CN113632026A - Fault diagnosis method and system for rotary mechanical equipment and storage medium - Google Patents

Abstract

A method of fault diagnosis for rotating machinery, comprising: acquiring operation data on at least one detection point in a rotating mechanical device (S110); respectively performing anomaly detection analysis on at least one anomaly type of the detection points by using the operation data to obtain an analysis result corresponding to each anomaly type (S120); and diagnosing at least one fault in the rotary mechanical equipment by using the obtained at least one analysis result to output a corresponding fault diagnosis result (S130). By continuously reducing the dimension of the acquired data and presetting key indexes, the data are subjected to feature extraction, and the accuracy is ensured under the condition of a small amount of data samples.

Description

Fault diagnosis method and system for rotary mechanical equipment and storage medium Technical Field

The present disclosure relates to the field of fault detection technologies, and in particular, to a method and a system for fault diagnosis of a rotating machine, and a storage medium.

Background

When the rotary mechanical equipment in industrial production breaks down, the phenomena of equipment vibration abnormity and the like can be caused, and the possible faults of the equipment can be diagnosed by analyzing the vibration condition of the equipment. However, a large amount of data samples are generally required as input when a fault diagnosis model is constructed, and the number of samples in actual industrial production is limited, so that the accuracy of diagnosing abnormal phenomena by the fault diagnosis method in the prior art is low.

Disclosure of Invention

In view of the above-mentioned shortcomings of the prior art, the present application aims to provide a fault diagnosis method, a fault diagnosis system and a storage medium for a rotary machine, which are used to solve the problem of low diagnosis accuracy caused by the fact that the fault diagnosis method in the prior art cannot acquire a large number of sample inputs for training.

To achieve the above and other related objects, a first aspect of the present application provides a fault diagnosis method for a rotary mechanical apparatus, including the steps of: acquiring operation data on at least one detection point in a rotating mechanical device; respectively carrying out anomaly detection analysis on at least one anomaly type of the detection points by using the running data to obtain an analysis result corresponding to each anomaly type; and diagnosing at least one fault in the rotary mechanical equipment by using the obtained at least one analysis result so as to output a corresponding fault diagnosis result.

In certain embodiments of the first aspect of the present application, the anomaly type includes at least one of the following anomalies reflected by respective detection points during operation of the rotating mechanical equipment based on operating conditions and process requirements: an abnormality type set separately based on each vibration abnormality of at least one spatial dimension, an abnormality type set based on a temperature abnormality, an abnormality type set separately based on each power abnormality.

In certain embodiments of the first aspect of the present application, the operational data is used to generate the detection data and the reference data; the step of respectively performing anomaly detection analysis on at least one anomaly type of the detection points by using the operation data comprises the following steps: performing feature extraction on the acquired detection data based on the reference data of the detection point during normal operation of the rotating mechanical equipment to obtain at least one feature element for detecting at least one abnormal type of the detection point; and analyzing the characteristic elements corresponding to each abnormal type to obtain an analysis result of the corresponding abnormal type.

In certain embodiments of the first aspect of the present application, further comprising the steps of: and performing further data processing on the obtained at least one characteristic element to analyze the at least one characteristic element after the further data processing, so as to obtain an analysis result of the corresponding abnormal type.

In certain embodiments of the first aspect of the present application, the reference data comprises at least one or more of: the working condition data of the rotating mechanical equipment is data extracted from process data in the acquired operation data, and the data extracted from mechanical vibration data in the acquired operation data.

In certain embodiments of the first aspect of the present application, further comprising the steps of: and analyzing the fundamental frequency data in the corresponding reference data when the rotating mechanical equipment normally operates from the acquired operation data.

In certain embodiments of the first aspect of the present application, the acquired operational data comprises mechanical vibration data; the step of performing feature extraction on the acquired detection data based on the reference data of the detection point during normal operation of the rotating machinery to obtain at least one feature element for detecting at least one abnormal type of the detection point comprises: extracting at least one frequency feature element in the acquired frequency spectrum of the mechanical vibration data, which is related to the fundamental frequency data in the reference data of the detection point, for detecting at least one anomaly type of the detection point.

In certain embodiments of the first aspect of the present application, the acquired operational data comprises process data and/or operating condition data; the method further comprises the steps of: and analyzing the acquired process data and/or working condition data to obtain benchmark data in the reference data of the detection point during normal operation of the rotary mechanical equipment.

In certain embodiments of the first aspect of the present application, the step of performing feature extraction on the acquired detection data based on reference data of the detection point during normal operation of the rotating machine to obtain at least one feature element for detecting at least one anomaly type of the detection point comprises: obtaining at least one deviation feature element based on a deviation of the acquired detection data from the reference data in normal operation of the rotating mechanical device, for detecting at least one anomaly type of the detection point.

In certain embodiments of the first aspect of the present application, the step of performing a diagnostic process on at least one fault in the rotating mechanical device using the obtained at least one analysis result to output a corresponding fault diagnosis result includes: presenting the obtained fault diagnosis result in a display interface of a control system of the rotary mechanical equipment.

A second aspect of the present application further provides a server, including: the interface unit is used for carrying out data communication with the sensor on at least one dimension on the detection point of the rotary mechanical equipment; a storage unit for storing at least one program; and a processing unit for calling the at least one program to coordinate the interface unit and the storage unit to execute and implement the fault diagnosis method for the rotating machinery according to any one of the first aspect of the present application.

The third aspect of the present application also provides a first failure diagnosis system of a rotary mechanical apparatus, including: the server according to the second aspect of the present application; and the detection device is arranged at each detection point of the rotating mechanical equipment, is in communication connection with the server and is used for providing operation data of each detection point.

The fourth aspect of the present application also provides a computer-readable storage medium storing at least one program that, when invoked, executes and implements the rotating machine equipment failure diagnosis method according to any one of the first aspects of the present application.

The fifth aspect of the present application also provides a second failure diagnosis system of a rotary mechanical apparatus, including: the data acquisition module is used for acquiring operation data on at least one detection point in a piece of rotary mechanical equipment; and the data processing module is used for respectively carrying out abnormality detection analysis on at least one abnormality type of the detection points by using the operation data so as to obtain an analysis result corresponding to each abnormality type, and diagnosing at least one fault in the rotary mechanical equipment by using the obtained at least one analysis result so as to output a corresponding fault diagnosis result.

In certain embodiments of the fifth aspect of the present application, the anomaly type includes at least one of the following anomalies reflected by respective detection points during operation of the rotating mechanical device based on operating conditions and process requirements: an abnormality type set separately based on each vibration abnormality of at least one spatial dimension, an abnormality type set based on a temperature abnormality, an abnormality type set separately based on each power abnormality.

In certain embodiments of the fifth aspect of the present application, the operational data is used to generate the test data and the reference data; the data processing module performs feature extraction on the acquired detection data based on the reference data of the detection points during normal operation of the rotary mechanical equipment to obtain at least one feature element for detecting at least one abnormal type of the detection points; and analyzing the characteristic elements corresponding to each abnormal type to obtain the analysis result of the corresponding abnormal type.

In some embodiments of the fifth aspect of the present application, the data processing module performs further data processing on the obtained at least one feature element, so as to analyze the at least one feature element after the further data processing, thereby obtaining an analysis result of the corresponding exception type.

In certain embodiments of the fifth aspect of the present application, the reference data comprises at least one or more of: the working condition data of the rotating mechanical equipment is data extracted from process data in the acquired operation data, and the data extracted from mechanical vibration data in the acquired operation data.

In certain embodiments of the fifth aspect of the present application, the data processing module further analyzes fundamental frequency data in the reference data corresponding to normal operation of the rotating machine from the acquired operation data.

In certain embodiments of the fifth aspect of the present application, the acquired operational data comprises mechanical vibration data, and the data processing module extracts at least one frequency feature element in the frequency spectrum of the acquired mechanical vibration data associated with the fundamental frequency data in the reference data of the detection point for detecting at least one anomaly type of the detection point.

In certain embodiments of the fifth aspect of the present application, the acquired operational data comprises process data and/or operating condition data; the data processing module is also used for analyzing the acquired process data and/or working condition data to obtain datum data in the reference data of the detection point during normal operation of the rotary mechanical equipment.

In certain embodiments of the fifth aspect of the present application, the data processing module derives at least one deviation feature for detecting at least one anomaly type of the detection point based on a deviation of the acquired detection data from the reference data in the reference data during normal operation of the rotating machine.

In certain embodiments of the fifth aspect of the present application, a fault display device is further included to present the obtained fault diagnosis result.

A sixth aspect of the present application also provides a management system of a rotary mechanical device, including: the detection device is arranged at each detection point of the rotating mechanical equipment and used for providing operation data of each detection point; the control system of the rotating mechanical equipment is in data connection with each detection device and used for collecting and forwarding each operation data; the server according to the second aspect of the present application is communicatively connected to the control system, and configured to receive each piece of operation data and execute a corresponding fault diagnosis method based on the received operation data.

As described above, the method, system, and storage medium for diagnosing a fault of a rotating machine according to the present application have the following advantageous effects: according to the method and the device, the acquired data are subjected to continuous dimensionality reduction processing, and the data are subjected to feature extraction through preset key indexes, so that the accuracy is guaranteed under the condition of a small amount of data samples. Secondly, the method can flexibly expand the process parameters, the newly added parameters do not affect the machine learning model of the existing parameters, the existing model does not need to be retrained, only the newly added parameters need to be newly built, and the relation model of the newly added parameters and the existing parameters is added into the comprehensive diagnosis model. In addition, the three-layer model structure comprises a characteristic mechanism model and a machine learning model, and the position, type and the like of the detection point are analyzed through a comprehensive diagnosis model, so that management experience is fused with a machine learning algorithm, and the accuracy of a fault diagnosis result is ensured.

Drawings

Fig. 1 is a schematic view showing a fault diagnosis method of a rotary machine according to the present application.

Fig. 2 is a schematic view of an embodiment of setting a detection point on a wind turbine in the present application.

Fig. 3a to 3b are schematic diagrams illustrating an embodiment of the fault diagnosis system for performing abnormality detection analysis according to the present application.

Fig. 4 a-4 b are schematic diagrams illustrating the time domain waveforms of fig. 3a and 3b being converted into frequency domain spectrums according to the present application.

FIG. 5 is a schematic diagram of an embodiment of an anomaly detection analysis using operational data as described herein.

Fig. 6 is a schematic diagram of an embodiment of a fault diagnosis process performed in the present application.

Fig. 7 is a schematic diagram illustrating an exemplary architecture of a server in the present application.

Detailed Description

The following description of the embodiments of the present application is provided for illustrative purposes, and other advantages and capabilities of the present application will become apparent to those skilled in the art from the present disclosure.

Although the terms first, second, etc. may be used herein to describe various elements in some instances, these elements should not be limited by these terms. These terms are only used to distinguish one element from another. For example, a first fault diagnosis system may be referred to as a second fault diagnosis system, and similarly, a second fault diagnosis system may be referred to as a first fault diagnosis system, without departing from the scope of the various described embodiments. The first fault diagnosis system and the second fault diagnosis system are both describing one fault diagnosis system, but they are not the same fault diagnosis system unless the context clearly indicates otherwise.

Also, as used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, unless the context indicates otherwise. It will be further understood that the terms "comprises," "comprising," "includes" and/or "including," when used in this specification, specify the presence of stated features, steps, operations, elements, components, items, species, and/or groups, but do not preclude the presence, or addition of one or more other features, steps, operations, elements, components, species, and/or groups thereof. The terms "or" and/or "as used herein are to be construed as inclusive or meaning any one or any combination. Thus, "A, B or C" or "A, B and/or C" means "any of the following: a; b; c; a and B; a and C; b and C; A. b and C ". An exception to this definition will occur only when a combination of elements, functions, steps or operations are inherently mutually exclusive in some way.

As described in the background section, in industrial production, data such as vibration and temperature of a rotating machine during normal operation and abnormal operation may vary. Through fault feature analysis under different faults, a machine learning model can be constructed, and therefore fault diagnosis can be carried out by inputting the detected data of the self-rotating mechanical equipment into the model. However, in some embodiments, analyzing various faults that may occur in a rotating machine using a model with integrated analysis capabilities requires a large number of data samples. For example, in an experimental scenario, the sample size may be made sufficient by artificially destroying various components of the rotating machine and acquiring data to meet the requirement of requiring a large amount of sample input when constructing the machine learning model. However, the rotating machinery in actual industrial production is expensive, and obviously, the destructive method cannot be applied to actual industrial production, and if enough samples are lacked, the accuracy of the analysis result is affected. Therefore, although these machine models can be theoretically used to diagnose faults in rotating machinery, the accuracy of the diagnosis in actual production applications is greatly reduced in the absence of large data samples.

To this end, the present application provides a fault diagnosis method for a rotary mechanical apparatus, which is mainly performed by a fault diagnosis system. Wherein the fault diagnosis system can be executed by a server. The fault diagnosis system may be a software system configured at a server. The server includes, but is not limited to, a single server, a server cluster, a distributed server cluster, a cloud server, and the like. Here, according to the actual design, the server side where the fault diagnosis system is located may be configured in a server device located in a machine room on the rotating mechanical device side. For example, a single server or a server cluster in which the fault diagnosis system is located in a machine room on the rotating mechanical equipment side. Alternatively, the fault diagnosis system may be configured in a cloud service terminal provided by a cloud provider according to actual design. The Cloud Service end comprises a Public Cloud (Public Cloud) Service end and a Private Cloud (Private Cloud) Service end, wherein the Public or Private Cloud Service end comprises Software-as-a-Service (SaaS), Platform-as-a-Service (PaaS), Infrastructure as a Service (IaaS), and Infrastructure as a Service (IaaS). The private cloud service end is used for example for an Aliskian cloud computing service platform, an Amazon cloud computing service platform, a Baidu cloud computing platform, a Tencent cloud computing platform and the like.

In this case, the service end can be connected in communication with the control system of the rotating machine, depending on the actual design of the fault diagnosis system. The control system is a software system running on the computer device, collects running data detected by sensors arranged at detection points of the rotating mechanical device by means of the computer device, acquires device parameters of the rotating mechanical device, outputs a control instruction to the rotating mechanical device, and the like. For example, vibration sensors, temperature sensors, current sensors, etc. are distributed on the rotating machine, and the control system obtains data provided by any one or more of the above sensors and transmits the data to the fault diagnosis system via a communication network. Or, according to the actual design of the fault diagnosis system, the server may also be directly in communication connection with the rotating mechanical device and the sensors disposed at the detection points of the rotating mechanical device, so as to collect the operation data detected by the sensors disposed at the detection points of the rotating mechanical device, and obtain the device parameters of the rotating mechanical device.

It should be understood that, since the fault types of the rotating machinery are various, and the abnormal condition represented by each fault type is different, in some embodiments, although the diagnosis result can be obtained through one detection point, in order to ensure the correct diagnosis rate, a plurality of detection points can be arranged on the rotating machinery, and a more reliable diagnosis result is obtained through comprehensive analysis of the plurality of detection points. Wherein, the setting position of the detection point can be set according to actual requirements. Taking a fan as an example, please refer to fig. 2, which shows a schematic diagram of an embodiment of setting a detection point on the fan in the present application, as shown in the figure, a motor 102 drives a fan 100 to rotate through a transmission component, in this embodiment, a vibration sensor may be disposed in each of three directions (i.e., axial direction, vertical direction, and horizontal direction) of a fan bearing 101, so as to obtain mechanical vibration data in 6 directions. If the fan blade fails, the overall shaking of the equipment is increased, and the mechanical vibration data on the two bearings are abnormal; and if one bearing fails, the mechanical vibration data of the failed bearing is abnormal, and the mechanical vibration data of the other bearing is not obviously abnormal. For another example, when the rotating machine is a generator set, 1 temperature sensor may be provided on each bearing shell of each bearing of the generator set. Thus, the malfunction of the rotary mechanical apparatus can be diagnosed more accurately.

Here, the operation data acquired from the sensor includes at least one of the following according to the selected detection point and the sensor type thereof: mechanical vibration data, process data and operating condition data. Wherein the mechanical vibration data is sensed data provided by a sensor sensing mechanical vibration, examples of which include but are not limited to: and one or more of vibration speed, vibration impact, vibration displacement, vibration acceleration, vibration frequency spectrum and the like of the detection point. The process data is one or more data related to the production process in the rotating mechanical equipment; the production process is generally a work, a method and a technology for processing or treating various raw materials, semi-finished products and the like and making the raw materials, the semi-finished products and the like into finished products. The rotating mechanical device is a device in the process of performing a machining or treatment. The rotating mechanical equipment is used for providing specific indexes required to be reached by some manufacturing stages in the finished product manufacturing process. For example, a blower is used to provide combustion supporting wind power to the molten iron raw material to achieve a temperature index of the molten iron raw material. For another example, a gear machine is used for providing mechanical energy for driving a mechanical arm to grab molten steel. For this purpose, the process data are working data set for providing respective indexes in a manufacturing process reflecting the execution of the respective production process of the rotary machine, which are obtained by sensor detection provided at the detection point or the detection point itself, depending on the type of the rotary machine and its function in the production process. Wherein the detection point may be one or more, depending on the number and type of process data that the rotating machine itself can provide, and the number and type of sensors that are otherwise provided in the rotating machine with the rotation-related device. Taking a fan as an example, a plurality of detection points may be configured on the fan, such as a fan inlet, a fan outlet, a motor output power (or a driving voltage, a driving current, etc.), a fan blade, and the like. By way of example and generalization of the fan to other rotating machines, the process data acquired reflects, according to their role in the production process and by means of sensors built-in or external to the various detection points, at least one of the following provided by the rotating machine at the various detection points during the execution of the corresponding production process: temperature, current, rotational speed, flow, wind temperature, air door opening, inlet pressure, outlet pressure, etc. For example: when the rotating mechanical equipment is a fan and the detection point is positioned at a valve for controlling the flow to enter, the acquired process data can comprise the opening degree of an air door, the flow and the like; when the rotating mechanical device is a fan and the detection point is located on a drive bearing of the fan, the acquired process data may include temperature, etc. The working condition refers to the working state of the equipment under the condition that the equipment has direct relation with the action of the equipment. The rotating machinery generally has a plurality of operating modes, and the energy generated in each operating mode is different, so that the energy required by actual production requirements can be met by setting the operating mode of the rotating machinery. For example, the electric machine can provide different rotational speeds to the fan to meet the wind demand, in which process the required electrical energy is in turn met by the current. For this purpose, the operating condition data, which reflect the operating state of the rotating machine during operation, may be obtained by a sensor provided at the detection point or the detection point itself, depending on the type of the rotating machine and the energy that can be provided during operation. For example, the fault diagnosis system obtains the current vibration frequency of the motor through a sensor and analyzes the current rotating speed of the motor, so as to infer the current working mode of the motor from the rotating speed, and thus obtain the working condition data. The number and type of the detection points may be one or more, depending on the number and type of the operating condition data that the rotating machine itself can provide, and the number and type of the sensors that are additionally installed in the rotating machine. Taking the water pump as an example, the water pump may be configured with a plurality of detection points, such as a water outlet, a motor output power, a motor output shaft, and the like. By way of example and extension of the water pump to other rotating machines, the operating condition data may be obtained by the rotating machine from at least one of the following data provided at the various detection points during the supply of the energy required for production, depending on its role in production and by means of sensors built-in or externally on the various detection points: vibration, temperature, rotational speed, flow, water temperature, valve opening, inlet pressure, outlet pressure, and the like. For example: when the rotating mechanical equipment is a water pump and the detection point is positioned at a valve for controlling the water yield, the working condition data can be obtained through flow and the like; when the rotating mechanical equipment is a water pump and the detection point is positioned at the output shaft of the motor of the water pump, the working condition data can be obtained through the rotating speed; when the rotating mechanical equipment is a water pump and the detection point is positioned on the motor side for driving the water pump, the working condition data can be obtained through current and the like. In some embodiments, the operating condition data may be obtained by the rotating machinery itself or a management system of the rotating machinery. For example, the rotating machine device directly transmits the operation mode in which the rotating machine device is currently operating to the failure diagnosis system. For another example, the management system of the rotating machinery acquires the working mode in which the rotating machinery is currently working and sends the working mode to the fault diagnosis system.

Here, the present application describes an implementation of diagnosing a fault of a rotating machine by taking the fault diagnosis system as an example to diagnose the rotating machine.

Please refer to fig. 1, which is a schematic diagram illustrating a fault diagnosis method for a rotating machine according to the present application. As shown, in step S110, the fault diagnosis system acquires operation data at least one detection point in a rotating machine.

In some embodiments, the mechanical vibration data, process data, is directly acquired by a sensor. For example: various types of sensors are arranged on the detection points to acquire various types of mechanical vibration data and/or process data required. In other embodiments, some mechanical vibration data and/or process data may be calculated from the acquired correlation data. For example: the vibration displacement and the vibration speed can be calculated through the vibration acceleration acquired by the sensor, and a vibration displacement sensor and a vibration speed sensor do not need to be additionally arranged at the detection point.

Referring to fig. 1, in step S120, the fault diagnosis system performs an anomaly detection analysis on at least one anomaly type of the detection points by using the operation data, so as to obtain an analysis result corresponding to each anomaly type. It should be understood that, since the rotating mechanical equipment may have multiple faults at the same time in some cases, in order to ensure that the fault diagnosis system can diagnose multiple fault conditions of the rotating mechanical equipment, the obtained operation data is used to analyze the type of the abnormality corresponding to the corresponding detection point.

Wherein the anomaly type includes at least one of the following anomalies reflected by the respective detection points during operation of the rotating mechanical apparatus based on operating conditions and process requirements: an abnormality type set separately based on each vibration abnormality of at least one spatial dimension, an abnormality type set based on a temperature abnormality, an abnormality type set separately based on each power abnormality. Wherein the anomaly types that are set separately based on each vibration anomaly of at least one spatial dimension include, but are not limited to: abnormal unbalance of an impeller, abnormal misalignment of a coupling, abnormal vibration of a rolling bearing, abnormal vibration of a sliding bearing and the like; the abnormal type set based on the temperature abnormality includes but is not limited to bearing temperature variation abnormality and the like; the abnormality type set separately based on each power abnormality includes, but is not limited to, a motor current variation abnormality, and the like.

Here, the failure diagnosis system is preset with an abnormality detection model that is set according to each type of abnormality expression corresponding to a detection point at each failure, wherein the abnormality detection model includes an algorithm that determines a possibility of belonging or not belonging to a corresponding abnormality type based on input operation data. The fault diagnosis system inputs the received operation data of the detection points into the corresponding abnormity detection model to obtain an analysis result of whether the corresponding detection points show one type of abnormity. Wherein, different types of faults according to the rotating machinery are reflected on the abnormal performance of each detection point, and the abnormal types are exemplified and not limited as follows: the motor is characterized by comprising the following steps of abnormal impeller unbalance, abnormal misalignment of a coupling, abnormal vibration of a rolling bearing, abnormal vibration of a sliding bearing, abnormal temperature change of the bearing, abnormal current change of a motor and the like. For example, the fault diagnosis system performs an impeller imbalance abnormality detection analysis on the acquired operation data to obtain an analysis result of whether the detection point has an impeller imbalance abnormality.

Here, since the data required for each abnormality type when performing the abnormality detection analysis is different, the fault diagnosis system also generates the data required for each abnormality type when performing the abnormality detection analysis by associating the acquired operation data with each other. In order to diagnose the abnormality, reference data of the normal operation of the rotating mechanical equipment under the current working condition during the execution of the current production process needs to be determined.

In some embodiments, the reference data is static data that is pre-stored locally. The reference data comprise initial parameters and/or predetermined calibration parameters of the rotating machine. The initial parameters include inherent mechanical vibration data, rated process data, rated working condition data and the like of the rotary mechanical equipment in the factory or after maintenance. The preset calibration parameters are data obtained by calibrating all or part of mechanical vibration data, process data and the like of the rotary mechanical equipment by a manager of the rotary mechanical equipment according to experience. For example: in some scenarios, a problem of a geographical environment (such as insufficient installation base strength) may cause abnormal vibration of the device, which results in that mechanical vibration data of the rotating mechanical device when the rotating mechanical device operates in the environment is greatly different from mechanical vibration data when the rotating mechanical device leaves a factory, but since the abnormal mechanical vibration data is not caused by a fault of the rotating mechanical device, a manager of the rotating mechanical device may calibrate the mechanical vibration data of the rotating mechanical device according to experience, so as to use the calibrated mechanical vibration data as reference data.

In other embodiments, the operating data is used to generate the detection data and the reference data, i.e. the reference data is dynamic data. In this case, the fault diagnosis system analyzes the acquired operating data for reference data during normal operation of the rotating machine. For example, the fault diagnosis system analyzes mechanical vibration data in the acquired operation data to obtain a fundamental frequency of the rotating mechanical device, and uses the fundamental frequency as reference data, and simultaneously converts the acquired mechanical vibration data into a vibration spectrum, and performs abnormality detection analysis on the mechanical vibration data by using the fundamental frequency. In this embodiment, the operation data acquired by the fault diagnosis system is real-time data provided by the detection point, which directly or indirectly provides detection data and at least part of reference data for anomaly detection and analysis. Wherein the reference data is used to represent normal data for normal operation of the rotating mechanical device under the current operating conditions during execution of the current production process. The sensed data is indicative of current data of the rotating machine at a current operating condition during execution of the current production process. And the fault diagnosis system performs anomaly detection analysis on the detection data based on the reference data and obtains a corresponding analysis result. In some specific examples, depending on the type of anomaly, some of the acquired operational data may be used directly as reference data, such as torque of the drive motor, etc.

It should be understood that the fault diagnosis system may acquire the reference data once each time the fault diagnosis is performed; the reference data acquired once may be stored in the storage medium, and the reference data in the storage medium may be called each time the failure diagnosis needs to be performed.

It should be understood that the reference data and the detected data should be under the same operating condition data, i.e., the reference data and the detected data are both data when the rotating mechanical device operates in the same operating mode. Thus, the reference data comprises at least one or more of: the working condition data of the rotating mechanical equipment is data extracted from process data in the acquired operation data, and the data extracted from mechanical vibration data in the acquired operation data.

In some embodiments, the reference data of the detection points may be obtained dynamically in the above embodiments and/or statically in normal operation of the rotating machine. For example, in some embodiments, the reference data of the detection point is obtained in a dynamic manner in the above embodiments; for another example, in some embodiments, the reference data of the detection point is obtained in a static manner in the above embodiments; for another example, in some embodiments, a part of the reference data of the detection point is obtained in a dynamic manner in the above embodiments, another part of the reference data of the detection point is obtained in a static manner in the above embodiments, and the like.

In an exemplary embodiment, when the acquired operational data comprises process data and/or operating condition data; the step S1201 further includes the steps of: and analyzing the acquired process data and/or working condition data to obtain benchmark data in the reference data of the detection point during normal operation of the rotary mechanical equipment.

In some specific examples, the fault diagnosis system calculates corresponding reference data by using process data meeting normal operating conditions according to the abnormal type corresponding to the detection point. For example, the fault diagnosis system detects abnormal air outlet at an air outlet of a fan, and uses the temperature, flow, pressure, density and the like of inlet air to calculate reference data of the fan energy efficiency of the fan under the current working condition. In some specific examples, the fault diagnosis system obtains the reference data by using the operating condition data obtained from the detection point or the locally pre-stored operating condition data according to the abnormal type corresponding to the detection point. For example, in order to detect the bearing rotation abnormality, the fault diagnosis system determines that the reference data includes rotation speed data according to rotation speed data corresponding to the same working condition mode, which is continuously acquired for multiple times. In some specific examples, the fault diagnosis system obtains the reference data by using the operating condition data obtained from the detection points or the locally pre-stored operating condition data and the process data according to the abnormal type corresponding to the detection points. For example, the fault diagnosis system detects an abnormality of an air outlet of a fan, and uses a working condition mode of the air outlet and a motor current as reference data.

Here, when the detection data includes process data, operating condition data, or both process data and operating condition data, the fault diagnosis system analyzes the process data, the operating condition data, or both the process data and the operating condition data, thereby obtaining reference data used as a standard in reference data of the detection point when the rotary mechanical device is operating normally, so as to determine whether the detection data is abnormal using the reference data.

The reference data can be obtained in a static manner, in addition to the dynamic manner. The reference data comprise initial parameters and/or predetermined calibration parameters of the rotating machine. The initial parameter includes, for example, factory process data of the rotating mechanical device. The preset calibration parameters are data obtained by calibrating all or part of process data and the like of the rotating mechanical equipment by a manager of the rotating mechanical equipment according to experience, for example: in some scenarios, due to a problem of a geographical environment, such as a high temperature of a production environment, a temperature abnormality of the device may be caused, so that a difference between temperature data of the rotating mechanical device when the rotating mechanical device operates in the environment and temperature data when the rotating mechanical device leaves a factory is large, but since the abnormality of the temperature data is not caused by a fault of the rotating mechanical device, a manager of the rotating mechanical device may calibrate the temperature data of the rotating mechanical device according to experience, so as to use the calibrated temperature data as reference data. It should be understood that the fault diagnosis system may acquire the reference data once each time the fault diagnosis is performed; the reference data acquired once may be stored in the storage medium, and the reference data in the storage medium may be called each time the failure diagnosis needs to be executed.

In an exemplary embodiment, referring to fig. 5, which shows a schematic diagram of an embodiment of performing anomaly detection analysis by using operation data in the present application, as shown in step S1201, the fault diagnosis system further performs feature extraction on the acquired detection data based on reference data of the detection points during normal operation of the rotating mechanical equipment to obtain at least one feature element for detecting at least one anomaly type of the detection points.

In some embodiments, after the fault diagnosis system acquires the operation data, there may be a situation where the operation data acquired from the sensor or the like is disturbed or a measurement error occurs. In contrast, after preprocessing the acquired operation data, the fault diagnosis system generates detection data (or detection data and reference data) from the operation data, and performs feature extraction on the detection data. The preprocessing method includes but is not limited to noise reduction, outlier rejection, and the like.

In some embodiments, the fault diagnosis system performs feature extraction on the detection data according to the reference data, thereby obtaining at least one feature element for detecting at least one anomaly type of the detection point. The feature extraction method includes, but is not limited to, Mean value calculation, RMS (Root Mean Square), FFT (Fast Fourier transform), envelope spectrum extraction, etc., and the feature elements obtained by feature extraction are used as input for analyzing the type of the anomaly.

In some embodiments, after the fault diagnosis system performs feature extraction on the detection data according to the reference data, the fault diagnosis system further performs feature engineering on the result after the feature extraction, so as to process the result after the feature extraction into feature elements required by each abnormal type. The feature engineering includes but is not limited to feature combination, feature dimension reduction, feature processing, feature normalization, and the like.

In some embodiments, to facilitate the analysis of the abnormality, the fault diagnosis system further includes performing further data processing on the obtained at least one feature element, so as to analyze the at least one feature element after the further data processing, thereby obtaining an analysis result of the corresponding abnormality type. Here, the data processing includes, but is not limited to, mathematical operations and the like. The feature elements processed by the data can also be used for inputting into an anomaly detection model for analysis, so that accurate analysis is realized under the condition of insufficient data sample amount. For example, the component at the frequency 2 times the fundamental frequency, the component at the frequency 3 times the fundamental frequency, the component at the frequency 4 times the fundamental frequency, the component at the frequency 5 times the fundamental frequency, etc. are combined in increments of several higher order multiples to form a feature element.

Because different types of abnormality need to use different types and different descriptions of feature elements as judgment, each obtained feature element can correspond to one or more abnormality types. In other words, the same feature element may be reused in different exception types. The set characteristic extraction mode can be determined according to the experience of a manager of the rotary mechanical equipment, so that the criticality of the extracted characteristics is ensured, and the accuracy and the efficiency of an analysis result are ensured.

In some embodiments, the fault diagnosis system further includes a feature mechanism model, the input of the feature mechanism model is detection data and reference data, and the output of the feature mechanism model is feature elements corresponding to each abnormality type. In this case, the feature mechanism model performs feature extraction on the detection data by referring to the data. The feature mechanism model determines features to be extracted according to a preset rule, and performs feature extraction on detection data by using reference data to generate feature values by combining algorithms such as average value calculation, effective value calculation, frequency spectrum extraction and envelope spectrum extraction. Wherein the reference data may be obtained statically and/or obtained dynamically. Meanwhile, the characteristic mechanism model further processes the characteristic value into each characteristic element corresponding to each abnormal type through data processing, so that each characteristic element is analyzed to obtain an analysis result of the corresponding abnormal type. The characteristic mechanism model can be constructed through the mechanism model by utilizing pre-marked historical operating data, reference data and the like. For example, historical operation data of the rotating machine is acquired from a manager, a control system, or the like of the rotating machine, and initial parameters of the rotating machine are acquired from a manager, a control system, a network, or the like of the rotating machine. It should be understood that the mechanism model, also known as the white box model. Which is an accurate mathematical model built from objects, internal mechanisms of the production process, or transport mechanisms of the material flow. The algorithm in the feature mechanism model includes but is not limited to feature value calculation, feature engineering, and the like, and the feature value calculation includes but is not limited to: taking mechanical vibration data as an example, the characteristic value calculation comprises processing the acquired historical mechanical vibration data into a frequency domain spectrum, acquiring a fundamental frequency from the mechanical vibration data and the like to obtain at least one characteristic value. The feature engineering includes, but is not limited to, subjecting the at least one feature value to feature combination dimension reduction, feature processing, feature normalization, and the like to obtain at least one feature element. And if the accuracy of the output result of the trained feature mechanism model reaches a preset accuracy threshold, finishing the training.

Here, a process of calculating the feature element when the acquired operation data includes mechanical vibration data or the acquired detection data includes process data and/or condition data is exemplified, respectively.

Wherein, when the acquired operation data includes mechanical vibration data, the step S1201 includes: extracting at least one frequency feature element in the acquired frequency spectrum of the mechanical vibration data, which is related to the fundamental frequency data in the reference data of the detection point, for detecting at least one anomaly type of the detection point. The fundamental frequency data corresponds to a frequency or a frequency interval with low frequency and high intensity in a vibration spectrum generated when the rotary mechanical equipment normally operates under the current working condition during the execution of the current production process. According to the actual working conditions during the production process executed by the rotary mechanical equipment, the fundamental frequency data of different production processes and different working conditions are not completely consistent. To this end, in some specific examples, the fundamental frequency data is extracted from operational data. For example, the mechanical vibration data of a certain dimension of the blade is subjected to frequency domain conversion, and the fundamental frequency data is obtained through the analysis of the frequency spectrum distribution. In another specific example, the fundamental frequency data is selected from a plurality of corresponding relations of the locally stored process, the operating condition and the fundamental frequency data according to the acquired operating condition data, the process data and the like.

It is understood that rotary mechanical devices (e.g., fans, motors, water pumps, gear boxes, etc.) in industrial operations are rotated and perform work under the drive of motors to output energy. Therefore, the motor, the rotating part and the like have the rotating speeds of the motor and the rotating part, the rotating speeds can cause equipment vibration to a certain extent, and the fundamental frequency of the vibration can be calculated through the rotating speeds of the equipment. Meanwhile, the vibration frequency spectrum caused by the rotating speed is based on the fundamental frequency, and frequency spectrum components such as integral multiple, fractional multiple, characteristic multiple, high-frequency modulation and the like of the fundamental frequency are additionally superposed. These spectral components tend to be small during normal operation of the device (less vibrational energy during normal operation), while different faults may exhibit different vibrational spectra during a failure of the device.

It will be appreciated that in some embodiments, the fundamental frequency may be calculated from the rotational speed, for example: n/60, wherein: f is the frequency (unit: HZ) and n is the rotational speed (unit: rpm). In other embodiments, the fundamental frequency can also be analyzed from the vibration spectrum by means of spectral analysis.

The frequency feature elements include, but are not limited to, a multiplying power spectral component that is a fundamental frequency, and the like. For example: the frequency domain conversion operation is carried out on the obtained mechanical vibration data, and the frequency spectrum distribution is analyzed, so that frequency characteristic elements such as components on 2-frequency multiplication and components on 4-frequency multiplication of the fundamental frequency are obtained. In some embodiments, the frequency feature element may also be a spectral component of other specific frequencies, such as a component at the frequency 3 multiplied by 0.5.

When the acquired detection data includes process data and/or operating condition data, the step S1201 further includes: obtaining at least one deviation feature element based on a deviation of the acquired detection data from the reference data in normal operation of the rotating mechanical device, for detecting at least one anomaly type of the detection point.

In this case, the acquired test data are compared with reference data, so that at least one deviation characteristic element is obtained, which includes, but is not limited to: increments, percentages of increments, etc. For example: the acquired detection data is temperature data, and the reference temperature data in the reference data and the temperature data in the detection data are compared when the rotary mechanical equipment normally operates, so that the difference value of the temperature change and/or the percentage of the difference value to the reference temperature data are extracted. And respectively using the difference and/or the percentage of the difference relative to the reference temperature data as a deviation characteristic element so as to detect at least one abnormal type of the detection point.

After obtaining the at least one feature element through the various manners of the above-described embodiments, the failure diagnosis system provides the at least one feature element to step S1202.

Referring to fig. 5, in step S1202, the feature elements corresponding to each anomaly type are analyzed to obtain an analysis result of the corresponding anomaly type.

The method for analyzing the feature elements corresponding to each anomaly type includes, but is not limited to, analyzing by using an anomaly detection model, where the anomaly detection model is a machine learning model, and the machine learning model includes, but is not limited to, a KNN (k-Nearest Neighbor, k-Neighbor algorithm) -based machine learning model, and the like. In some embodiments, each anomaly type has a separate machine learning model, such as: the method comprises the following steps that an impeller unbalance model, a coupling misalignment model, a rolling bearing vibration abnormity and a sliding bearing vibration abnormity are respectively corresponding to the impeller unbalance model, the coupling misalignment model, the rolling bearing fault model and the sliding bearing fault model; the bearing temperature abnormity and the motor current abnormity correspond to a bearing temperature abnormity model, a motor current abnormity model and the like. Since the data required in analyzing each anomaly type is different, the failure diagnosis system converts the detection data into the inputs, i.e., feature elements, required by the models to obtain the corresponding analysis results. For example: and the coupler misalignment model needs a component on 2 frequency multiplication of the fundamental frequency, a component on 3 frequency multiplication of the fundamental frequency and a component on 4 frequency multiplication of the fundamental frequency as input, and the fault diagnosis system performs feature extraction on the detection data, and then enters the coupler misalignment model for analysis by taking two feature elements of the component on 2 frequency multiplication of the fundamental frequency, the component on 3 frequency multiplication of the fundamental frequency, the component on 4 frequency multiplication of the fundamental frequency and the fundamental frequency as input, so as to obtain an analysis result of the coupler misalignment model.

At least the machine learning algorithm in the anomaly detection model can be trained through pre-marked historical operating data and the like to obtain parameters in the algorithm. For example, historical operation data of a rotating machine and historical faults corresponding to the historical operation data are acquired from a manager, a control system, or the like of the rotating machine. And processing the acquired data into sample data required by a corresponding algorithm and training the algorithm to obtain an anomaly detection model. Wherein the process is used to convert the acquired data into data that can be processed by an algorithm, including but not limited to: normalization processing, data conversion according to a preset conversion formula and the like. And if the accuracy of the trained anomaly detection model reaches a preset accuracy threshold, finishing the training. The input of the abnormality detection model is a feature element required by the model, and the fault diagnosis system determines an abnormality type analysis result corresponding to each abnormality of the rotary mechanical equipment by using the abnormality detection model.

Here, the fault diagnosis system performs step S130 after obtaining the analysis result corresponding to each abnormality type at the at least one detection point.

With continued reference to fig. 1, in step S130, the fault diagnosis system performs a diagnosis process on at least one fault in the rotating mechanical equipment by using the obtained at least one analysis result to output a corresponding fault diagnosis result.

Here, the fault diagnosis system obtains at least one analysis result according to the actual conditions such as the number of the operation data provided by the detection points, the network transmission efficiency and the like, and performs fault diagnosis processing according to the at least one analysis result. For example, the fault diagnosis system performs diagnosis of the imbalance fault of the blade or the non-alignment fault of the coupling according to the analysis result about the abnormal type of the vibration of the blade at the corresponding blade detection point obtained in step S120, or performs diagnosis of the imbalance fault of the blade and the non-alignment fault of the coupling respectively, and obtains a fault diagnosis result through a diagnosis and evaluation system.

It should be understood that when different parts of the rotating machinery fail, there are different anomalies. Taking a fan as an example, when a bearing of the fan breaks down, the vibration of the bearing part is strong and the temperature is increased, but the current abnormality of a fan motor is not caused; for another example, when a blade of a fan fails, the temperature of the bearing portion does not rise, but the fan as a whole is shaken. Therefore, if only the abnormality type analysis result of a single detection point is taken as the failure diagnosis result, the result is low in accuracy, whereas if the abnormality type analysis results of a plurality of detection points are comprehensively diagnosed, a diagnosis result with higher accuracy is obtained.

Here, the fault diagnosis system further includes a comprehensive diagnosis model, so that the fault diagnosis result of the rotary machine is obtained by comprehensively diagnosing and processing the abnormality type analysis result of each detection point on the rotary machine and by comprehensively processing the position and type of each detection point. The comprehensive diagnosis model is a mechanism model, the input of the comprehensive diagnosis model is the analysis result corresponding to each abnormal type on at least one detection point, and the output of the comprehensive diagnosis model is the fault diagnosis result of the rotary mechanical equipment.

In an exemplary embodiment, the integrated diagnostic model includes a plurality of independent fault models, such as: the system comprises an impeller unbalance fault model, a coupling misalignment fault model, a fan side bearing fault model, a motor side bearing fault model and the like, wherein the fault diagnosis system correspondingly inputs the abnormal type analysis results of all detection points into the fault model. Where the inputs required for each fault type are different and the same input may also be used for different fault types. For example: the input of the unbalanced impeller fault model corresponds to the analysis result of the unbalanced type of the impeller on the detection point, the input of the misaligned coupling fault model corresponds to the analysis result of the misaligned coupling fault type on the detection point and the analysis result of the abnormal type of the current on the detection point, the input of the fan side bearing fault model corresponds to the analysis result of the abnormal type of the rolling bearing and/or the abnormal type of the sliding bearing on the detection point, and the input of the motor side bearing fault model corresponds to the analysis result of the abnormal type of the rolling bearing and/or the abnormal type of the sliding bearing on the detection point, the analysis result of the abnormal type of the temperature on the detection point and the like.

The fault diagnosis system presets diagnosis rules of a plurality of fault models in a comprehensive diagnosis model so as to carry out fault diagnosis on the rotating mechanical equipment through the analysis result of each abnormal type on at least one detection point. For example: taking the fan as an example, the fault diagnosis system presets that when the analysis result of the rolling bearing abnormal type and the analysis result of the temperature abnormal type at the detection point are both abnormal, and the analysis results of other abnormal types at the detection point are both normal, the comprehensive diagnosis model outputs the fault diagnosis result as the fault diagnosis result of the motor side bearing. It should be appreciated that, since multiple faults may exist in the rotating machine at the same time, in some embodiments, the output of the comprehensive diagnostic model has multiple fault diagnosis results. In still other embodiments, when the fault diagnosis system fails to diagnose a fault in the rotating machine, a diagnosis of an unknown fault is output.

In an exemplary embodiment, the fault diagnosis system displays the fault diagnosis result in order to facilitate management or operation by a manager of the rotary machine. For this purpose, the step S130 further includes: presenting the obtained fault diagnosis result in a display interface of a control system of the rotary mechanical equipment. In some embodiments, a display may be connected to the computer device where the control system is located, and a manager of the rotating mechanical device may view the fault diagnosis result through the display. In still other embodiments, after obtaining the fault diagnosis result, the manager of the rotating mechanical equipment overhauls or checks the rotating mechanical equipment according to the fault diagnosis result, and feeds back a conclusion whether the fault diagnosis result is correct to the fault diagnosis system through the control system.

For convenience of understanding, the process of fault diagnosis by the fault diagnosis system for rotating mechanical equipment will be illustrated by using 5 detection points in conjunction with fig. 3a to 6, it should be understood that this embodiment is only used for explaining the present application, and is not used for limiting the present application, and the detection points configured on the rotating mechanical equipment may further include a sixth detection point, a seventh detection point, an eighth detection point, and the like according to actual requirements.

In this embodiment, the fault diagnosis system first obtains operation data of a sensor at a first detection point, where the sensor is a vibration sensor, and the vibration sensor provides mechanical vibration data to the fault diagnosis system. After the fault diagnosis system preprocesses the mechanical vibration data, the preprocessed mechanical vibration data are input into the characteristic mechanism model. The characteristic mechanism model firstly generates vibration mechanical data into detection data. Meanwhile, the fault diagnosis system acquires reference data corresponding to the detection data, namely mechanical vibration data of the rotating mechanical equipment during normal operation, and inputs the reference data to the characteristic mechanism model. The characteristic mechanism model records the detection data and the reference data into time domain waveforms, please refer to fig. 3a to fig. 3b, which show schematic diagrams of an embodiment of the fault diagnosis system for performing anomaly detection analysis in the present application, wherein fig. 3b shows time domain waveforms of the detection data at detection points in the rotating machinery equipment, that is, waveforms (unit: mm/s) of real-time mechanical vibration speed data of the detection points; fig. 3a shows a time domain waveform of reference data corresponding to the detection data, that is, a time domain waveform of mechanical vibration data of the detection point when the rotating mechanical device is in normal operation. It can be seen that it is difficult to analyze the reference data and the detection data with reference to fig. 3a and 3b, so the feature mechanism model further converts the time domain waveforms of the reference data and the detection data into frequency domain spectrums.

Referring to fig. 4a to 4b, which are schematic diagrams illustrating a time domain waveform of fig. 3a to 3b is converted into a frequency domain spectrum according to the present application, as shown in the figure, the anomaly in the detection data can be clearly seen from the frequency domain spectrum, and the feature mechanism model processes feature values corresponding to the anomaly to obtain a plurality of first feature elements. The processing method comprises the step of converting the characteristic values into frequency spectrum components of specific frequency multiplying power so as to input the frequency spectrum components into various anomaly detection models for analysis. After obtaining the plurality of first feature elements, some of the first feature element data is further processed, so as to reduce the dimension and form a plurality of second feature elements.

The characteristic mechanism model provides the first characteristic element and/or the second characteristic element to a corresponding abnormality detection model, wherein the abnormality detection model comprises an impeller imbalance model, a coupling misalignment model, a rolling bearing fault model and a sliding bearing fault model, each abnormality detection model respectively outputs an analysis result, namely whether the possibility of a fault corresponding to the model exists, for example, the possibility of the impeller imbalance model outputting the impeller imbalance abnormality is 30%, the possibility of the coupling misalignment model outputting the coupling misalignment abnormality is 80%, and the like. Here, the failure diagnosis system obtains an analysis result output by each abnormality detection model at the first detection point.

The second detection point and the third detection point are both vibration sensors, and the way of obtaining the analysis result is the same as that of the first detection point, so that the repeated description is omitted.

Secondly, the fault diagnosis system also obtains the operation data of a sensor on a fourth detection point, wherein the sensor is a temperature sensor, and the vibration sensor provides the temperature data for the fault diagnosis system. After the fault diagnosis system preprocesses the temperature data, the preprocessed temperature data are input into the characteristic mechanism model. The characteristic mechanism model firstly generates detection data from temperature data. Meanwhile, the fault diagnosis system acquires reference data corresponding to the detection data, namely temperature data of the rotating mechanical equipment during normal operation, and inputs the reference data to the characteristic mechanism model. The feature mechanism model compares the detection data with the reference data to calculate the increment percentage between the detection data and the reference data, and uses the increment percentage as a feature element, for example, the detection data is 30% higher than the reference data, and the like. The feature mechanism model provides the feature elements to a corresponding abnormality detection model, where the abnormality detection model includes a temperature abnormality model that outputs an analysis result, that is, whether there is a possibility of a fault corresponding to the model, for example, the temperature abnormality model outputs a possibility of a temperature abnormality of 10%. Here, the failure diagnosis system obtains an analysis result output by the abnormality detection model at the fourth detection point.

Meanwhile, the fault diagnosis system also acquires operation data of a sensor on a fifth detection point, wherein the sensor is a current sensor, and the vibration sensor provides the current data to the fault diagnosis system. After the fault diagnosis system preprocesses the current data, the preprocessed current data are input into the characteristic mechanism model. The characteristic mechanism model firstly generates current data into detection data. Meanwhile, the fault diagnosis system acquires reference data corresponding to the detection data, namely current data of the rotating mechanical equipment in normal operation, and inputs the reference data to the characteristic mechanism model. The feature mechanism model compares the detection data with the reference data to calculate the increment percentage between the detection data and the reference data, and uses the increment percentage as a feature element, for example, the detection data is 30% higher than the reference data, and the like. The feature mechanism model provides the feature elements to a corresponding abnormality detection model, where the abnormality detection model includes a current variation model, and the current variation model outputs an analysis result, that is, whether there is a possibility of a fault corresponding to the model, for example, the current variation model has a possibility of outputting a current abnormality of 10%. Here, the failure diagnosis system obtains an analysis result output by the abnormality detection model at the fifth detection point.

Referring to fig. 6, which is a schematic diagram of an embodiment of the fault diagnosis process performed in the present application, an analysis result output by each anomaly detection model at each of 5 detection points is provided to a comprehensive diagnosis model, so as to comprehensively diagnose an anomaly type analysis result at each detection point, and obtain a fault diagnosis result of the rotary machine in combination with position and type comprehensive processing of each detection point.

According to the fault diagnosis scheme provided by the application, a complex single machine learning model is split into a single anomaly detection model, so that the model complexity is reduced, and a high-precision fault detection effect can be obtained by adopting less sample learning; in addition, the scheme provided by the application does not depend on the collected comprehensive abnormal types, and the corresponding fault diagnosis result can be given according to the analysis result of the abnormal types which can be collected actually, so that the coordination flexibility among the machine learning models is effectively improved.

In an embodiment of the second aspect of the present application, a server is provided, please refer to fig. 7, which is a schematic structural diagram of a server. As shown, the server includes an interface unit 11, a storage unit 12, and a processing unit 13. The storage unit 12 includes a nonvolatile memory, a storage server, and the like. The nonvolatile memory is, for example, a solid state disk or a usb disk. The storage server is used for storing various acquired operation data, reference data and the like. The interface unit 11 includes a network interface, a data line interface, and the like. Wherein the network interfaces include, but are not limited to: network interface devices based on ethernet, network interface devices based on mobile networks (3G, 4G, 5G, etc.), network interface devices based on near field communication (WiFi, bluetooth, etc.), and the like. The data line interface includes, but is not limited to: USB interface, RS232, etc. The interface unit is connected with data such as sensors, a fault diagnosis system, the Internet and the like arranged at detection points on the rotary mechanical equipment. The processing unit 13 is connected to the interface unit 11 and the storage unit 12, and includes: a CPU or a chip integrated with a CPU, a programmable logic device (FPGA), and a multi-core processor. The processing unit 13 also includes memories, registers, etc. for temporarily storing data.

The interface unit 11 is used for data communication with a sensor in at least one dimension at a detection point of a rotating mechanical device. Here, the interface unit 11 is, for example, a network card, and may be communicatively connected to the computer device through the internet or a built-up dedicated network.

The storage unit 12 is used to store at least one program. Here, the storage unit 12 includes, for example, a hard disk provided at a server side and stores the at least one program, and in addition, various information acquired to the interface unit 11 is stored in the storage unit 12 according to external data required to be acquired during the program operation. Wherein the various information includes the aforementioned reference data of the rotating machine device, and the like.

The processing unit 13 is configured to invoke the at least one program to coordinate the interface unit and the storage unit to execute the fault diagnosis method of the rotary machine according to any one of the above examples. The method for diagnosing faults of rotating mechanical equipment is shown in fig. 1 and corresponding description, and is not repeated herein.

In an embodiment of the third aspect of the present application, a first fault diagnosis system for a rotary machine is provided, where the first fault diagnosis system includes a service end as described in the embodiment of the second aspect of the present application, and a detection device disposed at each detection point of the rotary machine. The detection device of each detection point of the rotating mechanical equipment is in communication connection with the server, so that the operation data of each detection point is provided for the server, and the server can diagnose the fault of the rotating mechanical equipment through the provided operation data of each detection point.

It should be noted that, through the above description of the embodiments, those skilled in the art can clearly understand that part or all of the present application can be implemented by software and combined with necessary general hardware platform. With this understanding in mind, an embodiment of a fourth aspect of the present application provides a computer-readable storage medium storing at least one program that, when invoked, performs and implements any of the aforementioned methods of rotary machine fault diagnosis.

At the same time, it will be appreciated that aspects of the present disclosure, in essence, or as part of the prior art, may be embodied in a software product that may include one or more machine-readable media having stored thereon machine-executable instructions that, when executed by one or more machines such as a computer, network of computers, or other electronic devices, may cause the one or more machines to perform operations in accordance with embodiments of the present disclosure. For example, each step in the rotating machine failure diagnosis method is performed. The machine-readable medium may include, but is not limited to, floppy diskettes, optical disks, CD-ROMs (compact disc-read only memories), magneto-optical disks, ROMs (read only memories), RAMs (random access memories), EPROMs (erasable programmable read only memories), EEPROMs (electrically erasable programmable read only memories), magnetic or optical cards, flash memory, or other type of media/machine-readable medium suitable for storing machine-executable instructions. The storage medium may be located in a server or in a third-party server, such as a server in a mall providing some application. The specific application mall is not limited, such as the millet application mall, the Huawei application mall, and the apple application mall.

The application is operational with numerous general purpose or special purpose computing system environments or configurations. For example: personal computers, server computers, hand-held or portable devices, tablet-type devices, multiprocessor systems, microprocessor-based systems, set top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments that include any of the above systems or devices, and the like.

The application may be described in the general context of computer-executable instructions, such as program modules, being executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform particular tasks or implement particular abstract data types. The application may also be practiced in distributed computing environments where tasks are performed by remote processing devices that are linked through a communications network. In a distributed computing environment, program modules may be located in both local and remote computer storage media including memory storage devices.

In an embodiment of the fifth aspect of the present application, a second fault diagnosis system is provided.

In an exemplary embodiment, first, the second fault diagnosis system acquires operation data on at least one detection point in a rotary mechanical device.

In some embodiments, the mechanical vibration data, process data, is directly acquired by a sensor. For example: various types of sensors are arranged on the detection points to acquire various types of mechanical vibration data and/or process data required. In other embodiments, some mechanical vibration data and/or process data may be calculated from the acquired correlation data. For example: the vibration displacement and the vibration speed can be calculated through the vibration acceleration acquired by the sensor, and a vibration displacement sensor and a vibration speed sensor do not need to be additionally arranged at the detection point.

And the second fault diagnosis system respectively performs anomaly detection analysis on at least one anomaly type of the detection points by using the operation data to obtain an analysis result corresponding to each anomaly type. It should be understood that, since the rotating mechanical equipment may have multiple faults at the same time in some cases, in order to ensure that the second fault diagnosis system can diagnose multiple fault conditions of the rotating mechanical equipment, the obtained operation data is used to analyze the type of the abnormality corresponding to the corresponding detection point.

Wherein the anomaly type includes at least one of the following anomalies reflected by the respective detection points during operation of the rotating mechanical apparatus based on operating conditions and process requirements: an abnormality type set separately based on each vibration abnormality of at least one spatial dimension, an abnormality type set based on a temperature abnormality, an abnormality type set separately based on each power abnormality. Wherein the anomaly types that are set separately based on each vibration anomaly of at least one spatial dimension include, but are not limited to: abnormal unbalance of an impeller, abnormal misalignment of a coupling, abnormal vibration of a rolling bearing, abnormal vibration of a sliding bearing and the like; the abnormal type set based on the temperature abnormality includes but is not limited to bearing temperature variation abnormality and the like; the abnormality type set separately based on each power abnormality includes, but is not limited to, a motor current variation abnormality, and the like.

Here, the second failure diagnosis system is preset with an abnormality detection model that is set according to each type of abnormality expression corresponding to the detected point at the time of each failure, wherein the abnormality detection model includes an algorithm that determines the possibility of belonging or not belonging to the corresponding abnormality type based on the input operation data. And the second fault diagnosis system inputs the received operation data of the detection points into the corresponding abnormity detection model to obtain an analysis result of whether the corresponding detection points show one type of abnormity. Wherein, different types of faults according to the rotating machinery are reflected on the abnormal performance of each detection point, and the abnormal types are exemplified and not limited as follows: the motor is characterized by comprising the following steps of abnormal impeller unbalance, abnormal misalignment of a coupling, abnormal vibration of a rolling bearing, abnormal vibration of a sliding bearing, abnormal temperature change of the bearing, abnormal current change of a motor and the like. For example, the second fault diagnosis system performs an impeller imbalance abnormality detection analysis on the acquired operation data to obtain an analysis result of whether the detection point has an impeller imbalance abnormality.

Here, since the data required for each abnormality type when performing the abnormality detection analysis is different, the second failure diagnosis system further generates the data required for each abnormality type when performing the abnormality detection analysis by associating the acquired operation data with each other. In order to diagnose the abnormality, reference data of the normal operation of the rotating mechanical equipment under the current working condition during the execution of the current production process needs to be determined.

In some embodiments, the reference data is static data that is pre-stored locally. The reference data comprise initial parameters and/or predetermined calibration parameters of the rotating machine. The initial parameters include inherent mechanical vibration data, rated process data, rated working condition data and the like of the rotary mechanical equipment in the factory or after maintenance. The preset calibration parameters are data obtained by calibrating all or part of mechanical vibration data, process data and the like of the rotary mechanical equipment by a manager of the rotary mechanical equipment according to experience. For example: in some scenarios, a problem of a geographical environment (such as insufficient installation base strength) may cause abnormal vibration of the device, which results in that mechanical vibration data of the rotating mechanical device when the rotating mechanical device operates in the environment is greatly different from mechanical vibration data when the rotating mechanical device leaves a factory, but since the abnormal mechanical vibration data is not caused by a fault of the rotating mechanical device, a manager of the rotating mechanical device may calibrate the mechanical vibration data of the rotating mechanical device according to experience, so as to use the calibrated mechanical vibration data as reference data.

In other embodiments, the operating data is used to generate the detection data and the reference data, i.e. the reference data is dynamic data. In this case, the second fault diagnosis system also analyzes, from the acquired operating data, corresponding reference data during normal operation of the rotating machine. For example, the second fault diagnosis system analyzes mechanical vibration data in the acquired operation data to obtain a fundamental frequency of the rotating mechanical device, and uses the fundamental frequency as reference data, and simultaneously converts the acquired mechanical vibration data into a vibration spectrum, and performs abnormality detection analysis on the mechanical vibration data by using the fundamental frequency. In this embodiment, the operation data acquired by the second fault diagnosis system is real-time data provided by the detection point, which directly or indirectly provides detection data and at least part of reference data for anomaly detection and analysis. Wherein the reference data is used to represent normal data for normal operation of the rotating mechanical device under the current operating conditions during execution of the current production process. The sensed data is indicative of current data of the rotating machine at a current operating condition during execution of the current production process. And the second fault diagnosis system carries out anomaly detection analysis on the detection data based on the reference data and obtains a corresponding analysis result. In some specific examples, depending on the type of anomaly, some of the acquired operational data may be used directly as reference data, such as torque of the drive motor, etc.

It should be understood that the second failure diagnosis system may acquire the reference data once each time the failure diagnosis is performed; the reference data acquired once may be stored in the storage medium, and the reference data in the storage medium may be called each time the failure diagnosis needs to be performed.

It should be understood that the reference data and the detected data should be under the same operating condition data, i.e., the reference data and the detected data are both data when the rotating mechanical device operates in the same operating mode. Thus, the reference data comprises at least one or more of: the working condition data of the rotating mechanical equipment is data extracted from process data in the acquired operation data, and the data extracted from mechanical vibration data in the acquired operation data.

In some embodiments, the reference data of the detection points may be obtained dynamically in the above embodiments and/or statically in normal operation of the rotating machine. For example, in some embodiments, the reference data of the detection point is obtained in a dynamic manner in the above embodiments; for another example, in some embodiments, the reference data of the detection point is obtained in a static manner in the above embodiments; for another example, in some embodiments, a part of the reference data of the detection point is obtained in a dynamic manner in the above embodiments, another part of the reference data of the detection point is obtained in a static manner in the above embodiments, and the like.

In an exemplary embodiment, when the acquired operational data comprises process data and/or operating condition data; the second fault diagnosis system also analyzes the acquired process data and/or operating condition data to obtain benchmark data in the reference data of the detection point during normal operation of the rotary mechanical equipment.

In some specific examples, the second fault diagnosis system calculates corresponding reference data by using process data meeting normal operation conditions according to the abnormal type corresponding to the detection point. For example, the second fault diagnosis system is configured to detect an abnormal air outlet at an air outlet of the fan, and use the temperature, flow rate, pressure, density, and the like of inlet air to calculate reference data of the fan energy efficiency of the fan under the current operating condition. In some specific examples, the second fault diagnosis system obtains the reference data by using the operating condition data obtained from the detection point or the locally pre-stored operating condition data according to the abnormality type corresponding to the detection point. For example, the second fault diagnosis system determines that the reference data includes rotational speed data according to rotational speed data corresponding to the same operating mode, which is continuously acquired for multiple times, for detecting rotational abnormality of the bearing. In some specific examples, the second fault diagnosis system obtains the reference data by using the operating condition data obtained from the detection point or the locally pre-stored operating condition data and the process data according to the abnormality type corresponding to the detection point. For example, the second fault diagnosis system detects an abnormality of the air outlet of the fan, and uses the operating mode of the air outlet and the motor current as reference data.

Here, when the detected data includes process data, operating condition data, or both process data and operating condition data, the second fault diagnosis system analyzes the process data, the operating condition data, or both the process data and the operating condition data, thereby obtaining reference data used as a standard in reference data of the detected point when the rotary mechanical device is operating normally, so as to determine whether the detected data is abnormal using the reference data.

The reference data can be obtained in a static manner, in addition to the dynamic manner. The reference data comprise initial parameters and/or predetermined calibration parameters of the rotating machine. The initial parameter includes, for example, factory process data of the rotating mechanical device. The preset calibration parameters are data obtained by calibrating all or part of process data and the like of the rotating mechanical equipment by a manager of the rotating mechanical equipment according to experience, for example: in some scenarios, due to a problem of a geographical environment, such as a high temperature of a production environment, a temperature abnormality of the device may be caused, so that a difference between temperature data of the rotating mechanical device when the rotating mechanical device operates in the environment and temperature data when the rotating mechanical device leaves a factory is large, but since the abnormality of the temperature data is not caused by a fault of the rotating mechanical device, a manager of the rotating mechanical device may calibrate the temperature data of the rotating mechanical device according to experience, so as to use the calibrated temperature data as reference data. It should be understood that the second failure diagnosis system may acquire the reference data once each time the failure diagnosis is performed; the reference data acquired once may be stored in the storage medium, and the reference data in the storage medium may be called each time the failure diagnosis needs to be executed.

In an exemplary embodiment, the second fault diagnosis system further performs feature extraction on the acquired detection data based on reference data of the detection point in normal operation of the rotary mechanical device to obtain at least one feature element for detecting at least one abnormality type of the detection point.

In some embodiments, after the second fault diagnosis system acquires the operation data, there may be a case where the operation data acquired from the sensor or the like is disturbed or a measurement error occurs. In this regard, after preprocessing the acquired operation data, the second fault diagnosis system generates detection data (or detection data and reference data) from the operation data, and performs feature extraction on the detection data. The preprocessing method includes but is not limited to noise reduction, outlier rejection, and the like.

In some embodiments, the second fault diagnosis system performs feature extraction on the detection data according to the reference data, thereby obtaining at least one feature element for detecting at least one anomaly type of the detection point. The feature extraction method includes, but is not limited to, average value calculation, effective value calculation, spectrum extraction, envelope spectrum extraction, etc., and the feature elements obtained by feature extraction are used as input for analyzing the anomaly type.

In some embodiments, after the second fault diagnosis system performs feature extraction on the detection data according to the reference data, the second fault diagnosis system further performs feature engineering on the result after the feature extraction, so as to process the result after the feature extraction into feature elements required for each abnormality type. The feature engineering includes but is not limited to feature combination, feature dimension reduction, feature processing, feature normalization, and the like.

In some embodiments, to facilitate the analysis of the abnormality, the second fault diagnosis system further includes performing further data processing on the obtained at least one feature element, so as to analyze the at least one feature element after the further data processing, thereby obtaining an analysis result of the corresponding abnormality type. Here, the data processing includes, but is not limited to, mathematical operations and the like. The feature elements processed by the data can also be used for inputting into an anomaly detection model for analysis, so that accurate analysis is realized under the condition of insufficient data sample amount. For example, the component at the frequency 2 times the fundamental frequency, the component at the frequency 3 times the fundamental frequency, the component at the frequency 4 times the fundamental frequency, the component at the frequency 5 times the fundamental frequency, etc. are combined in increments of several higher order multiples to form a feature element.

Because different types of abnormality need to use different types and different descriptions of feature elements as judgment, each obtained feature element can correspond to one or more abnormality types. In other words, the same feature element may be reused in different exception types. The set characteristic extraction mode can be determined according to the experience of a manager of the rotary mechanical equipment, so that the criticality of the extracted characteristics is ensured, and the accuracy and the efficiency of an analysis result are ensured.

In some embodiments, the second fault diagnosis system further includes a feature mechanism model, the input of the feature mechanism model is detection data and reference data, and the output of the feature mechanism model is feature elements corresponding to each abnormality type. In this case, the feature mechanism model performs feature extraction on the detection data by referring to the data. The feature mechanism model determines features to be extracted according to a preset rule, and performs feature extraction on detection data by using reference data to generate feature values by combining algorithms such as average value calculation, effective value calculation, frequency spectrum extraction and envelope spectrum extraction. Wherein the reference data may be obtained statically and/or obtained dynamically. Meanwhile, the characteristic mechanism model further processes the characteristic value into each characteristic element corresponding to each abnormal type so as to analyze each characteristic element and obtain the analysis result of the corresponding abnormal type. The characteristic mechanism model can be constructed through the mechanism model by utilizing pre-marked historical operating data, reference data and the like. For example, historical operation data of the rotating machine is acquired from a manager, a control system, or the like of the rotating machine, and initial parameters of the rotating machine are acquired from a manager, a control system, a network, or the like of the rotating machine. It should be understood that the mechanism model, also known as the white box model. Which is an accurate mathematical model built from objects, internal mechanisms of the production process, or transport mechanisms of the material flow. The algorithm in the feature mechanism model includes but is not limited to feature value calculation, feature engineering, and the like, and the feature value calculation includes but is not limited to: taking mechanical vibration data as an example, the characteristic value calculation comprises processing the acquired historical mechanical vibration data into a frequency domain spectrum, acquiring a fundamental frequency from the mechanical vibration data and the like to obtain at least one characteristic value. The feature engineering includes, but is not limited to, subjecting the at least one feature value to feature combination dimension reduction, feature processing, feature normalization, and the like to obtain at least one feature element. And if the accuracy of the output result of the trained feature mechanism model reaches a preset accuracy threshold, finishing the training.

Here, a process of calculating the feature element when the acquired operation data includes mechanical vibration data or the acquired detection data includes process data and/or condition data is exemplified, respectively.

Wherein, when the acquired operation data contains mechanical vibration data, the second fault diagnosis system extracts at least one frequency feature element in the frequency spectrum of the acquired mechanical vibration data, which is related to fundamental frequency data in the reference data of the detection point, for detecting at least one abnormality type of the detection point. The fundamental frequency data corresponds to a frequency or a frequency interval with low frequency and high intensity in a vibration spectrum generated when the rotary mechanical equipment normally operates under the current working condition during the execution of the current production process. According to the actual working conditions during the production process executed by the rotary mechanical equipment, the fundamental frequency data of different production processes and different working conditions are not completely consistent. To this end, in some specific examples, the fundamental frequency data is extracted from operational data. For example, the mechanical vibration data of a certain dimension of the blade is subjected to frequency domain conversion, and the fundamental frequency data is obtained through the analysis of the frequency spectrum distribution. In another specific example, the fundamental frequency data is selected from a plurality of corresponding relations of the locally stored process, the operating condition and the fundamental frequency data according to the acquired operating condition data, the process data and the like.

It is understood that rotary mechanical devices (e.g., fans, motors, water pumps, gear boxes, etc.) in industrial operations are rotated and perform work under the drive of motors to output energy. Thus, the motor, the rotating part and the like have own rotating speeds, the rotating speeds can induce equipment vibration to a certain extent, and the fundamental frequency of the vibration can be calculated through the rotating speeds of the equipment. Meanwhile, the vibration frequency spectrum caused by the rotating speed is based on the fundamental frequency, and frequency spectrum components such as integral multiple, fractional multiple, characteristic multiple, high-frequency modulation and the like of the fundamental frequency are additionally superposed. These spectral components tend to be small during normal operation of the device (less vibrational energy during normal operation), while different faults may exhibit different vibrational spectra during a failure of the device.

It will be appreciated that in some embodiments, the fundamental frequency may be calculated from the rotational speed, for example: n/60, wherein: f is the frequency (unit: HZ) and n is the rotational speed (unit: rpm). In other embodiments, the fundamental frequency can also be analyzed from the vibration spectrum by means of a spectral analysis.

The frequency feature elements include, but are not limited to, a multiplying power spectral component that is a fundamental frequency, and the like. For example: the frequency domain conversion operation is carried out on the obtained mechanical vibration data, and the frequency spectrum distribution is analyzed, so that frequency characteristic elements such as components on 2-frequency multiplication and components on 4-frequency multiplication of the fundamental frequency are obtained. In some embodiments, the frequency feature element may also be a spectral component of other specific frequencies, such as a component at the frequency 3 multiplied by 0.5.

When the acquired detection data includes process data and/or operating condition data, the second fault diagnosis system further obtains at least one deviation feature element based on a deviation of reference data in normal operation of the rotary mechanical equipment from the acquired detection data, for detecting at least one abnormality type of the detection point.

In this case, the acquired test data are compared with reference data, so that at least one deviation characteristic element is obtained, which includes, but is not limited to: increments, percentages of increments, etc. For example: the acquired detection data is temperature data, and the reference temperature data in the reference data and the temperature data in the detection data are compared when the rotary mechanical equipment normally operates, so that the difference value of the temperature change and/or the percentage of the difference value to the reference temperature data are extracted. And respectively using the difference and/or the percentage of the difference relative to the reference temperature data as a deviation characteristic element so as to detect at least one abnormal type of the detection point.

After obtaining the at least one feature element in the various manners of the above embodiment, the second fault diagnosis system analyzes each feature element corresponding to each abnormality type to obtain an analysis result of the corresponding abnormality type.

The method for analyzing the feature elements corresponding to each anomaly type includes, but is not limited to, analyzing by using an anomaly detection model, where the anomaly detection model is a machine learning model, and the machine learning model includes, but is not limited to, a KNN (k-Nearest Neighbor, k-Neighbor algorithm) -based machine learning model, and the like. In some embodiments, each anomaly type has a separate machine learning model, such as: the method comprises the following steps that an impeller unbalance model, a coupling misalignment model, a rolling bearing vibration abnormity and a sliding bearing vibration abnormity are respectively corresponding to the impeller unbalance model, the coupling misalignment model, the rolling bearing fault model and the sliding bearing fault model; the bearing temperature abnormity and the motor current abnormity correspond to a bearing temperature abnormity model, a motor current abnormity model and the like. Since the data required when analyzing each anomaly type is different, the second failure diagnosis system converts the detection data into inputs, i.e., feature elements, required for these models to obtain the corresponding analysis results. For example: and the coupler misalignment model needs a component on 2 frequency multiplication of the fundamental frequency, a component on 3 frequency multiplication of the fundamental frequency and a component on 4 frequency multiplication of the fundamental frequency as input, and the fault diagnosis system performs feature extraction on the detection data, and then enters the coupler misalignment model for analysis by taking two feature elements of the component on 2 frequency multiplication of the fundamental frequency, the component on 3 frequency multiplication of the fundamental frequency, the component on 4 frequency multiplication of the fundamental frequency and the fundamental frequency as input, so as to obtain an analysis result of the coupler misalignment model. At least the machine learning algorithm in the anomaly detection model can be trained through pre-marked historical operating data and the like to obtain parameters in the algorithm. For example, historical operation data of a rotating machine and historical faults corresponding to the historical operation data are acquired from a manager, a control system, or the like of the rotating machine. And processing the acquired data into sample data required by a corresponding algorithm and training the algorithm to obtain an anomaly detection model. Wherein the process is used to convert the acquired data into data that can be processed by an algorithm, including but not limited to: normalization processing, data conversion according to a preset conversion formula and the like. And if the accuracy of the trained anomaly detection model reaches a preset accuracy threshold, finishing the training. The input of the abnormality detection model is a feature element required by the model, and the second fault diagnosis system determines an abnormality type analysis result corresponding to each abnormality of the rotary machine using the abnormality detection model.

Here, after the second fault diagnosis system obtains the analysis result corresponding to each abnormality type at the at least one detection point, the second fault diagnosis system performs diagnosis processing on at least one fault in the rotary mechanical device using the obtained at least one analysis result to output a corresponding fault diagnosis result.

Here, according to the actual conditions such as the number of the operation data provided by the detection points, the network transmission efficiency, and the like, the second fault diagnosis system obtains at least one analysis result, and the second fault diagnosis system performs fault diagnosis processing according to the at least one analysis result. For example, the second fault diagnosis system performs diagnosis of the imbalance fault of the blade or the non-alignment fault of the coupler according to the obtained analysis result of the blade vibration abnormal type of the corresponding blade detection point, or performs diagnosis of the imbalance fault of the blade and the non-alignment fault of the coupler respectively, and obtains a fault diagnosis result through a diagnosis evaluation system.

It should be understood that when different parts of the rotating machinery fail, there are different anomalies. Taking a fan as an example, when a bearing of the fan breaks down, the vibration of the bearing part is strong and the temperature is increased, but the current abnormality of a fan motor is not caused; for another example, when a blade of a fan fails, the temperature of the bearing portion does not rise, but the fan as a whole is shaken. Therefore, if only the abnormality type analysis result of a single detection point is taken as the failure diagnosis result, the result is low in accuracy, whereas if the abnormality type analysis results of a plurality of detection points are comprehensively diagnosed, a diagnosis result with higher accuracy is obtained.

Here, the second fault diagnosis system further includes a comprehensive diagnosis model, so that the abnormality type analysis result of each detection point on the rotary machine is comprehensively diagnosed and processed, and the fault diagnosis result of the rotary machine is obtained by combining the position and type comprehensive processing of each detection point. The comprehensive diagnosis model is a mechanism model, the input of the comprehensive diagnosis model is the analysis result corresponding to each abnormal type on at least one detection point, and the output of the comprehensive diagnosis model is the fault diagnosis result of the rotary mechanical equipment.

In an exemplary embodiment, the integrated diagnostic model includes a plurality of independent fault models, such as: the second fault diagnosis system correspondingly inputs the abnormal type analysis results of all detection points into the fault model. Where the inputs required for each fault type are different and the same input may also be used for different fault types. For example: the input of the unbalanced fault model of the impeller corresponds to the analysis result of the unbalanced abnormal type of the impeller on the detection point, the input of the unbalanced fault model of the coupling corresponds to the analysis result of the unbalanced abnormal type of the coupling on the detection point and the analysis result of the abnormal type of the current on the detection point, the input of the fault model of the fan side bearing corresponds to the analysis result of the abnormal type of the rolling bearing and/or the abnormal type of the sliding bearing on the detection point, the input of the fault model of the motor side bearing corresponds to the analysis result of the abnormal type of the rolling bearing and/or the abnormal type of the sliding bearing on the detection point, the analysis result of the abnormal type of the temperature on the detection point and the like.

The second fault diagnosis system presets diagnosis rules of a plurality of fault models in the comprehensive diagnosis model so as to diagnose faults of the rotating mechanical equipment through the analysis result of each abnormal type on at least one detection point. For example: taking the fan as an example, the second fault diagnosis system presets that when the analysis result of the rolling bearing abnormal type and the analysis result of the temperature abnormal type at the detection point are both abnormal, and the analysis results of other abnormal types at the detection point are both normal, the comprehensive diagnosis model outputs the fault diagnosis result as the fault diagnosis result of the motor side bearing. It should be appreciated that, since multiple faults may exist in the rotating machine at the same time, in some embodiments, the output of the comprehensive diagnostic model has multiple fault diagnosis results. In still other embodiments, when the second fault diagnosis system fails to diagnose a fault of the rotary machine, a diagnosis result of an unknown fault is output.

In an exemplary embodiment, the second fault diagnosis system displays the fault diagnosis result in order to facilitate management or operation by a manager of the rotary machine. To this end, the second fault diagnosis system also presents the obtained fault diagnosis result in a display interface of a control system of the rotary machine apparatus. In some embodiments, a display may be connected to the computer device where the control system is located, and a manager of the rotating mechanical device may view the fault diagnosis result through the display. In still other embodiments, after obtaining the fault diagnosis result, the manager of the rotating machine overhauls or checks the rotating machine according to the fault diagnosis result, and feeds back a conclusion whether the fault diagnosis result is correct to the second fault diagnosis system through the control system.

For convenience of understanding, the process of diagnosing the fault of the rotating mechanical equipment by the second fault diagnosis system will be illustrated below by using 5 detection points, and it should be understood that the present embodiment is only used for explaining the present application, and is not used for limiting the present application.

In this embodiment, the second fault diagnosis system first obtains the operation data of the sensor at the first detection point, where the sensor is a vibration sensor, and the vibration sensor provides the mechanical vibration data to the second fault diagnosis system. And after the second fault diagnosis system preprocesses the mechanical vibration data, inputting the preprocessed mechanical vibration data into the characteristic mechanism model. The characteristic mechanism model firstly generates vibration mechanical data into detection data. Meanwhile, the second fault diagnosis system acquires reference data corresponding to the detection data, namely mechanical vibration data when the rotating mechanical equipment normally operates, and inputs the reference data to the characteristic mechanism model. The feature mechanism model records the detection data and the reference data into time domain waveforms, and further converts the time domain waveforms of the reference data and the detection data into frequency domain spectrums.

The characteristic mechanism model processes characteristic values corresponding to the anomalies to obtain a plurality of first characteristic elements. The processing method comprises the step of converting the characteristic values into frequency spectrum components of specific frequency multiplying power so as to input the frequency spectrum components into various anomaly detection models for analysis. After obtaining the plurality of first feature elements, some of the first feature element data is further processed, so as to reduce the dimension and form a plurality of second feature elements.

The characteristic mechanism model provides the first characteristic element and/or the second characteristic element to a corresponding abnormality detection model, wherein the abnormality detection model comprises an impeller imbalance model, a coupling misalignment model, a rolling bearing fault model and a sliding bearing fault model, each abnormality detection model respectively outputs an analysis result, namely whether the possibility of a fault corresponding to the model exists, for example, the possibility of the impeller imbalance model outputting the impeller imbalance abnormality is 30%, the possibility of the coupling misalignment model outputting the coupling misalignment abnormality is 80%, and the like. Here, the second failure diagnosis system obtains an analysis result output by each abnormality detection model at the first detection point.

The second detection point and the third detection point are both vibration sensors, and the way of obtaining the analysis result is the same as that of the first detection point, so that the repeated description is omitted.

Secondly, the second fault diagnosis system also acquires the operation data of a sensor on a fourth detection point, wherein the sensor is a temperature sensor, and the vibration sensor provides the temperature data for the second fault diagnosis system. And after preprocessing the temperature data, the second fault diagnosis system inputs the preprocessed temperature data into the characteristic mechanism model. The characteristic mechanism model firstly generates detection data from temperature data. Meanwhile, the second fault diagnosis system acquires reference data corresponding to the detection data, namely temperature data of the rotating mechanical equipment during normal operation, and inputs the reference data to the characteristic mechanism model. The feature mechanism model compares the detection data with the reference data to calculate the increment percentage between the detection data and the reference data, and uses the increment percentage as a feature element, for example, the detection data is 30% higher than the reference data, and the like. The feature mechanism model provides the feature elements to a corresponding abnormality detection model, where the abnormality detection model includes a temperature abnormality model that outputs an analysis result, that is, whether there is a possibility of a fault corresponding to the model, for example, the temperature abnormality model outputs a possibility of a temperature abnormality of 10%. Here, the second failure diagnosis system obtains an analysis result output by the abnormality detection model at the fourth detection point.

Meanwhile, the second fault diagnosis system also acquires operation data of a sensor on a fifth detection point, wherein the sensor is a current sensor, and the vibration sensor provides the current data to the second fault diagnosis system. And after preprocessing the current data, the second fault diagnosis system inputs the preprocessed current data into the characteristic mechanism model. The characteristic mechanism model firstly generates current data into detection data. Meanwhile, the second fault diagnosis system acquires reference data corresponding to the detection data, namely current data of the rotating mechanical equipment during normal operation, and inputs the reference data to the characteristic mechanism model. The feature mechanism model compares the detection data with the reference data to calculate the increment percentage between the detection data and the reference data, and uses the increment percentage as a feature element, for example, the detection data is 30% higher than the reference data, and the like. The feature mechanism model provides the feature elements to a corresponding abnormality detection model, where the abnormality detection model includes a current variation model, and the current variation model outputs an analysis result, that is, whether there is a possibility of a fault corresponding to the model, for example, the current variation model has a possibility of outputting a current abnormality of 10%. Here, the second failure diagnosis system obtains an analysis result output by the abnormality detection model at the fifth detection point.

Here, the analysis result output by each anomaly detection model at each of the 5 detection points is provided to the comprehensive diagnosis model, so that the analysis result of the anomaly type at each detection point is comprehensively diagnosed and processed, and the fault diagnosis result of the rotary machine is obtained by combining the position and type comprehensive processing of each detection point.

The second fault diagnosis system splits a complex single machine learning model into a single abnormal detection model, so that the complexity of the model is reduced, and a high-precision fault detection effect can be obtained by adopting less sample learning; in addition, the second fault diagnosis system in the application does not depend on the collected comprehensive abnormal types, and can provide corresponding fault diagnosis results according to the analysis results of the abnormal types which can be collected actually, so that the flexibility of matching among machine learning models is effectively improved.

An embodiment of a sixth aspect of the present application provides a management system for a rotating machine, where the management system for a rotating machine includes a detection device disposed at each detection point of the rotating machine, a control system for the rotating machine, and a server as described in the embodiment of the second aspect of the present application.

The detection device configured at each detection point of the rotary mechanical equipment is used for providing operation data of each detection point on the rotary mechanical equipment. And the control system of the rotary mechanical equipment is in data connection with each detection device so as to collect and forward each operation data to the server, so that the server in communication connection with the control system receives each operation data and executes a corresponding fault diagnosis method based on the received operation data. The fault diagnosis method is consistent with the fault diagnosis method in the foregoing embodiments, and therefore, the details are not repeated.

As described above, the method, system, and storage medium for diagnosing a fault of a rotating machine according to the present application have the following advantageous effects: according to the method and the device, the acquired data are subjected to continuous dimensionality reduction processing, and the data are subjected to feature extraction through preset key indexes, so that the accuracy is guaranteed under the condition of a small amount of data samples. Secondly, the method can flexibly expand the process parameters, the newly added parameters do not affect the machine learning model of the existing parameters, the existing model does not need to be retrained, only the newly added parameters need to be newly built, and the relation model of the newly added parameters and the existing parameters is added into the comprehensive diagnosis model. In addition, the three-layer model structure comprises a characteristic mechanism model and a machine learning model, and the position, type and the like of the detection point are analyzed through a comprehensive diagnosis model, so that management experience is fused with a machine learning algorithm, and the accuracy of a fault diagnosis result is ensured.

The above embodiments are merely illustrative of the principles and utilities of the present application and are not intended to limit the application. Any person skilled in the art can modify or change the above-described embodiments without departing from the spirit and scope of the present application. Accordingly, it is intended that all equivalent modifications or changes which can be made by those skilled in the art without departing from the spirit and technical concepts disclosed in the present application shall be covered by the claims of the present application.

Claims (24)

1. A fault diagnosis method for rotary mechanical equipment is characterized by comprising the following steps:

   acquiring operation data on at least one detection point in a rotating mechanical device;

   respectively carrying out anomaly detection analysis on at least one anomaly type of the detection points by using the running data to obtain an analysis result corresponding to each anomaly type;

   and diagnosing at least one fault in the rotary mechanical equipment by using the obtained at least one analysis result so as to output a corresponding fault diagnosis result.
2. The method of diagnosing a malfunction of a rotary mechanical device according to claim 1, wherein the abnormality type includes at least one of the following abnormalities reflected by respective detection points during operation of the rotary mechanical device based on operating conditions and process requirements: an abnormality type set separately based on each vibration abnormality of at least one spatial dimension, an abnormality type set based on a temperature abnormality, an abnormality type set separately based on each power abnormality.
3. The method of claim 1, wherein the operational data is used to generate test data and reference data; the step of respectively performing anomaly detection analysis on at least one anomaly type of the detection points by using the operation data comprises the following steps:

   performing feature extraction on the acquired detection data based on the reference data of the detection point during normal operation of the rotating mechanical equipment to obtain at least one feature element for detecting at least one abnormal type of the detection point;

   and analyzing the characteristic elements corresponding to each abnormal type to obtain an analysis result of the corresponding abnormal type.
4. The method of diagnosing faults in rotating machinery according to claim 3, further comprising the steps of: and performing further data processing on the obtained at least one characteristic element to analyze the at least one characteristic element after the further data processing, so as to obtain an analysis result of the corresponding abnormal type.
5. The method of fault diagnosis of a rotating mechanical device according to claim 3 or 4, characterized in that said reference data comprise at least one or more of the following: the working condition data of the rotating mechanical equipment is data extracted from process data in the acquired operation data, and the data extracted from mechanical vibration data in the acquired operation data.
6. The method of diagnosing faults in rotating machinery according to claim 3, further comprising the steps of: and analyzing the fundamental frequency data in the corresponding reference data when the rotating mechanical equipment normally operates from the acquired operation data.
7. The rotating machine equipment failure diagnosis method according to claim 3 or 6, characterized in that the acquired operation data contains mechanical vibration data; the step of performing feature extraction on the acquired detection data based on the reference data of the detection point during normal operation of the rotating machinery to obtain at least one feature element for detecting at least one abnormal type of the detection point comprises:

   extracting at least one frequency feature element in the acquired frequency spectrum of the mechanical vibration data, which is related to the fundamental frequency data in the reference data of the detection point, for detecting at least one anomaly type of the detection point.
8. The method of claim 3, wherein the acquired operational data comprises process data and/or operating condition data; the method further comprises the steps of: and analyzing the acquired process data and/or working condition data to obtain benchmark data in the reference data of the detection point during normal operation of the rotary mechanical equipment.
9. The method according to claim 8, wherein the step of performing feature extraction on the acquired detection data based on the reference data of the detection points in normal operation of the rotary machine to obtain at least one feature element for detecting at least one abnormality type of the detection points comprises:

   obtaining at least one deviation feature element based on a deviation of the acquired detection data from the reference data in normal operation of the rotating mechanical device, for detecting at least one anomaly type of the detection point.
10. The method according to claim 1, wherein the step of performing a diagnostic process on at least one fault in the rotary machine using the obtained at least one analysis result to output a corresponding fault diagnosis result includes: presenting the obtained fault diagnosis result in a display interface of a control system of the rotary mechanical equipment.
11. A server, comprising:

    the interface unit is used for carrying out data communication with the sensor on at least one dimension on the detection point of the rotary mechanical equipment;

    a storage unit for storing at least one program; and

    a processing unit for calling the at least one program to coordinate the execution of the interface unit and the storage unit and to implement the fault diagnosis method of the rotating machinery according to any one of claims 1 to 10.
12. A first fault diagnosis system of a rotary mechanical apparatus, comprising:

    the server of claim 11; and

    and the detection device is arranged at each detection point of the rotating mechanical equipment, is in communication connection with the server and is used for providing operation data of each detection point.
13. A computer-readable storage medium characterized by storing at least one program which, when called, executes and implements a rotary mechanical device failure diagnosis method according to any one of claims 1 to 10.
14. A second fault diagnosis system of a rotary mechanical apparatus, comprising:

    the data acquisition module is used for acquiring operation data on at least one detection point in a piece of rotary mechanical equipment;

    and the data processing module is used for respectively carrying out abnormality detection analysis on at least one abnormality type of the detection points by using the operation data so as to obtain an analysis result corresponding to each abnormality type, and diagnosing at least one fault in the rotary mechanical equipment by using the obtained at least one analysis result so as to output a corresponding fault diagnosis result.
15. The second failure diagnosis system of a rotary mechanical apparatus according to claim 14, wherein the abnormality type includes at least one of the following abnormalities reflected by the respective detection points during operation of the rotary mechanical apparatus based on the operating conditions and the process requirements: an abnormality type set separately based on each vibration abnormality of at least one spatial dimension, an abnormality type set based on a temperature abnormality, an abnormality type set separately based on each power abnormality.
16. The second failure diagnosis system for a rotary mechanical apparatus according to claim 14, wherein the operation data is used to generate detection data and reference data; the data processing module performs feature extraction on the acquired detection data based on the reference data of the detection points during normal operation of the rotary mechanical equipment to obtain at least one feature element for detecting at least one abnormal type of the detection points; and analyzing the characteristic elements corresponding to each abnormal type to obtain the analysis result of the corresponding abnormal type.
17. The system of claim 16, wherein the data processing module further processes the obtained at least one feature element to analyze the at least one feature element after further processing, so as to obtain an analysis result of the corresponding abnormality type.
18. The second fault diagnosis system of a rotary mechanical apparatus according to claim 16 or 17, wherein the reference data includes at least one or more of: the working condition data of the rotating mechanical equipment is data extracted from process data in the acquired operation data, and the data extracted from mechanical vibration data in the acquired operation data.
19. The system of claim 16, wherein the data processing module further analyzes fundamental frequency data of the reference data corresponding to normal operation of the rotating machine from the acquired operation data.
20. The second failure diagnosis system for a rotary mechanical apparatus according to claim 16 or 19, wherein the acquired operation data contains mechanical vibration data, and the data processing module extracts at least one frequency feature element in the frequency spectrum of the acquired mechanical vibration data, which is associated with the fundamental frequency data in the reference data of the detection point, for detecting at least one abnormality type of the detection point.
21. The second fault diagnosis system of a rotary mechanical apparatus according to claim 16, wherein the acquired operation data includes process data and/or operating condition data; the data processing module is also used for analyzing the acquired process data and/or working condition data to obtain datum data in the reference data of the detection point during normal operation of the rotary mechanical equipment.
22. The system of claim 21, wherein the data processing module obtains at least one deviation feature element for detecting at least one abnormality type of the detection point based on a deviation of the acquired detection data from the reference data in the normal operation of the rotating machine.
23. The second failure diagnosis system for a rotary mechanical apparatus according to claim 21, further comprising a failure display device for presenting the obtained failure diagnosis result.
24. A management system for a rotating machine, comprising:

    the detection device is arranged at each detection point of the rotating mechanical equipment and used for providing operation data of each detection point;

    the control system of the rotating mechanical equipment is in data connection with each detection device and used for collecting and forwarding each operation data;

    the server of claim 11, communicatively coupled to the control system, configured to receive each of the operational data and to perform a corresponding fault diagnostic method based on the received operational data.

Images