A method for diagnosing electromechanical equipment by training a multi-dimensional operation and maintenance data model and operation and maintenance equipment

Abstract

The application discloses a kind of multi-dimensional operation and maintenance data training model electromechanical equipment diagnostic method and operation and maintenance equipment, it is related to electromechanical equipment operation and maintenance technical field, deployment multidimensional sensor network is carried out staggered cross sampling, obtains electromechanical current data, introduces first verification port and second verification port, constructs diagnostic network architecture, into network to electromechanical current data, by executing the aggregation screening based on first verification port and the data structure conversion and encapsulation based on second verification port, back to data center, based on built-in operation and maintenance diagnostic instrument, operation and maintenance fault decision and trend prediction are carried out, and output fault diagnosis single. For solving the technical problems that lack intelligent diagnostic architecture in the prior art, cannot guarantee the effectiveness and accuracy of data sampling and diagnostic decision, carry out flexible sampling with demand as the guide, construct systematic diagnostic architecture, to improve the intelligence and accuracy of electromechanical equipment diagnostic decision.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of mechanical and electrical equipment operation and maintenance, and particularly relates to a mechanical and electrical equipment diagnosis method of a multi-dimensional operation and maintenance data training model and an operation and maintenance device. BACKGROUND

[0002] In the current field of mechanical and electrical equipment operation and maintenance, traditional diagnosis methods mainly rely on manual inspection and experience judgment. This method is not only inefficient, but also difficult to accurately capture early fault signals of the equipment. With the development of Internet of Things technology, multi-dimensional sensor networks have begun to be applied to data collection of mechanical and electrical equipment. However, existing technologies often ignore the optimization of data sampling methods, resulting in uneven data quality. In addition, the lack of effective data processing methods causes data to be mixed, affecting the diagnosis logic. At the same time, existing diagnosis systems are mostly based on a single fault model, which is difficult to adapt to the complex and variable fault characteristics of mechanical and electrical equipment.

[0003] Therefore, how to efficiently and accurately collect and process mechanical and electrical equipment operation and maintenance data and construct an intelligent diagnosis architecture has become a technical problem to be solved. SUMMARY

[0004] The present application provides a mechanical and electrical equipment diagnosis method of a multi-dimensional operation and maintenance data training model and an operation and maintenance device, which is used to solve the technical problem that the existing technology lacks an intelligent diagnosis architecture and cannot guarantee the effectiveness and accuracy of data sampling and diagnosis decision.

[0005] In view of the above problems, the present application provides a mechanical and electrical equipment diagnosis method of a multi-dimensional operation and maintenance data training model and an operation and maintenance device.

[0006] In a first aspect, the present application provides a mechanical and electrical equipment diagnosis method of a multi-dimensional operation and maintenance data training model, which comprises: deploying a multi-dimensional sensor network in a mechanical and electrical equipment area, acquiring mechanical and electrical flow data by adopting a staggered cross-sampling method; introducing a first verification port and a second verification port to construct a diagnosis network architecture, wherein the first verification port is arranged at a transmission starting segment of the sensing side and performs multi-dimensional data aggregation, and the second verification port is arranged at a transmission middle segment and performs data vector encapsulation; based on the diagnosis network architecture, the mechanical and electrical flow data are networked, and through the execution of the aggregation screening based on the first verification port and the data structure conversion and encapsulation based on the second verification port, the data are returned to a data center, based on a built-in operation and maintenance diagnosis device, operation and maintenance fault decision and trend prediction are performed, and a fault diagnosis single is output.

[0007] In a second aspect, the present application provides an operation and maintenance device, which is composed of a multi-dimensional sensing network, a diagnosis network architecture and an operation and maintenance diagnostic device; wherein the multi-dimensional sensing network performs front-end sampling and on-site early warning, and the diagnosis network architecture and the operation and maintenance diagnostic device perform data interaction and processing, and cooperatively constitute a diagnosis system of the electromechanical equipment.

[0008] The one or more technical solutions provided in the present application have at least the following technical effects or advantages:

[0009] The electromechanical equipment diagnosis method of the multi-dimensional operation and maintenance data training model provided in the embodiments of the present application deploys a multi-dimensional sensing network in an electromechanical equipment area, acquires electromechanical current data by adopting a staggered cross-sampling mode, introduces a first verification port and a second verification port, constructs a diagnosis network architecture, the first verification port is arranged at a transmission starting segment of the sensing side and performs multi-dimensional data aggregation, and the second verification port is arranged at a transmission middle segment and performs data vector encapsulation; based on the diagnosis network architecture, the electromechanical current data is networked, and through performing the aggregation screening based on the first verification port and the data structure conversion and encapsulation based on the second verification port, the data is returned to a data center, based on the built-in operation and maintenance diagnostic device, operation and maintenance fault decision and trend prediction are performed, and a fault diagnosis sheet is output. The technical problem that the existing technology lacks an intelligent diagnosis architecture and cannot guarantee the effectiveness and accuracy of data sampling and diagnosis decision is solved, flexible sampling is performed according to the demand, a systematic diagnosis architecture is constructed, and the intelligence and accuracy of the electromechanical equipment diagnosis decision are improved. BRIEF DESCRIPTION OF DRAWINGS

[0010] Figure 1 An electromechanical equipment diagnosis method flowchart of the multi-dimensional operation and maintenance data training model is provided for the present application.

[0011] Figure 2 A sampling mode configuration flowchart of the electromechanical equipment diagnosis method of the multi-dimensional operation and maintenance data training model is provided for the present application. DETAILED DESCRIPTION

[0012] The present application provides an electromechanical equipment diagnosis method of a multi-dimensional operation and maintenance data training model and an operation and maintenance device, deploys a multi-dimensional sensing network for staggered cross-sampling, acquires electromechanical current data, introduces a first verification port and a second verification port, constructs a diagnosis network architecture, networks the electromechanical current data, returns the data to a data center through performing the aggregation screening based on the first verification port and the data structure conversion and encapsulation based on the second verification port, based on the built-in operation and maintenance diagnostic device, performs operation and maintenance fault decision and trend prediction, and outputs a fault diagnosis sheet, which is used to solve the technical problem that the existing technology lacks an intelligent diagnosis architecture and cannot guarantee the effectiveness and accuracy of data sampling and diagnosis decision.

[0013] Embodiment one:

[0014] As Figure 1 indicated, the present application provides a kind of mechanical and electrical equipment diagnosis method of multi-dimensional operation and maintenance data training model, the method comprises:

[0015] S1: multi-dimensional sensing network is deployed in mechanical and electrical equipment area, obtains mechanical current data by using staggered cross sampling mode.

[0016] In the embodiment of the application, multi-dimensional sensing network refers to the network system composed of multiple different types of sensors that can sense multiple physical quantities. These sensors are deployed according to the parameter types required for monitoring the operation of mechanical and electrical equipment, such as voltage, current, vibration and temperature, etc. The deployment position is carefully planned to fully cover the key parts of the mechanical and electrical equipment to accurately capture various information during the operation of the equipment.

[0017] At the same time, staggered cross sampling mode is used to obtain more comprehensive and accurate mechanical current data. Staggered cross sampling, that is, according to the data characteristics of different sensing sources, such as the speed of data change, whether it has periodicity, etc., and the fault characteristics mined from the historical fault records of the equipment, to flexibly and specifically set the sampling time point and frequency. For example, for sensing sources with fast data change and non-periodic non-stationary fault characteristics, the sampling frequency may be increased and the sampling time may be staggered with other sensing sources.

[0018] In the implementation process, first, the data characteristics of the numerous multi-dimensional sensing sources in the mechanical and electrical equipment area are determined in detail, including the sensing source type (such as current sensor, vibration sensor, etc.) and the specific deployment position. Then, the historical fault records accumulated for a long time in this mechanical and electrical equipment area are called to deeply mine the fault characteristics, such as identifying whether the fault source is from the overheating fault of the motor or the short circuit fault of the line, etc. Then, according to the determined data characteristics and fault characteristics, the sampling mode is scientifically and reasonably configured. Finally, the multi-dimensional sensing network is driven by the sampling mode to obtain mechanical current data. This data collection method can effectively avoid data redundancy and improve the effectiveness and reliability of the data, providing a solid data foundation for the subsequent accurate diagnosis of mechanical and electrical equipment.

[0019] Further, by using staggered cross sampling mode, step S1 of the present application comprises:

[0020] The data characteristics of the multi-dimensional sensing source of the electromechanical equipment area are determined, wherein the sensing source type and the deployment position are identified; the historical fault record of the electromechanical equipment area is called to mine the fault characteristics, wherein the fault source is identified; and the sampling mode is configured according to the data characteristics and the fault characteristics; wherein the data characteristics include linear and nonlinear classes, and the fault characteristics include periodic stationary fault class, periodic non-stationary fault class, non-periodic stationary fault class and non-periodic non-stationary fault class.

[0021] Specifically, the multi-dimensional sensing source is a collection of various types of sensors deployed in a specific area of the electromechanical equipment to comprehensively collect information during equipment operation.

[0022] The sensing source types are diverse, such as current sensors that monitor and collect current values in real time during equipment operation through electromagnetic induction and other principles. The data obtained under normal equipment operation and different fault states will exhibit specific change rules. Different types of sensors are deployed in appropriate positions of the electromechanical equipment according to sensing requirements, for example, temperature sensors are deployed near the winding of the motor. Since the winding generates heat due to current passing through during operation, the temperature change here can directly reflect the load and operating state of the motor, and its data characteristics may exhibit complex change relationships related to factors such as motor operating time and load size.

[0023] Further, the data characteristics of each sensing source are determined, which are classified as linear or nonlinear in this application. Further, the historical fault record is a detailed record of various fault conditions that have occurred during the past operation of the electromechanical equipment. By mining fault characteristics, information with regularity and representing the essence of the fault is extracted from the historical record, and the corresponding fault source, i.e., the root cause of the equipment failure, is determined.

[0024] For example, when a key component inside the equipment suddenly breaks, this fault source will cause a dramatic and irregular change in the equipment operating parameters, exhibiting non-periodic non-stationary fault class characteristics. Such faults are extremely harmful to the normal operation of the equipment due to their suddenness and complexity. Different faults occur at different equipment locations, and different fault characteristics have different sampling requirements.

[0025] Further, the sampling mode is configured according to the data characteristics and the fault characteristics. The targeted configuration can guarantee the effectiveness and information completeness of the sampled stream data. In the data characteristics, the linear type means that the data collected by the sensor source has a relatively regular linear relationship with some factors. For example, within a certain load range, the current value of the motor may have a linear growth relationship with the load size. The nonlinear type means that the data variation law is more complex and cannot be described by a simple linear relationship, such as some sensor data affected by multiple complex factors.

[0026] The fault characteristics are divided into a periodic stationary fault type, such as the vibration anomaly caused by the periodic relaxation of the belt. This type of fault has periodicity in time and a relatively stable fault state. The periodic non-stationary fault type means that the fault occurrence has periodicity, but the operating parameters and other characteristics of the equipment are not stable within the fault period, that is, the fault characteristics present non-stationary changes within the period. The non-periodic stationary fault type means that the fault does not occur at a fixed period, but the fault state of the equipment is relatively stable after the fault occurs. The non-periodic non-stationary fault type, such as sudden fracture of a component, has no time regularity, and the operating state of the equipment is also in an extremely unstable state after the fault occurs.

[0027] Through staggered cross-sampling, a scientific and reasonable sampling mode is constructed to obtain high-quality data that meets the precise diagnosis requirements of the mechanical and electrical equipment.

[0028] Further, as shown in Figure 2 The sampling mode is configured, and the step S1 of the present application includes:

[0029] The data characteristics and the fault characteristics are combined to determine a first sampling type, wherein the first sampling type is any combination of characteristics. For the first sampling type, a first sensing part is located, and a first sampling condition is set, wherein the sampling condition at least includes a sampling frequency. The first sampling condition is identified, a time sequence is sampled, each sampling condition is cross-fused to determine a sampling mode, and data collection driving of the multi-dimensional sensing network is performed according to the sampling mode.

[0030] The data characteristics and the fault characteristics are combined, for example, the combination of the linear type data characteristics and the periodic stationary fault type, or the combination of the nonlinear type data characteristics and the non-periodic non-stationary fault type. The first sampling type is determined, and the first sampling type is any combination. The significance of this combination is to develop a more accurate sampling strategy for different combinations of data and fault characteristics.

[0031] For the determined first sampling class, determine the sensing detection part that meets the combination mode as the first sensing part. The first sensing part refers to the set of sensing sources corresponding to the first sampling class. The data characteristics of these sensing sources and the fault characteristics associated with them meet the characteristics of the current first sampling class.

[0032] Further, set the sampling condition based on the sampling requirement for the first sensing part as the first sampling condition, that is, the rule guiding the data collection of the sensor, which at least contains the sampling frequency. The sampling frequency refers to the number of data collection times of the sensing source per unit time, which is crucial for accurately capturing data changes. For different sampling classes, appropriate sampling frequencies need to be set according to their characteristics.

[0033] For example, for the combination of non-linear class data characteristics and non-periodic non-stationary fault class, due to the complex data changes and irregular fault bursts, a higher sampling frequency is needed to capture the data changes at the moment of fault occurrence in time. For the combination of linear class data characteristics and periodic stationary fault class, a lower sampling frequency can be set due to the relatively regular data changes.

[0034] Preferably, in addition to the sampling frequency, the sampling condition may also include parameters such as sampling duration and sampling accuracy, which together constitute the complete sampling condition.

[0035] Specifically, the sampling time sequence refers to a series of sampling time points arranged in chronological order. Cross-fusion is to reasonably arrange the sampling conditions of different combinations in the time dimension to avoid mutual interference of data collection of various sensing sources, while fully utilizing system resources. Through cross-fusion, their sampling time points can be staggered, so that there will not be too many sensing sources collecting data at the same time, reducing the load pressure of the system. By setting the sampling mode, each sensing source in the multi-dimensional sensing network can work cooperatively according to the predetermined rules, achieving efficient and accurate data collection.

[0036] Finally, according to the determined sampling mode, the data collection driving of the multi-dimensional sensing network is carried out. This means that the sampling mode is converted into actual control signals and sent to each sensor in the multi-dimensional sensing network, so that comprehensive, accurate and representative mechanical and electrical equipment operation data can be obtained, providing a solid data foundation for subsequent fault diagnosis and operation decision-making.

[0037] S2: Introduce a first verification port and a second verification port to build a diagnostic network architecture, wherein the first verification port is arranged at the transmission starting segment of the sensing side to perform multi-dimensional data aggregation, and the second verification port is arranged at the middle segment of the transmission to perform data vector packaging.

[0038] First of all, the purpose of introducing the first verification port and the second verification port is clear. In the monitoring process of electromechanical equipment, a large amount of operation data will be collected by the multidimensional sensor network. In order to ensure the quality of the data used for diagnosis and improve the accuracy and reliability of diagnosis, a specific verification and processing mechanism needs to be introduced.

[0039] In this application, two ports are used for processing in the data transmission process, which avoids the complexity of single-stage data processing affecting the continuity of the whole process and provides a coherent and standardized diagnostic architecture.

[0040] Among them, the first verification port is arranged at the transmission starting segment of the sensing side. The transmission starting segment refers to the initial segment position where the data collected from various sensors starts to be transmitted. Here, the multi-linear sensor data is aggregated and pre-processed, and the data collected from different types and locations of sensors is integrated and aggregated.

[0041] Specifically, it includes classification, statistics, and removal of some obvious errors or repeated data, making the data more orderly and refined to form a data set containing multiple information, preparing for subsequent processing and analysis.

[0042] Then, the second verification port is arranged at the transmission middle segment. The transmission middle segment refers to the intermediate position of the data from the sensing side to the data center. Because the data needs further format conversion and packaging after the preliminary processing by the first verification port, so as to facilitate subsequent data processing and analysis.

[0043] If all the data is processed in the first verification port, it cannot guarantee the continuity of the transmission of new and old sampling data, and the targeted port configuration can guarantee the accuracy of data processing.

[0044] The main function of the second verification port is to perform data vector packaging. That is, the data after preliminary processing is converted into a vector form of a specific format. For example, the second verification port will perform feature extraction and conversion on the data aggregated by the first verification port, and convert it into a series of vectors with specific dimensions and meanings. These vectors contain key information of the running state of electromechanical equipment and have a unified format, which is convenient for subsequent machine learning algorithms or diagnostic models to process.

[0045] By introducing the first verification port and the second verification port and arranging them at appropriate positions to perform corresponding multidimensional data aggregation and data vector packaging functions, a complete diagnostic network architecture is constructed, which can effectively process and optimize the data collected by the multidimensional sensor network of electromechanical equipment, and provide high-quality data support for subsequent fault diagnosis and operation and maintenance decision-making.

[0046] Further, the step S2 includes: connecting the multi-dimensional sensing network and the data center of the electromechanical diagnosis system through a transmission path, deploying a first verification port and a second verification port in the transmission path, and constructing a diagnosis network architecture;

[0047] The constructing the diagnosis network architecture includes: for the multi-dimensional sensing network, taking a sensing deployment position as a starting position and taking the data center of the electromechanical diagnosis system as a terminal position, building a transmission path; performing multi-thread aggregation of sensing sampling return, deploying the first verification port in the transmission path; and deploying the second verification port in the transmission path, wherein the second verification port is located behind the first verification port.

[0048] Firstly, the transmission path is built. For the multi-dimensional sensing network, it is a network system composed of a large number of sensors distributed in different positions of the electromechanical equipment and having different functions. These sensors are responsible for real-time collection of various data in the running process of the equipment. The sensing deployment position is taken as the starting position, which means that the data transmission starting point is planned from the actual installation position of each sensor. The data center of the electromechanical diagnosis system is the core hub of the entire diagnosis system, which is responsible for receiving, storing, analyzing and processing a large amount of data from the multi-dimensional sensing network to make accurate fault diagnosis and operation and maintenance decisions, so it is taken as the terminal position of the transmission path.

[0049] When building the transmission path, a variety of factors need to be considered, such as the stability of data transmission, transmission speed, anti-interference ability, etc.

[0050] Next, the first verification port is deployed through multi-thread aggregation of sensing sampling return. Each sensor in the multi-dimensional sensing network corresponds to a data thread, and the multi-thread aggregation is implemented, that is, the multi-dimensional data is aggregated and summarized. After the multi-thread data is aggregated and summarized according to the first verification port, single-thread transmission is determined for subsequent transmission. The aggregated data is preliminarily verified and processed. For example, the data is de-duplicated and filtered to remove some duplicate or incorrect data and improve the quality of the data.

[0051] Further, the second verification port is deployed in the transmission path. After the preliminary processing by the first verification port, the data is further processed for feature vectorization and encapsulation to ensure the convenience of subsequent diagnosis. Data vector encapsulation is to convert the preliminarily processed data into a vector form of a specific format, which contains key information of the running state of the electromechanical equipment and has a unified format, facilitating subsequent processing by machine learning algorithms or diagnosis models. In the entire transmission path, the second verification port plays a role of connecting the previous and the next, which further optimizes the quality and format of the data and provides strong support for accurate diagnosis of the data center.

[0052] Through the above steps, we successfully built a complete diagnostic network architecture, which can effectively and accurately transmit the data collected by the multi-dimensional sensor network to the data center of the mechanical and electrical diagnosis system on the basis of ensuring the efficiency of data flow, providing a solid foundation for fault diagnosis and operation and maintenance decision of mechanical and electrical equipment.

[0053] For example, the configuration of the first verification port and the second verification port can be oriented towards port-specific functions, trained to convergence using sample-driven training, and further deployed in the transmission path to ensure automatic operation and processing.

[0054] S3: Based on the diagnostic network architecture, the mechanical and electrical flow data is connected to the network, and through the execution of the aggregation screening based on the first verification port and the data structure conversion and packaging based on the second verification port, it is returned to the data center. Based on the built-in operation and maintenance diagnostic device, operation and maintenance fault decision and trend prediction are performed, and the fault diagnosis single is output.

[0055] First, based on the diagnostic network architecture, the mechanical and electrical flow data is connected to the network. The diagnostic network architecture is an organic whole composed of multi-dimensional sensor network, transmission path and first verification port and second verification port deployed therein, which provides a stable and efficient framework for the transmission and processing of mechanical and electrical flow data. As the key data reflecting the operation state of mechanical and electrical equipment, the mechanical and electrical flow data is connected to the network, that is, the mechanical and electrical flow data is connected to the transmission channel built by the diagnostic network architecture, so that it can be transmitted according to the established path and rules.

[0056] Subsequently, the aggregation screening based on the first verification port is performed. Aggregation screening refers to the aggregation and integration of scattered mechanical and electrical flow data from different sensing sources, and the screening and checking of the data to ensure the quality and effectiveness of the data. At the same time, according to the pre-set rules such as normal current and voltage range, the data is screened. For data that exceeds the normal range, further analysis is made to determine whether it is real abnormal data or false data caused by sensor failure or interference. If it is found that the current data transmitted by a sensor deviates significantly from the normal operating range and the duration is short, it may be a transient anomaly caused by electromagnetic interference, and the first verification port can correct or mark the data for subsequent processing.

[0057] Next, the data structure conversion and packaging based on the second verification port is performed, that is, the data after the aggregation screening by the first verification port is converted from its original collection format to a unified data structure more suitable for subsequent storage, transmission and analysis, and converted into the form of feature vector. Packaging is to add specific identification and description information to the data based on data structure conversion, so that it becomes a complete, easily identifiable and processed data unit.

[0058] After the above processing, it is returned to the data center. The data center, as the core of the entire mechanical and electrical equipment diagnosis system, has powerful data storage, calculation and analysis capabilities. It is responsible for receiving the processed mechanical and electrical flow data transmitted from the diagnosis network architecture and performing subsequent in-depth analysis. For example, the data center receives the mechanical and electrical data packets returned by the second verification port through a high-speed network interface, stores them in a special database, and waits for further processing and analysis.

[0059] Finally, based on the built-in operation and maintenance diagnostic device, operation and maintenance fault decision and trend prediction are performed, and a fault diagnosis sheet is output. The operation and maintenance diagnostic device is an intelligent diagnostic tool trained based on specific algorithms and models, which is built into the data center. It uses the received mechanical and electrical flow data, historical data and fault cases to evaluate and diagnose the running state of the mechanical and electrical equipment through machine learning, data analysis and other technical means, to judge whether the equipment is currently faulty and the type and severity of the fault, and then give corresponding treatment suggestions. For example, if the operation and maintenance diagnostic device analyzes the mechanical and electrical flow data and finds that the current fluctuation of the motor is abnormal and exceeds the normal range, combined with historical fault data, it is judged that the motor winding may have a short circuit fault, at which time the operation and maintenance diagnostic device will give the decision suggestion of stopping the equipment running, checking and repairing the motor winding.

[0060] At the same time, it makes trend prediction, that is, the operation and maintenance diagnostic device predicts the possible fault conditions of the equipment in the future period of time according to the trend of the mechanical and electrical flow data and the running history of the equipment. For example, through the trend analysis of the gradual rise of the motor current in a period of time, the operation and maintenance diagnostic device predicts that the motor may fail due to overload, and reminds the maintenance personnel to check the equipment and adjust the load in advance.

[0061] Finally, the operation and maintenance diagnostic device sorts out these diagnosis results and decision suggestions into a fault diagnosis sheet and outputs it, providing a strong basis for the maintenance and management of mechanical and electrical equipment. The fault diagnosis sheet usually contains detailed information such as equipment name, fault type, fault occurrence time, fault cause analysis and treatment suggestions, which facilitates maintenance personnel to quickly understand the equipment status and take corresponding measures.

[0062] Further, the aggregation screening based on the first verification port is performed, and the step S3 of the present application comprises:

[0063] The mechanical and electrical flow data is aggregated to the first verification port, data preprocessing and verification filtering are performed, and verification data is determined, wherein the data processing calibration or resampling compensation is used as the preprocessing method; the verification data is grouped based on the data correlation degree based on the fault characteristics, and the aggregated flow data is determined.

[0064] Specifically, the motor current data covers multi-dimensional sensing information during the operation of the motor, which is collected by sensors distributed throughout the device. Under the transmission system of the multi-dimensional sensing network, the data flows to the first verification port according to the predetermined path, that is, the motor current data flows to the first verification port.

[0065] Then, data preprocessing and verification filtering are carried out, aiming to improve data quality and provide reliable basis for subsequent analysis. There are two common ways in the data preprocessing link, that is, data processing calibration and resampling compensation. Data processing calibration is to correct the data with collection errors such as missing data by using interpolation and the like. For example, due to the zero drift of the sensor after long-term use, the collected current data may deviate, and data processing calibration can make it closer to the true value.

[0066] Resampling compensation is to carry out targeted sampling and add it into the motor current data for the problem of inconsistent sampling frequency in the data collection process. For example, if some motor current data sampling points are lost due to network fluctuations in a certain period of time, resampling compensation can be carried out according to the change trend of the data before and after the missing data points to fill in the missing data points, so that the data sequence is complete and smooth.

[0067] After data preprocessing, verification filtering is carried out. That is, the preprocessed data is screened according to certain rules, that is, it is preliminarily determined whether the data is a data value representing a fault of the motor. Specifically, by setting a threshold according to the data critical value of the fault, if the threshold is met, it indicates that there is a potential risk of fault in the current data state, and the data is filtered and returned for diagnosis and decision; if the current data state represents a normal operation state, it is intercepted and does not participate in the subsequent diagnosis and decision.

[0068] After completing the data verification, the verification data is grouped based on the data correlation degree of the fault characteristics, and then the aggregated flow data is determined. The data correlation degree represents the correlation degree between different fault characteristics. For example, for a fault caused by motor winding short circuit, the fault characteristic is that the current increases sharply and the voltage decreases. When grouping the verification data, those data with increased current and decreased voltage are grouped together, and the correlation degree of these data with the winding short circuit fault characteristic is higher.

[0069] Through this grouping method based on the data correlation degree of the fault characteristics, the verification data is divided into different sets, and each set is an aggregated flow data, so as to carry out in-depth analysis and processing for different types of data respectively in the subsequent process, and improve the accuracy and efficiency of fault diagnosis.

[0070] Further, verification filtering is carried out to determine the verification data, and step S3 of the present application comprises:

[0071] With the fault characteristics, a data barrier based on rigid data threshold is set, wherein the non-fault data threshold is set; according to the data barrier, the preprocessed data is screened, and the verification data is determined, wherein the data part not meeting the rigid data threshold is passed, and the data part meeting the rigid data threshold is intercepted.

[0072] Firstly, with the fault characteristics, a data barrier based on rigid data threshold is set. The rigid data threshold is a numerical standard with clear limits determined based on the data range of the normal operation of the equipment. The non-fault data threshold is set, that is, a reasonable critical value is determined according to the fluctuation range of various data under the normal operation condition of the equipment, that is, as long as it is within this critical value, there is no probability of mechanical and electrical operation failure. For example, for a motor with a rated working current of 5-8 amperes, after a large number of normal operation data statistical analysis, it is determined that the current fluctuation range during normal operation is between 6-7.5 amperes, so 6 amperes can be set as the lower limit rigid data threshold, and 7.5 amperes can be set as the upper limit rigid data threshold, to build a barrier for motor current data.

[0073] Next, after completing the setting of the rigid data threshold barrier, it is applied to the mechanical and electrical data after preprocessing such as data processing calibration or resampling compensation. The data screening process is like a strict quality detection process, which checks the preprocessed data one by one. For example, if the current data of the motor is being processed, the current value collected at a certain moment is 5.5 amperes, which is lower than the set lower limit rigid data threshold of 6 amperes, and belongs to the data part that does not meet the rigid data threshold. According to the set rule, this part of data will be judged as data with fault risk, and will be passed, that is, it is allowed to enter the subsequent further analysis link, because the data may imply that the motor has potential operation problems and needs to be further explored. Conversely, if the collected current value is 7 amperes, it is within the rigid data threshold range of 6-7.5 amperes, that is, it meets the rigid data threshold, and this part of data is considered as normal data and is intercepted, and does not enter the subsequent deep analysis process for abnormal data, because it meets the data characteristics under the normal operation state of the equipment.

[0074] Through careful screening operation, the verification data that may be related to fault can be accurately screened from a large amount of preprocessed data, providing a reliable data basis for subsequent in-depth analysis and accurate diagnosis based on fault characteristics, effectively improving the efficiency and accuracy of fault diagnosis, and avoiding the influence on the judgment of the real operation state of the equipment due to the interference of a large amount of normal data.

[0075] Further, the data structure conversion and encapsulation based on the second verification port are performed. The step S3 includes: determining a unified data structure; converting the aggregated flow data stream to the second verification port, identifying feature states, and converting the feature states based on the unified data structure; and performing encapsulation and routing coupling based on the feature vectors for the converted feature states to determine encapsulated data.

[0076] First, a unified data structure is determined. In the process of running the electromechanical equipment, the initial formats of the electromechanical flow data collected by the multi-dimensional sensor network are various, and the sources thereof include sensors of different types and different manufacturers, and each sensor generates data of a unique format according to the characteristics of the physical quantity monitored by the sensor. For example, a current sensor outputs a numerical sequence reflecting the change of the current intensity with time, and a vibration sensor outputs complex waveform data related to the vibration amplitude and frequency of the equipment.

[0077] In order to facilitate the unified processing, storage and transmission of subsequent data, a unified data structure needs to be determined. This structure specifies the organization mode of the data, the types and mutual relationships of the data elements, etc.

[0078] Then, the aggregated flow data, that is, the data obtained by preprocessing and verifying the data passing through the first verification port and grouping the data according to the data correlation degree of the fault characteristics, is aggregated to the second verification port. These data flow along the transmission path of the diagnostic network architecture to the second verification port located in the middle of the transmission.

[0079] After reaching the second verification port, the feature states, that is, various state characteristics of the running electromechanical equipment reflected by the aggregated flow data, including the numerical values of the equipment running parameters, the change trend, etc. are identified. For example, the current value in the electromechanical flow data, the stability of the voltage, the frequency of the current fluctuation, etc. belong to the feature states. Identifying the feature states is to extract these key feature information from the aggregated flow data.

[0080] Then, based on the determined unified data structure, the identified feature states are converted, and the converted data is reorganized to obtain converted feature states conforming to the unified data structure.

[0081] Further, encapsulation and routing coupling based on the feature vectors are performed for the converted feature states. Specifically, the feature vectors are quantized and vectorized representations of the key feature information in the converted feature states, so as to facilitate efficient processing and analysis by the computer. For example, the current value, the voltage value, the vibration amplitude value, etc. of the equipment are quantized as a multi-dimensional vector, and each dimension corresponds to a feature value. The encapsulation is to add necessary identification and description information to the feature vectors, so that they become an independent and complete data unit. For example, a time stamp is added to the feature vector to indicate the running time of the equipment corresponding to the data; and a device number is added to indicate the equipment to which the data belongs.

[0082] The multiple encapsulated units of the encapsulated data vector are coupled and routed. A unified data system is determined as the encapsulated data, which is then efficiently and accurately transmitted to the data center in the diagnostic network architecture, providing reliable data support for subsequent equipment fault diagnosis and operation and maintenance decision-making.

[0083] Further, based on the feature vector, the encapsulation and routing coupling of the converted feature state are performed to determine the encapsulated data. Step S3 of the present application comprises:

[0084] The converted feature state is traversed to determine the feature vectors and determine multiple feature layers according to the feature progression relationship. Each feature vector is independently encapsulated. The feature transmission relationship between the multiple feature layers is determined to determine the directed transmission mode and perform routing coupling to determine the encapsulated data.

[0085] First, the converted feature state, i.e., the data after the unified data structure conversion, is traversed. Then, the feature vectors are determined and the multiple feature layers are determined according to the feature progression relationship. The feature vector is a quantized and vectorized representation of each feature of the equipment operation, so that the computer can more efficiently process and analyze these data. Determining the feature vector is to extract representative features from the converted feature state and convert them into vector form. For example, from the operation data of the motor, the current value, voltage value, and vibration amplitude are extracted to form a three-dimensional feature vector, and each dimension corresponds to a feature value.

[0086] The feature progression relationship refers to the logical sequence, causal relationship, difference in importance, and other data logical relationships between different features. For example, for motor fault diagnosis, the change in current value may be the first feature to reflect the abnormal operation of the motor, followed by the fluctuation of the voltage value, and then the increase in vibration amplitude, which forms a feature progression relationship.

[0087] According to the feature progression relationship, multiple feature layers are determined. Features closely related to motor operation and playing a key role in fault diagnosis are divided into a feature layer, such as current value, voltage value, and other features directly reflecting the electrical performance of the motor are classified as an electrical feature layer; vibration amplitude, temperature, and other features reflecting the mechanical state of the motor are classified as a mechanical feature layer. In this process, each feature vector is independently encapsulated, i.e., each feature vector is processed individually, and necessary identification and description information is added to make it an independent data unit. For example, for the current value feature vector, a time stamp is added to indicate the time corresponding to the current value, and a device number is added to specify the motor to which the data belongs.

[0088] Finally, the feature transmission relationship between the plurality of feature layers is used to determine a directed transmission mode and perform routing coupling. The routing coupling is the directed transmission mode, which is used to associate and couple different packaging units to obtain a standardized and data structure intuitive data system as the packaging data. The feature transmission relationship between the plurality of feature layers describes the flow direction and mutual influence of information between different feature layers. For example, in the operation of a motor, the change of current and voltage values of the electrical feature layer can cause the vibration amplitude and temperature change of the mechanical feature layer. Determining the directed transmission mode is to determine the transmission direction of data between different feature layers according to the feature transmission relationship. For example, it is specified that data is transmitted from the electrical feature layer to the mechanical feature layer, and then to other related feature layers. For example, how to transmit from the lower layer to the upper layer. In the case of multi-source data, the lower layer may be the cause and the upper layer may be the result. The related packaging units of the lower layer are weighted and identified to obtain the upper layer.

[0089] The packaging data not only contains the key feature information of the operation of the electromechanical equipment, but also has the necessary properties for accurate transmission in the network, thereby providing reliable data support for the equipment fault diagnosis and operation and maintenance decision of the data center.

[0090] Further, the fault diagnosis sheet is output, and step S3 of the present application includes:

[0091] The historical fault record is called to adjust the historical fault record based on the data vector packaging architecture to determine sample data, wherein the sample data includes packaging sample-diagnosis sample-prediction sample. According to the sample data, the operation and maintenance diagnostic apparatus is constructed by supervised training to convergence and is built in the data center. The operation and maintenance diagnostic apparatus receives the packaging data and decides to output the fault diagnosis sheet. The fault diagnosis sheet is displayed and warned on the terminal, and the source sensor is located, and the reverse transmission and the on-site warning based on the source sensor are performed in combination with the transmission path.

[0092] Specifically, the historical fault record is called, which is a collection of fault related information of the past operation of the electromechanical equipment, and includes the occurrence time, phenomenon, corresponding operating parameters and solving measures of the equipment fault and the like. The historical fault record is adjusted based on the data vector packaging architecture. The architecture organizes and represents data according to specific rules, and converts complex fault information into a vector form that is easy for a computer to process. In this process, the information of the historical fault record is filtered, extracted and converted, such as the operating parameters of current, voltage and temperature closely related to fault diagnosis, which are packaged into a vector in order and format, so as to convert the historical fault record into sample data.

[0093] The sample data includes encapsulated samples, diagnostic samples, and prediction samples. The encapsulated samples are historical fault records processed by data vector encapsulation, storing the key information of equipment faults in vector form. The diagnostic samples are data annotated with fault types and diagnostic results based on the encapsulated samples, providing clear targets for supervised training. The prediction samples are used for performance verification of the operation and maintenance diagnostic device and prediction of future faults.

[0094] Through the above sample processing, the data system suitable for the technical architecture is converted to ensure that the subsequent training results are highly consistent with the technical solution.

[0095] According to the sample data, the model is trained using the diagnostic samples with labeled information through a supervised training method, so that the model learns the mapping relationship between fault features and fault types. During training, the model continuously adjusts parameters to minimize the error between the prediction results and the labeled information. When the error reaches a minimum value and no longer changes significantly, the training converges. For example, using a neural network model to train sample data, the output of the encapsulated sample is predicted for fault types, and the labeled information of the diagnostic sample is compared. The weights and biases are adjusted through the backpropagation algorithm until the error meets the preset convergence condition. The trained operation and maintenance diagnostic device is embedded in the data center, which is the core hub of the mechanical and electrical equipment diagnostic system and has strong storage and computing capabilities. It provides a stable operating environment and efficient data processing support for the operation and maintenance diagnostic device, enabling it to receive real-time encapsulated data from multi-dimensional sensor networks and quickly and accurately diagnose faults.

[0096] The operation and maintenance diagnostic device receives encapsulated data of real-time operation of mechanical and electrical equipment processed by the second verification port and organized according to the data vector encapsulation architecture. Based on the learned mapping relationship between fault features and fault types, it analyzes and judges the current operating state and possible fault types of the equipment. For example, if encapsulated data of a motor is received, it is found that the current value is abnormally high and the temperature is outside the normal range, indicating that the motor may be overloaded, and a fault diagnosis sheet containing information such as fault type, severity, and possible causes is generated.

[0097] Furthermore, the fault diagnosis sheet is displayed to the operation and maintenance personnel through terminal devices such as display screens to achieve terminal display warning, and the source sensor of the abnormal data is located to determine the specific location of the fault. For example, by analyzing the data source of the fault diagnosis sheet, it is determined that the temperature sensor near the motor winding collected abnormal high temperature data, and the fault is located at the motor winding. The fault diagnosis information is transmitted to the source sensor location in the reverse direction through data transmission to achieve reverse transmission, with the purpose of timely warning on site. Based on the source sensor, on-site warning is achieved through sound and light alarms, indicator lights, etc.

[0098] For example, after the operation and maintenance diagnostic device judges the motor fault, the fault information is transmitted in reverse through the transmission path to the source sensor near the motor, and the source sensor triggers the sound and light alarm to remind the on-site staff to handle the fault.

[0099] Through the above process, the historical fault records are used to build an operation and maintenance diagnostic device for real-time operation data fault diagnosis and timely warning of electromechanical equipment, which can effectively improve the reliability and operation efficiency of the equipment and reduce the loss caused by equipment failure.

[0100] The electromechanical equipment diagnosis method of the multi-dimensional operation and maintenance data training model provided in the present application has the following technical effects: 1. A variety of types of sensors are reasonably deployed on the electromechanical equipment, and cross-time sampling is performed according to a preset sampling frequency and rules, so that the operation data of the equipment are continuously and real-timely collected, ensuring the timeliness and accuracy of the data. At the same time, a field early warning mechanism is introduced, and early warning based on the corresponding sensors is performed on the diagnosis information, so that the field fault area can be quickly locked. 2. A first verification port is introduced to preliminarily correct and barrier screen the data, so as to protect the effective data related to potential faults. A second verification port is introduced to perform unified format conversion and feature vector packaging, so as to improve the systematicness and intuitiveness of the data. At the same time, based on the processed data features, an operation and maintenance diagnostic device is constructed and deployed in the data center. A systematic diagnosis network architecture is formed, which can adapt to continuous sampling and diagnosis in complex scenarios.

[0101] Embodiment two:

[0102] Based on the same inventive concept as the electromechanical equipment diagnosis method of the multi-dimensional operation and maintenance data training model in the foregoing embodiments, the present application provides an operation and maintenance device, which is composed of a multi-dimensional sensor network, a diagnosis network architecture and an operation and maintenance diagnostic device; wherein the multi-dimensional sensor network performs front-end sampling and field early warning, and the diagnosis network architecture and the operation and maintenance diagnostic device perform data interaction and processing, cooperatively forming a diagnosis system of electromechanical equipment.

[0103] Specifically, the multi-dimensional sensor network undertakes the key responsibilities of front-end sampling and field early warning in the operation and maintenance device. In terms of front-end sampling, it widely and accurately collects various data in the operation process of electromechanical equipment through various types of sensors, such as temperature sensors, pressure sensors, current sensors, etc. These data cover multiple dimensions of information such as temperature, pressure, current and vibration of the equipment, providing an original data basis for subsequent equipment state analysis. At the same time, in the field early warning, once the abnormal data reaches the preset early warning threshold, the multi-dimensional sensor network can immediately start the field early warning mechanism, and directly alarms the staff on the equipment site through sound and light alarm, indicator light flickering and other ways, reminding them to handle potential problems in time.

[0104] The multi-dimensional sensor network is connected to the diagnosis network architecture through wired or wireless transmission links. The data collected by the sensors are continuously transmitted to the first verification port of the diagnosis network architecture through these transmission links, starting the subsequent data processing process.

[0105] Various sensors work based on different physical principles, periodically collect equipment operation data according to the established sampling frequency and rules, and the network architecture is mainly responsible for data interaction and processing.

[0106] On the one hand, it receives raw data from the multi-dimensional sensor network, aggregates multi-dimensional data through the first verification port, integrates data collected by sensors of different types and different positions together, and performs preliminary data preprocessing and verification filtering to improve data quality. On the other hand, through the second verification port, the data vector encapsulation is performed on the data processed by the first verification port, the data is converted into a format suitable for subsequent analysis and transmission, and then the processed data is transmitted to the operation and maintenance diagnostic device. At the same time, it also receives the control information or instructions fed back by the operation and maintenance diagnostic device, and realizes the bidirectional interaction of data.

[0107] The operation and maintenance diagnostic device is the core decision-making component of the entire diagnostic system. It receives the encapsulated data processed by the diagnostic network architecture, analyzes the data in depth, judges the current running state of the electromechanical equipment, identifies whether there is a fault and the type and severity of the fault. At the same time, based on historical fault records and real-time data, it predicts the future running trend of the equipment, and finally outputs the fault diagnosis sheet, providing decision basis for the maintenance and management of the equipment.

[0108] In the overall operation logic, the multi-dimensional sensor network first collects data from the electromechanical equipment in all directions at the front end, and transmits the collected raw data to the diagnostic network architecture. The diagnostic network architecture aggregates, preprocesses, verifies and filters the data, and encapsulates the data vector, etc. A series of processes are performed to transmit high-quality data to the operation and maintenance diagnostic device. The operation and maintenance diagnostic device analyzes the data using the built-in diagnostic model, judges the equipment running state, outputs the fault diagnosis sheet, and feeds back the diagnostic results to the diagnostic network architecture and terminal equipment. Form a closed loop, efficient electromechanical equipment diagnostic system, to ensure the stable operation and timely maintenance of the equipment.

[0109] Through the foregoing detailed description of the electromechanical equipment diagnostic method of the multi-dimensional operation and maintenance data training model, those skilled in the art can clearly understand the electromechanical equipment diagnostic method of the multi-dimensional operation and maintenance data training model and the operation and maintenance equipment in the embodiment. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the related part is referred to the method part description.

[0110] The foregoing description of the disclosed embodiments enables a person skilled in the art to make or use the application. Modifications of these embodiments will occur to persons of skill in the art, and that the generic principles defined herein can be applied to other embodiments without departing from the spirit or scope of the application. Therefore, the present application is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for diagnosing electromechanical equipment using a multi-dimensional operation and maintenance data training model, characterized in that, The method includes: A multi-dimensional sensor network is deployed in the electromechanical equipment area, and the electromechanical current data is obtained by using a staggered cross sampling method. The electromechanical current data is the spatiotemporal flow data of the electromechanical equipment. A diagnostic network architecture is constructed by introducing a first verification port and a second verification port. The first verification port is deployed at the transmission start segment on the sensing side to perform multi-dimensional data aggregation, and the second verification port is deployed in the transmission middle segment to perform data vector encapsulation. Based on the diagnostic network architecture, the machine current data is entered into the network. Through the aggregation screening based on the first verification port and the data structure conversion and encapsulation based on the second verification port, it is sent back to the data center. Based on the built-in operation and maintenance diagnostic tool, operation and maintenance fault decision-making and trend prediction are performed, and a fault diagnosis report is output. Among them, the data center of the multi-dimensional sensor network and the electromechanical diagnostic system are connected by a transmission path, and a first verification port and a second verification port are deployed on the transmission path to build a diagnostic network architecture. The construction of the diagnostic network architecture includes: For the aforementioned multidimensional sensor network, a transmission path is established with the sensor deployment location as the starting point and the data center of the electromechanical diagnostic system as the ending point. Multi-threaded aggregation for sensor sampling and backhaul is performed, and a first verification port is deployed in the transmission path; A second verification port is deployed in the transmission path, wherein the second verification port is placed after the first verification port; The encapsulation of the data structure after conversion based on the second verification port includes: A unified data structure is used to transform the aggregated stream data after the first verification and determine the transformation characteristic state; For the aforementioned transformation feature state, feature vectors are decomposed and determined, and multiple feature layers are determined according to the feature progression relationship, wherein each feature vector is independently encapsulated; Based on the feature transmission relationships between the multiple feature layers, the directed transmission method is determined and routing coupling is performed to determine the encapsulated data.

2. The electromechanical equipment diagnosis method based on a multi-dimensional operation and maintenance data training model as described in claim 1, characterized in that, The staggered cross-sampling method is adopted, including: For the multi-dimensional sensor sources in the electromechanical equipment area, data characteristics are determined, wherein the sensor source type and deployment location are identified; The historical fault records of the electromechanical equipment area are retrieved to analyze the fault characteristics, in which the fault source is identified. Configure the sampling mode based on the data characteristics and the fault characteristics; The data characteristics include linear and nonlinear types, and the fault characteristics include periodic stationary faults, periodic non-stationary faults, non-periodic stationary faults, and non-periodic non-stationary faults.

3. The electromechanical equipment diagnosis method based on a multi-dimensional operation and maintenance data training model as described in claim 2, characterized in that, Configure sampling modes, including: Combine data characteristics and fault characteristics to determine a first sampling class, wherein the first sampling class is any combination of characteristics; For the first sampling class, locate the first sensing part and set the first sampling conditions, wherein the sampling conditions include at least the sampling frequency; Identify the first sampling condition, and cross-merge each sampling condition using the sampling time series to determine the sampling pattern; The data acquisition of the multidimensional sensor network is driven according to the sampling mode.

4. The electromechanical equipment diagnosis method based on a multi-dimensional operation and maintenance data training model as described in claim 1, characterized in that, Perform aggregated screening based on the first verification port, including: The machine current data is transferred and aggregated to the first verification port for data preprocessing and verification filtering to determine the verification data. The preprocessing method is data processing calibration or resampling compensation. The verification data is grouped based on the data correlation degree based on fault characteristics to determine the aggregated stream data.

5. The electromechanical equipment diagnosis method based on a multi-dimensional operation and maintenance data training model as described in claim 4, characterized in that, Perform verification filtering to determine the verification data, including: Based on the aforementioned fault characteristics, a data gate is set based on a rigid data threshold, wherein a non-faulty data threshold is used for setting. Based on the data checkpoint, the preprocessed data is screened to determine the verification data, wherein data portions that do not meet the rigid data threshold are passed, and data portions that meet the rigid data threshold are blocked.

6. The electromechanical equipment diagnosis method based on a multi-dimensional operation and maintenance data training model as described in claim 1, characterized in that, Output a fault diagnosis report, including: Historical fault records are retrieved and adjusted using a data vector encapsulation-based architecture to determine sample data, wherein the sample data includes encapsulated samples, diagnostic samples, and predictive samples. Based on the sample data, the operation and maintenance diagnostic tool is constructed and built into the data center through supervised training until convergence. The maintenance diagnostic tool receives the encapsulated data and outputs the fault diagnosis report; The fault diagnosis report is displayed on the terminal for early warning, and the source sensor is located. Reverse transmission and on-site early warning based on the source sensor are performed in combination with the transmission path.

7. A maintenance equipment, characterized in that, A diagnostic method for electromechanical equipment for executing a multi-dimensional operation and maintenance data training model as described in claims 1-6, wherein the operation and maintenance equipment is composed of a multi-dimensional sensor network, a diagnostic network architecture, and an operation and maintenance diagnostic tool. The multi-dimensional sensor network performs front-end sampling and on-site early warning, while the diagnostic network architecture and operation and maintenance diagnostic tool perform data interaction and processing, together forming a diagnostic system for electromechanical equipment.

Images