each of the processed time series data; Determine a target residual according to each of the processed time series data and the corresponding reconstructed data; Determining a conditional probability density function of the target residual based on the target residual and each of the reconstructed data; Determining an adaptive threshold value according to the conditional probability density function and a preset confidence level; Based on the size relationship between the adaptive threshold and the target residual, it is determined whether to issue a fault warning for the turbo expander.

[0007] Optionally, determining a target residual according to each of the processed time series data and the corresponding reconstructed data includes: For any of the processed time series data, calculating the difference between the processed time series data and the corresponding reconstructed data, and using the difference as the residual corresponding to the processed time series data; The residuals corresponding to all the processed time series data are averaged to obtain the residual mean and use it as the target residual.

[0008] Optionally, determining a conditional probability density function of the target residual based on the target residual and each of the reconstructed data includes: Performing density estimation on the target residual to obtain a corresponding first probability density function; performing density estimation on each of the reconstructed data to obtain a second probability density function corresponding to each of the reconstructed data; For any of the second probability density functions, determining a corresponding joint probability density estimation function based on the second probability density function and the first probability density function, and determining a ratio of the joint probability density estimation function to the second probability density function as a sub-conditional probability density function corresponding to the target residual; All the sub-conditional probability density functions corresponding to the target residual are taken as the conditional probability density function as a whole.

[0009] Optionally, determining the adaptive threshold according to the conditional probability density function and a preset confidence level includes: Integrate the conditional probability density function in the interval from negative infinity to the value to be solved to obtain a target integral formula; The integral result value of the integral expression is set as the preset confidence and the integral expression is solved to obtain a solution value corresponding to the value to be solved, and the solution value is used as the adaptive threshold.

[0010] Optionally, judging whether to issue a fault warning to the turbo expander based on the size relationship between the adaptive threshold and the target residual includes: Determining whether the target residual is greater than the adaptive threshold; If the target residual is greater than the adaptive threshold, it is determined that the turbo expander is abnormal, and a fault warning is issued for the turbo expander; If the target residual is not greater than the adaptive threshold, it is determined that the turbo expander is normal, and fault warning for the turbo expander is prohibited.

[0011] Optionally, the preprocessing each of the parameter value time series data to obtain processed time series data corresponding to each of the parameter value time series data includes: For any of the parameter value time series data, identify and delete abnormal values ​​in the parameter value time series data to obtain deleted time series data, perform missing value filling on the deleted time series data to obtain filled time series data, perform signal smoothing on the filled time series data to obtain smoothed time series data, normalize the smoothed time series data to obtain normalized time series data, add white Gaussian noise to the normalized time series data to obtain target time series data, and determine the target time series data as the processed time series data corresponding to the parameter value time series data.

[0012] Optionally, the multiple operating parameters include at least one of vibration value, torque, rotation speed, pressure, temperature, flow rate and power.

[0013] In a second aspect, the present invention provides a turbine expander fault warning device, comprising: A collection unit, used for collecting parameter value time series data of a plurality of operating parameters during the operation period of the turbo expander; A preprocessing unit, used for preprocessing each of the parameter value time series data to obtain processed time series data corresponding to each of the parameter value time series data; A reconstruction unit, used for inputting each of the processed time series data into a trained autoencoder for data reconstruction, to obtain reconstructed data corresponding to each of the processed time series data; A residual determination unit, configured to determine a target residual according to each of the processed time series data and the corresponding reconstructed data; A function determination unit, configured to determine a conditional probability density function of the target residual based on the target residual and each of the reconstructed data; A threshold determination unit, used to determine an adaptive threshold according to the conditional probability density function and a preset confidence level; The early warning unit is used to determine whether to issue a fault early warning to the turbine expander based on the size relationship between the adaptive threshold and the target residual.

[0014] In a third aspect, the present invention provides a computer device, comprising: a memory and a processor, the memory and the processor are communicatively connected to each other, computer instructions are stored in the memory, and the processor executes the turbine expander fault warning method of the first aspect or any corresponding embodiment thereof by executing the computer instructions.

[0015] In a fourth aspect, the present invention provides a computer-readable storage medium having computer instructions stored thereon, the computer instructions being used to enable a computer to execute the turbine expander fault warning method of the first aspect or any corresponding embodiment thereof.

[0016] The turbine expander fault warning method, device, equipment and medium provided by the present invention preprocess multiple parameter value time series data of the turbine expander to obtain corresponding multiple processed time series data, use an autoencoder to reconstruct the processed time series data, and calculate the target residual between the reconstructed data and the processed time series data, determine the conditional probability density function of the target residual based on the target residual and the reconstructed data, and determine an adaptive threshold based on a preset confidence, and determine whether to perform a fault warning based on the relationship between the adaptive threshold and the target residual. This can effectively achieve real-time intelligent warning of the turbine expander. Compared with the related technology of performing fault warning through a fixed threshold, this method has good universality and stability, is suitable for abnormal warning of turbine expanders under variable working conditions, and effectively reduces false alarms and missed alarms of fault warnings. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0018] Figure 1 A flow chart of a turbo expander fault warning method provided by an embodiment of the present invention; Figure 2 A flow chart of another turbo expander fault early warning method provided by an embodiment of the present invention; Figure 3 A schematic structural diagram of a turbo expander fault warning device provided by an embodiment of the present invention; Figure 4 A schematic diagram of the structure of a computer device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0019] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0020] In the related technologies, common typical failures of turboexpanders in industrial applications include equipment abnormalities and process abnormalities. Equipment abnormalities include bearing wear, abnormal equipment temperature, rotor imbalance, etc. Process abnormalities include abnormal changes in process parameters (such as heat exchanger structure failure, etc.) and abnormal changes in interference parameters (such as significant changes in ambient temperature, fluctuations in working fluid parameters, etc.).

[0021] In the related art, the mainstream fault warning method of the turbine expander is to calculate the alarm threshold based on the upper limit of the model estimation deviation or the lower limit of the similarity under normal conditions, and the warning information is triggered when the monitoring signal exceeds the threshold range. However, the fixed threshold has poor applicability and is difficult to cope with the changes of the equipment under different operating conditions. When the load fluctuates or the environment changes, the characteristic parameters will also deviate from the normal state with the change of the operating conditions, resulting in a decrease in the similarity with the normal conditions, and often omissions and false alarms will occur. In addition, setting a reasonable threshold requires a large amount of historical data and expert experience, and is highly subjective. Different operators may set different thresholds. The fixed threshold method has limited ability to identify complex faults caused by the combined action of multiple factors, and cannot effectively detect these potential faults. Both of the above situations will increase the probability of false alarms in the related art through fixed threshold warnings, and lack of intelligent warning methods for turbine expanders under variable operating conditions.

[0022] Combine the following Figure 1-Figure 2 The turbo expander fault early warning method of the present invention is described.

[0023] like Figure 1 As shown, this embodiment proposes a first turbine expander fault early warning method, which may include the following steps: S101. During the operation period of the turbo expander, collect time series data of parameter values ​​of a plurality of operating parameters.

[0024] The operating period may be a period during the operation of the turbo expander, and the specific duration of the operating period may be 1 hour, 1 day or 1 week, which is not limited in this embodiment.

[0025] The operating parameters refer to the operating parameters of the turbo expander.

[0026] Specifically, the parameter value time series data is a sequence of multiple parameter values ​​of a certain operating parameter of the turboexpander within an operating period, which are arranged in chronological order.

[0027] Specifically, in this embodiment, for any operating parameter, the parameter value of the operating parameter can be collected by a sensor. In this embodiment, multiple sensors, such as a temperature sensor and a rotation speed sensor, can be used to collect the time series data of the parameter value of each operating parameter during the operating period.

[0028] Optionally, the above-mentioned multiple operating parameters include at least one of vibration value, torque, rotation speed, pressure, temperature, flow rate and power.

[0029] Specifically, in this embodiment, the parameter value time series data of multiple operating parameters can be collected in real time by sensors, including the parameter value time series data of operating parameters such as vibration value, rotation speed and temperature. The set of the collected parameter value time series data can be expressed as: .

[0030] in, is the collection of collected parameter value time series data, For the The parameter value time series data of operating parameters, totaling operating parameters.

[0031] S102: Preprocess each parameter value time series data to obtain processed time series data corresponding to each parameter value time series data.

[0032] Specifically, this embodiment may preprocess each parameter value time series data to eliminate the impact of the environment and improve the setting accuracy of the adaptive threshold.

[0033] It should be noted that the preprocessing may include removing outliers, filling missing values, signal smoothing, normalization, and adding white Gaussian noise.

[0034] Optionally, step S102 may include: For any parameter value time series data, identify and delete abnormal values ​​in the parameter value time series data to obtain deleted time series data, fill missing values ​​in the deleted time series data to obtain filled time series data, perform signal smoothing on the filled time series data to obtain smoothed time series data, normalize the smoothed time series data to obtain normalized time series data, add white Gaussian noise to the normalized time series data to obtain target time series data, and determine the target time series data as the processed time series data corresponding to the parameter value time series data.

[0035] Specifically, regarding the removal of outliers, outliers may be caused by sensor measurement errors or other interference factors, which will have a greater impact on the training and prediction of the model. In this embodiment, the Z-score calculation formula can be used to identify outliers. The Z-score formula is:

[0036] in, For the The threshold corresponding to the time series data of parameter values, For the The parameter value time series data parameter values, For the The mean of all parameter values ​​in the parameter value time series data, For the The standard deviation of all parameter values ​​in the parameter value time series data.

[0037] It is understandable that if a parameter value exceeds the corresponding threshold , it is considered an outlier and is removed. , it is considered to be a normal value and is retained.

[0038] Specifically, regarding missing value filling, the missing values ​​of sensor signals will lead to inaccurate calculation results or even failure to perform calculations. Therefore, for missing parameter values, the mean of the sequence can be used to fill them:

[0039] in, For the The missing values ​​that need to be filled in the time series data after deletion, For the After deleting the time series data parameter values, For the The total number of parameter values ​​in the time series data after deletion of non-missing values.

[0040] Specifically, regarding signal smoothing, this embodiment requires smoothing to remove noise from the time series data and make it smoother and more continuous. This embodiment can use an exponentially weighted moving average method to smooth the time series data. Take the filled time series data as an example:

[0041] in, For the The time in the smoothed time series data The smoothed value of For the The time in the smoothed time series data Parameter value, is the smoothing factor (the value can be 0.2), For the The time in the smoothed time series data The smoothing value of .

[0042] Specifically, regarding normalization, in order to eliminate the impact between data of different scales and make the features comparable, maximum and minimum normalization is used to scale the data to the range of 0 to 1.

[0043] Specifically, with regard to adding white Gaussian noise, in order to enhance the effectiveness of the model's adaptive threshold under variable working conditions and ensure the accuracy of the model's reconstruction of signals under diverse environments, this embodiment needs to add white Gaussian noise with different signal-to-noise ratios. Take a time series data as an example, the signal after adding noise can be expressed as:

[0044] in, For the Target time series data, For the Normalized time series data, is white Gaussian noise.

[0045] It should be noted that this embodiment uses signal preprocessing and usefulness screening to eliminate outliers, fill missing values, smooth and normalize the signals of the collected turbine expander sensor signal sequence, thereby reducing the impact of the equipment environment on signal acquisition and transmission, and ensuring data quality and model performance.

[0046] S103, input each processed time series data into the trained autoencoder for data reconstruction, and obtain the reconstructed data corresponding to each processed time series data.

[0047] Specifically, this embodiment can train the relevant model based on the historical operation data of the turbo expander to obtain a trained autoencoder.

[0048] It should be noted that the autoencoder may include an encoder and a decoder. In order to obtain the reconstruction error to obtain the adaptive warning threshold, this embodiment constructs and trains an autoencoder to reconstruct the noise-free signal, and determines the adaptive warning threshold according to the output of the autoencoder to achieve the setting of the adaptive threshold under the variable working condition of the turbine expander.

[0049] Specifically, in this embodiment, when training the autoencoder, the encoder can compress the input time series data into a low-dimensional latent variable space, and the decoder reconstructs the time series data from the latent variable space. The autoencoder makes the data as close to the original reconstructed data as possible, and the loss function used is the mean square error.

[0050] It should be noted that the autoencoder reconstructs the original signal. In order to obtain the reconstruction error to obtain the adaptive warning threshold, the noise-free signal is reconstructed by building and training the autoencoder. The adaptive threshold is determined according to the output of the autoencoder to realize the setting of the adaptive threshold under the variable operating conditions of the turbine expander.

[0051] S104: Determine a target residual according to each processed time series data and the corresponding reconstructed data.

[0052] Specifically, this embodiment can calculate the target residual based on all processed time series data and reconstructed data. The target residual can be used to measure the gap between the processed time series data and the reconstructed data.

[0053] Optionally, step S104 may include: For any processed time series data, the difference between the processed time series data and the corresponding reconstructed data is calculated and used as the residual corresponding to the processed time series data; The residuals corresponding to all processed time series data are averaged to obtain the residual mean and use it as the target residual.

[0054] Specifically, for The processed time series data is used to calculate the reconstructed data from the encoder. The residual can be defined as the difference between the reconstructed data and the processed time series data. Take the processed time series data as an example, the residual can be expressed as:

[0055] in, For the The residual corresponding to the processed time series data is For the After processing, time series data, For the The reconstructed data corresponding to the processed time series data.

[0056] Since multiple processed time series data are integrated, each processed time series data has a corresponding residual, so the final target residual is the mean of all residuals:

[0057] in, is the total number of time series data after processing, For the The residual corresponding to the processed time series data is is the target residual.

[0058] S105: Determine a conditional probability density function of the target residual based on the target residual and each reconstructed data.

[0059] Specifically, this embodiment may determine the conditional probability density function of the target residual based on the target residue and all reconstructed data.

[0060] Optionally, step S105 may include: Density estimation is performed on the target residual to obtain a corresponding first probability density function; density estimation is performed on each reconstructed data to obtain a second probability density function corresponding to each reconstructed data; For any second probability density function, based on the second probability density function and the first probability density function, determine a corresponding joint probability density estimation function, and determine a ratio of the joint probability density estimation function to the second probability density function as a sub-conditional probability density function corresponding to the target residual; All sub-conditional probability density functions corresponding to the target residual are taken as the conditional probability density function as a whole.

[0061] Specifically, this embodiment can perform kernel density estimation on the target residual to obtain the probability density function of the target residual: At the same time, this embodiment also reconstructs each data Perform kernel density estimation to obtain the probability density function of each reconstructed data , through the joint probability density estimation function and reconstruct the marginal distribution of the data , calculate the sub-conditional probability density function of the target residual:

[0062] in, .

[0063] It should be noted that the adaptive threshold warning calculates the joint distribution of the residual and the reconstructed signal through kernel density estimation to obtain the adaptive threshold under variable operating conditions. Through dynamic adjustment and smoothing, the adaptive threshold can be used for real-time turbine expander fault warning. Compared with the fixed threshold warning in related technologies, it has good universality and stability, and is suitable for abnormal warning of turbine expanders under variable operating conditions.

[0064] S106: Determine an adaptive threshold according to the conditional probability density function and a preset confidence level.

[0065] The preset confidence level may be a confidence level set by a technician according to actual conditions, and this embodiment does not limit its specific size.

[0066] Optionally, step S106 may include: Integrate the conditional probability density function in the interval from negative infinity to the value to be solved to obtain the target integral formula; The integral result value of the integral expression is set as a preset confidence level and the integral expression is solved to obtain a solution value corresponding to the value to be solved, and the solution value is used as an adaptive threshold.

[0067] Specifically, this embodiment can rotate a suitable confidence level As the preset confidence, according to the marginal distribution of the target residual Determining the adaptive threshold , the following conditions are met:

[0068] It should be noted that the adaptive threshold To set the confidence level Under this condition, the residual falls within the adaptive threshold Values ​​within.

[0069] S107: Based on the relationship between the adaptive threshold and the target residual, determine whether to issue a fault warning for the turbine expander.

[0070] Specifically, this embodiment can compare the size relationship between the adaptive threshold and the target residual, determine whether there is an abnormality in the turbine expander based on the size relationship, and determine whether to issue a fault warning for the turbine expander.

[0071] Optionally, step S107 may include: Determine whether the target residual is greater than the adaptive threshold; If the target residual is greater than the adaptive threshold, it is determined that the turbo expander is abnormal, and a fault warning is issued for the turbo expander; If the target residual is not greater than the adaptive threshold, it is determined that the turbo expander is normal, and fault warning for the turbo expander is prohibited.

[0072] It should be noted that the present embodiment can first calculate the target residual, and then use kernel density estimation to estimate the distribution of the target residual. In addition to the target residual, the present embodiment can also perform kernel density estimation on the reconstructed data to determine its distribution. Then determine the joint distribution between the target residual and the reconstructed data to understand how the reconstructed signal affects the target residual. The present embodiment calculates the possibility of the residual under given reconstructed data by analyzing the relationship between the target residual and the reconstructed data. According to the calculated error distribution, a threshold, namely an adaptive threshold, is set, indicating that the residual will not exceed this value under normal circumstances. The adaptive threshold is adjusted according to a certain confidence level so that the probability of the residual exceeding the adaptive threshold is very low. If the residual exceeds the set adaptive threshold at a certain time, it can be determined that the data is abnormal, that the turbine expander is abnormal, and a fault warning is performed, thereby effectively realizing a fault warning for the turbine expander.

[0073] In practical applications, this embodiment can be verified using test data, indicating that the above method can adjust the optimal warning threshold in real time under variable operating conditions, and can effectively warn of abnormalities of the turbine expander under variable operating conditions.

[0074] Specifically, this embodiment reconstructs the original signal from the noisy signal through the autoencoder, and calculates the residual between the reconstructed signal and the original signal, uses kernel density estimation to estimate the joint probability density function of the reconstructed signal and the residual, determines the residual critical value corresponding to the reconstructed signal under variable operating conditions according to a predefined confidence level, compares the residual according to the calculated dynamic threshold, and realizes real-time intelligent early warning of the turbine expander. Compared with the related technology of early warning through fixed threshold, it has good universality and stability, and is suitable for abnormal early warning of turbine expanders under variable operating conditions.

[0075] This embodiment aims to solve the problems of missed alarms and false alarms caused by changes in operating conditions in turbine expander fault warnings. The original signal is reconstructed from a noisy signal through an autoencoder, and the residual between the reconstructed signal and the original signal is calculated. The joint probability density function of the reconstructed signal and the residual is estimated using a kernel density estimation method. According to a predefined confidence level, the residual critical value corresponding to the reconstructed signal under variable operating conditions is determined, and the residuals are compared according to the calculated dynamic threshold to achieve real-time intelligent early warning of the turbine expander.

[0076] The turbine expander fault warning method eliminated in the present embodiment preprocesses multiple parameter value time series data of the turbine expander to obtain multiple corresponding processed time series data, uses an autoencoder to reconstruct the processed time series data, and calculates the target residual between the reconstructed data and the processed time series data, determines the conditional probability density function of the target residual based on the target residual and the reconstructed data, and determines an adaptive threshold based on a preset confidence, and determines whether to perform a fault warning based on the relationship between the adaptive threshold and the target residual. This method can effectively realize real-time intelligent warning of the turbine expander. Compared with the related technology of performing fault warning through fixed thresholds, this method has good universality and stability, is suitable for abnormal warning of turbine expanders under variable working conditions, and effectively reduces false alarms and missed alarms of fault warnings.

[0077] like Figure 2 As shown, this embodiment proposes a second turbine expander fault warning method, in which this embodiment includes a training phase and a testing phase.

[0078] During the training phase, this embodiment can obtain the historical vibration signal of the turbine expander, and then perform signal preprocessing and usefulness screening, including missing value filling, outlier removal, signal smoothing and normalization, and then add Gaussian white noise, and perform hyperparameter setting and autoencoder training. After the autoencoder is trained, the autoencoder can be used to set an adaptive threshold.

[0079] During the testing phase, this embodiment can collect the field vibration signal of the turbine expander, perform signal preprocessing and usefulness screening on the field vibration signal, and then calculate the residual based on the trained autoencoder to determine whether the residual is greater than the adaptive threshold. If so, it can be determined that the turbine expander is abnormal and a fault warning is issued. If not, it can be determined that the turbine expander is normal and fault warning is prohibited.

[0080] The turbine expander fault warning method proposed in this embodiment can be applied to the turbine expander adaptive fault warning under variable operating conditions.

[0081] like Figure 3 As shown, this embodiment provides a turbine expander fault warning device, which may include: The collection unit 301 is used to collect parameter value time series data of multiple operating parameters during the operation period of the turbo expander; A preprocessing unit 302 is used to preprocess each parameter value time series data to obtain processed time series data corresponding to each parameter value time series data; A reconstruction unit 303 is used to input each processed time series data into the trained autoencoder for data reconstruction to obtain reconstructed data corresponding to each processed time series data; A residual determination unit 304, configured to determine a target residual according to each processed time series data and the corresponding reconstructed data; A function determination unit 305, configured to determine a conditional probability density function of the target residual based on the target residual and each reconstructed data; A threshold determination unit 306, configured to determine an adaptive threshold according to a conditional probability density function and a preset confidence level; The early warning unit 307 is used to determine whether to issue a fault early warning to the turbo expander based on the size relationship between the adaptive threshold and the target residual.

[0082] It should be noted that the processing processes of the acquisition unit 301, the preprocessing unit 302, the reconstruction unit 303, the residual determination unit 304, the function determination unit 305, the threshold determination unit 306 and the early warning unit 307 and the beneficial effects thereof can be respectively referred to in Figure 1 Steps S101 to S107 in the above are not described in detail.

[0083] Optionally, the residual determination unit 304 is further configured to: For any processed time series data, the difference between the processed time series data and the corresponding reconstructed data is calculated and used as the residual corresponding to the processed time series data; The residuals corresponding to all processed time series data are averaged to obtain the residual mean and use it as the target residual.

[0084] Optionally, the function determination unit 305 is further configured to: Density estimation is performed on the target residual to obtain a corresponding first probability density function; density estimation is performed on each reconstructed data to obtain a second probability density function corresponding to each reconstructed data; For any second probability density function, based on the second probability density function and the first probability density function, determine a corresponding joint probability density estimation function, and determine a ratio of the joint probability density estimation function to the second probability density function as a sub-conditional probability density function corresponding to the target residual; All sub-conditional probability density functions corresponding to the target residual are taken as the conditional probability density function as a whole.

[0085] Optionally, the threshold determination unit 306 is further configured to: Integrate the conditional probability density function in the interval from negative infinity to the value to be solved to obtain the target integral formula; The integral result value of the integral expression is set as a preset confidence level and the integral expression is solved to obtain a solution value corresponding to the value to be solved, and the solution value is used as an adaptive threshold.

[0086] Optionally, the early warning unit 307 is further configured to: Determine whether the target residual is greater than the adaptive threshold; If the target residual is greater than the adaptive threshold, it is determined that the turbo expander is abnormal, and a fault warning is issued for the turbo expander; If the target residual is not greater than the adaptive threshold, it is determined that the turbo expander is normal, and fault warning for the turbo expander is prohibited.

[0087] Optionally, the preprocessing unit 302 is further configured to: For any parameter value time series data, identify and delete abnormal values ​​in the parameter value time series data to obtain deleted time series data, fill missing values ​​in the deleted time series data to obtain filled time series data, perform signal smoothing on the filled time series data to obtain smoothed time series data, normalize the smoothed time series data to obtain normalized time series data, add white Gaussian noise to the normalized time series data to obtain target time series data, and determine the target time series data as the processed time series data corresponding to the parameter value time series data.

[0088] Optionally, the multiple operating parameters include at least one of vibration value, torque, rotation speed, pressure, temperature, flow rate and power.

[0089] The turbine expander fault warning device eliminated in this embodiment preprocesses multiple parameter value time series data of the turbine expander to obtain corresponding multiple processed time series data, uses an autoencoder to reconstruct the processed time series data, and calculates the target residual between the reconstructed data and the processed time series data, determines the conditional probability density function of the target residual based on the target residual and the reconstructed data, and determines an adaptive threshold based on a preset confidence, and determines whether to perform a fault warning based on the relationship between the adaptive threshold and the target residual. This can effectively achieve real-time intelligent warning of the turbine expander. Compared with the related technology of performing fault warning through fixed thresholds, this device has good universality and stability, is suitable for abnormal warning of turbine expanders under variable working conditions, and effectively reduces false alarms and missed alarms of fault warnings.

[0090] The turbine expander fault warning device in this embodiment is presented in the form of a functional unit, where the unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that executes one or more software or fixed programs, and / or other devices that can provide the above functions.

[0091] The embodiment of the present invention also provides a computer device having the above Figure 3 The turbo expander fault warning device shown.

[0092] See also Figure 4 , a schematic diagram of the structure of a computer device provided by an optional embodiment of the present invention, the computer device includes: one or more processors 10, a memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. The various components are connected to each other using different buses for communication, and can be installed on a common motherboard or installed in other ways as needed. The processor can process instructions executed in the computer device, including instructions stored in or on the memory to display graphical information of the GUI on an external input / output device (such as a display device coupled to the interface). In some optional embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Similarly, multiple computer devices can be connected, and each device provides part of the necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system). Figure 4 A processor 10 is taken as an example.

[0093] The processor 10 may be a central processing unit, a network processor or a combination thereof. The processor 10 may further include a hardware chip. The hardware chip may be a dedicated integrated circuit, a programmable logic device or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general purpose array logic or any combination thereof.

[0094] The memory 20 stores instructions executable by at least one processor 10, so that at least one processor 10 executes the method shown in the above embodiment.

[0095] The memory 20 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function. The data storage area may store data created according to the use of the computer device, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage devices. In some optional embodiments, the memory 20 may optionally include a memory remotely arranged relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0096] The memory 20 may include a volatile memory, such as a random access memory. The memory may also include a non-volatile memory, such as a flash memory, a hard disk or a solid state drive. The memory 20 may also include a combination of the above-mentioned types of memory.

[0097] The computer device further comprises a communication interface 30 for the computer device to communicate with other devices or a communication network.

[0098] The embodiment of the present invention also provides a computer-readable storage medium. The method according to the embodiment of the present invention can be implemented in hardware, firmware, or can be implemented as a computer code that can be recorded in a storage medium, or can be implemented as a computer code that is originally stored in a remote storage medium or a non-temporary machine-readable storage medium and will be stored in a local storage medium through a network download, so that the method described herein can be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only storage memory, a random access memory, a flash memory, a hard disk or a solid-state hard disk, etc.; further, the storage medium can also include a combination of the above types of memories. It can be understood that a computer, a processor, a microprocessor controller, or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by a computer, a processor, or hardware, the method shown in the above embodiment is implemented.

[0099] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A turbo expander fault early warning method, characterized in that: include: During the operation period of the turboexpander, collecting parameter value time series data of a plurality of operating parameters; Preprocessing each of the parameter value time series data to obtain processed time series data corresponding to each of the parameter value time series data; Input each of the processed time series data into a trained autoencoder for data reconstruction to obtain reconstructed data corresponding to each of the processed time series data; Determine a target residual according to each of the processed time series data and the corresponding reconstructed data; Determining a conditional probability density function of the target residual based on the target residual and each of the reconstructed data; Determining an adaptive threshold value according to the conditional probability density function and a preset confidence level; Based on the size relationship between the adaptive threshold and the target residual, it is determined whether to issue a fault warning for the turbo expander.

2. The method according to claim 1, characterized in that The determining the target residual according to each of the processed time series data and the corresponding reconstructed data comprises: For any of the processed time series data, calculating the difference between the processed time series data and the corresponding reconstructed data, and using the difference as the residual corresponding to the processed time series data; The residuals corresponding to all the processed time series data are averaged to obtain the residual mean and use it as the target residual.

3. The method according to claim 1, characterized in that: The step of determining a conditional probability density function of the target residual based on the target residual and each of the reconstructed data comprises: Performing density estimation on the target residual to obtain a corresponding first probability density function; performing density estimation on each of the reconstructed data to obtain a second probability density function corresponding to each of the reconstructed data; For any of the second probability density functions, determining a corresponding joint probability density estimation function based on the second probability density function and the first probability density function, and determining a ratio of the joint probability density estimation function to the second probability density function as a sub-conditional probability density function corresponding to the target residual; All the sub-conditional probability density functions corresponding to the target residual are taken as the conditional probability density function as a whole.

4. The method according to claim 1, characterized in that The step of determining the adaptive threshold according to the conditional probability density function and the preset confidence level includes: Integrate the conditional probability density function in the interval from negative infinity to the value to be solved to obtain a target integral formula; The integral result value of the integral expression is set as the preset confidence and the integral expression is solved to obtain a solution value corresponding to the value to be solved, and the solution value is used as the adaptive threshold.

5. The method according to claim 1, characterized in that: The determining whether to perform a fault warning on the turbo expander based on the magnitude relationship between the adaptive threshold and the target residual includes: Determining whether the target residual is greater than the adaptive threshold; If the target residual is greater than the adaptive threshold, it is determined that the turbo expander is abnormal, and a fault warning is issued for the turbo expander; If the target residual is not greater than the adaptive threshold, it is determined that the turbo expander is normal, and fault warning for the turbo expander is prohibited.

6. The method according to claim 1, characterized in that The preprocessing of each of the parameter value time series data to obtain processed time series data corresponding to each of the parameter value time series data includes: For any of the parameter value time series data, identify and delete abnormal values ​​in the parameter value time series data to obtain deleted time series data, perform missing value filling on the deleted time series data to obtain filled time series data, perform signal smoothing on the filled time series data to obtain smoothed time series data, normalize the smoothed time series data to obtain normalized time series data, add white Gaussian noise to the normalized time series data to obtain target time series data, and determine the target time series data as the processed time series data corresponding to the parameter value time series data.

7. The method according to any one of claims 1 to 6, characterized in that The plurality of operating parameters include at least one of vibration value, torque, rotation speed, pressure, temperature, flow rate and power.

8. A turbo expander fault warning device, characterized in that: include: A collection unit, used for collecting parameter value time series data of a plurality of operating parameters during the operation period of the turbo expander; A preprocessing unit, used for preprocessing each of the parameter value time series data to obtain processed time series data corresponding to each of the parameter value time series data; A reconstruction unit, used for inputting each of the processed time series data into a trained autoencoder for data reconstruction, to obtain reconstructed data corresponding to each of the processed time series data; A residual determination unit, configured to determine a target residual according to each of the processed time series data and the corresponding reconstructed data; A function determination unit, configured to determine a conditional probability density function of the target residual based on the target residual and each of the reconstructed data; A threshold determination unit, used to determine an adaptive threshold according to the conditional probability density function and a preset confidence level; The early warning unit is used to determine whether to issue a fault early warning to the turbine expander based on the size relationship between the adaptive threshold and the target residual.

9. A computer device, characterized in that: include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the turbine expander fault warning method according to any one of claims 1 to 7 by executing the computer instructions.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the turbine expander fault warning method according to any one of claims 1 to 7.

Images