State monitoring method, apparatus and system, storage medium and product

Abstract

The present disclosure relates to a state monitoring method, apparatus and system, a storage medium and a product, and relates to the field of engineering machinery. The state monitoring method comprises: on the basis of state data of a plurality of first devices, generating a training sample set, the plurality of first devices having a same model, and each first device being a construction machine; on the basis of types of samples in the training sample set, determining a diagnostic model; by using the training sample set, training the diagnostic model, so as to determine values of parameters of the diagnostic model; and sending the diagnostic model and the values of the parameters thereof to a plurality of second devices, so that each of the plurality of second devices can determine, by using the diagnostic model and the parameters thereof, whether its own state has had an anomaly, each second device being a construction machine of the same model as the first devices.

Description

State monitoring method, device, system, storage medium and product

[0001] Cross-reference to Related Applications

[0002] This application is based on and claims priority to the application with the Chinese application number 202410635740.6 and the filing date of May 21, 2024, the disclosure of which is hereby incorporated by reference in its entirety into the present application. TECHNICAL FIELD

[0003] The present disclosure relates to the field of construction machinery, and in particular, to a state monitoring method, device, system, storage medium and product. BACKGROUND

[0004] Remote monitoring of the state of construction machinery equipment is an important means to achieve comprehensive monitoring and management of construction machinery equipment. State monitoring of construction machinery equipment not only enables monitoring of the running state of construction machinery equipment, but also enables monitoring of abnormal states of construction machinery equipment. Abnormal state monitoring is based on analysis of working condition data during long-term operation of the equipment, establishment of a diagnostic model and identification of possible abnormal types, thereby enabling targeted maintenance of the equipment. SUMMARY

[0005] According to a first aspect of some embodiments of the present disclosure, a state monitoring method is provided, comprising: generating a training sample set based on state data of a plurality of first devices, wherein the plurality of first devices are of the same model, and the first devices are construction machinery equipment; determining a diagnostic model based on the types of the samples in the training sample set; training the diagnostic model using the training sample set to determine the values of the parameters of the diagnostic model; and sending the diagnostic model and the values of the parameters to a plurality of second devices, so that each second device in the plurality of second devices determines whether its own state is abnormal using the diagnostic model and the parameters, wherein the second devices are construction machinery equipment of the same model as the first devices.

[0006] In some embodiments, the state monitoring method further comprises: generating one or more types of abnormal samples based on alarm data and supplementing the abnormal samples to the training sample set to update the training sample set, wherein the alarm data is generated in response to one or more of the plurality of second devices being abnormal; re-determining the diagnostic model based on the types of the samples in the updated training sample set; training the re-determined diagnostic model using the updated training sample set to determine the values of the parameters of the re-determined diagnostic model; and sending the re-determined diagnostic model and the parameters to the plurality of second devices, so that each second device in the plurality of second devices re-determines whether its own state is abnormal using the re-determined diagnostic model and the parameters.

[0007] In some embodiments, the re-determining the diagnostic model based on the types of the samples in the updated training sample set comprises: in response to a number of at least one type of abnormal samples in the one or more types of abnormal samples being greater than a first threshold, re-determining the diagnostic model based on the types of the samples in the updated training sample set.

[0008] In some embodiments, in a case where the training sample set only includes normal samples, the diagnostic model is a single Gaussian model, and the parameters of the diagnostic model include the second threshold, a mean value of the single Gaussian model, and a covariance matrix.

[0009] In some embodiments, in a case where the training sample set includes normal samples and abnormal samples, the diagnostic model is a Gaussian mixture model, and the parameters of the diagnostic model include the second threshold, a weight, a mean value, and a covariance matrix of each single Gaussian model included in the Gaussian mixture model.

[0010] In some embodiments, the generating the training sample set based on the state data of the plurality of first devices comprises: selecting a plurality of feature parameters from the state data of the plurality of first devices; and generating the training sample set based on the state data corresponding to the plurality of feature parameters.

[0011] In some embodiments, the sending the diagnostic model and the values of the parameters of the diagnostic model to the plurality of second devices comprises: sending the diagnostic model and the values of the parameters of the diagnostic model to a server, so that the server forwards the diagnostic model and the values of the parameters of the diagnostic model to the plurality of second devices based on an MQTT protocol.

[0012] According to a second aspect of some embodiments of the present disclosure, a state monitoring apparatus is provided, comprising: a first generation module configured to generate a training sample set based on state data of a plurality of first devices, wherein the plurality of first devices are of the same type, and the first devices are engineering machinery devices; a first determination module configured to determine a diagnostic model based on types of samples in the training sample set; a first training module configured to train the diagnostic model using the training sample set to determine values of parameters of the diagnostic model; and a first sending module configured to send the diagnostic model and the values of the parameters of the diagnostic model to a plurality of second devices, so that each second device in the plurality of second devices determines whether an abnormality exists in a state of the second device using the diagnostic model and the parameters of the diagnostic model, wherein the second devices are engineering machinery devices of the same type as the first devices.

[0013] In some embodiments, the state monitoring apparatus further comprises: a second generation module configured to generate one or more types of abnormal samples based on alarm data and supplement the abnormal samples to the training sample set to update the training sample set, wherein the alarm data is generated in response to one or more of the plurality of second devices having an abnormality; a second determination module configured to re-determine the diagnostic model based on the types of the samples in the updated training sample set; a second training module configured to train the re-determined diagnostic model using the updated training sample set to determine values of parameters of the re-determined diagnostic model; and a second sending module configured to send the re-determined diagnostic model and the parameters to each of the plurality of second devices, so that each of the plurality of second devices re-determines whether its own state has an abnormality using the re-determined diagnostic model and the parameters.

[0014] According to a third aspect of some embodiments of the present disclosure, a state monitoring system is provided, comprising: a state monitoring apparatus as described above; and a second device configured to receive the diagnostic model and the parameters sent by the state monitoring apparatus, and determine whether its own state has an abnormality using the diagnostic model and the parameters.

[0015] In some embodiments, the second device is configured to: generate a plurality of test samples based on the state data of the second device; input the plurality of test samples into the diagnostic model to determine a probability density value corresponding to each of the plurality of test samples; and determine that the state of the second device has an unknown abnormality in a case where the probability density value corresponding to each of the plurality of test samples is less than the second threshold value.

[0016] In some embodiments, the second device is configured to: generate first alarm data based on the state data of the second device in a case where it is determined that the second device has an unknown abnormality.

[0017] In some embodiments, the second device is configured to: determine that the state of the second device is normal or has a known abnormality in a case where there is at least one test sample in the plurality of test samples whose corresponding probability density value is not less than the second threshold value.

[0018] In some embodiments, in a case where the trained model is a single Gaussian model, the second device is configured to: determine that the state of the second device is normal in a case where there is at least one test sample in the plurality of test samples whose corresponding probability density value is not less than the second threshold value.

[0019] In some embodiments, in the case where the diagnostic model is a Gaussian mixture model, the second device is configured to: input the plurality of test samples into each single Gaussian model included in the Gaussian mixture model, and determine a state monitoring result and a probability density value corresponding to each single Gaussian model, wherein for each single Gaussian model, the state monitoring result corresponding to the single Gaussian model is normal or there is a known abnormality and indicates a type of the known abnormality corresponding to the single Gaussian model; and determine the state monitoring result corresponding to the second device according to the probability density value corresponding to each single Gaussian model.

[0020] In some embodiments, in the case where the diagnostic model is a Gaussian mixture model, the second device is configured to: determine the state monitoring result of the single Gaussian model corresponding to the maximum value in the probability density values corresponding to each single Gaussian model as the state monitoring result corresponding to the second device.

[0021] In some embodiments, in the case where the diagnostic model is a Gaussian mixture model, the second device is configured to: in the case where the second device has a known abnormality, generate second alarm data based on the state data of the second device.

[0022] In some embodiments, the state monitoring system further comprises: a server configured to receive the diagnostic model and its parameters sent by the state monitoring device, and forward the diagnostic model and its parameters to the plurality of second devices based on the MQTT protocol.

[0023] In some embodiments, the state monitoring system further comprises: a monitoring device configured to receive the data sent by the second device, and forward the data sent by the second device to the state monitoring device.

[0024] In some embodiments, the server is configured to: receive the data sent by the second device based on the MQTT protocol, and forward the data sent by the second device to the monitoring device.

[0025] In some embodiments, the second device is configured to: send maintenance data to the server, wherein the maintenance data is generated in response to maintenance of the second device according to the alarm data of the second device and completion.

[0026] In some embodiments, the server is configured to: receive the data sent by the monitoring device based on the MQTT protocol, and forward the data sent by the monitoring device to one or more second devices according to the message identifier of the data sent by the monitoring device, so as to configure or control the one or more second devices by the monitoring device.

[0027] In some embodiments, the data sent by at least one of the second device and the monitoring device has a message identifier, and the message identifier comprises a plurality of hierarchical sub-identifiers.

[0028] In some embodiments, the sub-identities of the multiple levels include any one of a type of the device, a model of the device, a customer, a number of the device, a subject of the message.

[0029] In some embodiments, the subject of the message included in the message identifier of the data sent by the second device is any one of a status, an alarm, a maintenance, and the subject of the information included in the message identifier of the data sent by the monitoring device is any one of a control, a configuration.

[0030] In some embodiments, the server is configured to: receive data sent by a plurality of second devices; determine, according to the subscription information of the monitoring device, data to which the monitoring device subscribes from the data sent by the plurality of second devices; and send the data to which the monitoring device subscribes to the monitoring device, so that the monitoring device monitors the subscribed device.

[0031] In some embodiments, the server is configured to: determine a message identifier matching the subscription information of the monitoring device; and send data corresponding to the message identifier to the monitoring device.

[0032] In some embodiments, the server is configured to: for any level of the multiple levels, if the sub-identity of the level of the message identifier matches the value of the level of the subscription information of the monitoring device, determine that the message identifier matches the subscription information of the monitoring device.

[0033] In some embodiments, the value of one or more levels in the subscription information is a first wildcard, and for any level of the multiple levels, if the value of the level in the subscription information is the first wildcard, the value of the level in the subscription information matches any sub-identity.

[0034] In some embodiments, the last level in the subscription information is a second wildcard, wherein the second wildcard indicates that the level at which the second wildcard is located in the subscription information and the sub-level of the second wildcard match any sub-identity.

[0035] In some embodiments, the server is configured to: determine one or more second devices matching the message identifier of the data sent by the monitoring device; and send the data sent by the monitoring device to the one or more second devices.

[0036] According to a fourth aspect of some embodiments of the present disclosure, there is provided a state monitoring apparatus, comprising: a processor; and a memory coupled to the processor, for storing instructions, which, when executed by the processor, cause the processor to perform the state monitoring method as described above.

[0037] Other features and advantages of the present disclosure will become apparent from the following detailed description of exemplary embodiments thereof, taken in conjunction with the accompanying drawings. BRIEF DESCRIPTION OF DRAWINGS

[0038] In order to more clearly illustrate the technical solutions in the embodiments of the present disclosure or the related art, the following will briefly introduce the drawings needed to be used in the embodiments or the related art description. Obviously, the drawings in the following description only constitute some embodiments of the present disclosure, and for those skilled in the art, other drawings can also be obtained based on these drawings without creative labor.

[0039] FIG. 1 shows a flowchart of a state monitoring method according to some embodiments of the present disclosure.

[0040] FIG. 2 shows a schematic diagram of a diagnostic boundary of a single Gaussian model according to some embodiments of the present disclosure.

[0041] FIG. 3A shows a schematic diagram of a diagnostic boundary of a second-order Gaussian mixture model according to some embodiments of the present disclosure.

[0042] FIG. 3B shows a schematic diagram of a diagnostic boundary of a third-order Gaussian mixture model according to some embodiments of the present disclosure.

[0043] FIG. 4 shows a structural schematic diagram of a state detection apparatus according to some embodiments of the present disclosure.

[0044] FIG. 5 shows a structural schematic diagram of a state monitoring system according to some embodiments of the present disclosure.

[0045] FIG. 6 shows a schematic diagram of a subject structure of message identification according to some embodiments of the present disclosure.

[0046] FIG. 7 shows a schematic diagram of a state monitoring system according to some other embodiments of the present disclosure.

[0047] FIG. 8 shows a flowchart of state monitoring by a monitoring device according to some embodiments of the present disclosure.

[0048] FIGS. 9A-9C show schematic diagrams of state monitoring of a device by a monitoring device according to some embodiments of the present disclosure.

[0049] FIGS. 10A-10B show schematic diagrams of configuration of a device by a monitoring device according to some embodiments of the present disclosure.

[0050] FIG. 11 shows a flowchart of a state monitoring method according to some embodiments of the present disclosure.

[0051] FIG. 12 shows a sample probability density of a construction machine device in a normal state according to some embodiments of the present disclosure.

[0052] FIG. 13 shows a sample probability density of a construction machine device in an abnormal state according to some embodiments of the present disclosure.

[0053] FIG. 14 shows a structural schematic diagram of a state monitoring device according to some embodiments of the present disclosure.

[0054] FIG. 15 shows a structural schematic diagram of a state monitoring device according to some other embodiments of the present disclosure. DETAILED DESCRIPTION

[0055] The technical solutions in the embodiments of the present disclosure will be clearly and completely described below with reference to the drawings in the embodiments of the present disclosure. Obviously, the described embodiments are only some of the embodiments of the present disclosure, but not all the embodiments. The following description of at least one exemplary embodiment is actually only illustrative, but not as any limitation on the present disclosure and its application or use. Based on the embodiments in the present disclosure, all other embodiments obtained by those of ordinary skill in the art without creative labor fall within the scope of the present disclosure.

[0056] One technical problem to be solved by the present disclosure is how to improve the efficiency and accuracy of state monitoring of engineering machinery equipment.

[0057] The state monitoring of engineering machinery mainly includes two aspects of running state monitoring and abnormal state monitoring. Running state monitoring aims to collect key operating data of the equipment in real time, such as engine speed, working mode, running time, hydraulic oil temperature, valve port pressure, etc. By monitoring these data, the running condition, working load and energy efficiency performance of the equipment can be understood, so as to evaluate the running efficiency of the equipment and optimize the operation scheme. Abnormal state monitoring is to analyze the operating data of the equipment in long-term operation, establish a diagnostic model and identify possible abnormal types. By comprehensively monitoring and managing the running state and abnormal state of engineering machinery equipment, comprehensive monitoring and management of engineering machinery equipment can be achieved.

[0058] For running state monitoring, a communication protocol needs to be designed to upload equipment operating data in real time for processing and analysis. Compared with the traditional Transport Control Protocol (TCP) and User Datagram Protocol (UDP), the Message Queuing Telemetry Transport (MQTT) protocol has obvious advantages in saving bandwidth, network adaptability, flexibility, reliability, mobile support and security through mechanisms such as publish / subscribe mode, dynamic topic and Quality of Service (QoS), and is more suitable for the needs of Internet of Things applications such as engineering machinery state monitoring.

[0059] For abnormal state monitoring, a diagnosis model needs to be established according to the collected working condition data of the equipment to diagnose the equipment in real time. The traditional abnormal diagnosis method needs to use normal and various abnormal samples to train the diagnosis model, and a centralized central server is used to process the diagnosis tasks of all equipment. However, in the early stage of use of a certain type of equipment, there are few abnormal samples, and it is difficult to train a complete diagnosis model. In addition, the centralized execution of diagnosis using the central server may also cause the server and network to be overloaded.

[0060] Based on this, the present disclosure provides a state monitoring method.

[0061] FIG. 1 shows a flowchart of a state monitoring method according to some embodiments of the present disclosure. As shown in FIG. 1, the method of this embodiment includes steps S102-S108.

[0062] In step S102, a training sample set is generated based on state data of a plurality of first equipment, wherein the plurality of first equipment are of the same type, and the first equipment is an engineering machinery equipment.

[0063] The state of the engineering machinery equipment of the same type is monitored, that is, the training sample set is generated based on the state data of the engineering machinery equipment of the same type. The engineering machinery equipment of the same type is relative to the engineering machinery equipment of the same type, for example, the same type of excavator, the same type of crane, etc.

[0064] In some embodiments, generating the training sample set based on the state data of the plurality of first equipment includes: selecting a plurality of feature parameters from the state data of the plurality of first equipment; and generating the training sample set based on the state data corresponding to the plurality of feature parameters.

[0065] The state data of the first equipment includes state data corresponding to various working parameters, and when generating the training sample set, feature parameters capable of reflecting the state of the equipment are selected, for example, vibration signals of various parts, hydraulic valve port pressures, engine temperatures, etc., and the state data corresponding to these feature parameters is used as training samples for distinguishing abnormal states.

[0066] In step S104, a diagnosis model is determined based on the types of samples in the training sample set.

[0067] For each type of equipment, in the early stage of use of the equipment, the equipment is often in a normal state, therefore, the initial training sample set only includes normal samples, that is, the training sample set only includes one type of samples, and a single Gaussian model is used as the diagnosis model at this time. The single Gaussian model can distinguish between normal and abnormal states.

[0068] With the accumulation of the use time of the equipment, various types of abnormalities may occur to the equipment, and after corresponding maintenance is performed on the equipment, the types of abnormalities of the equipment can be marked, so that the abnormality samples corresponding to various types of abnormalities can be generated based on the abnormality state data of the equipment, and the abnormality samples are supplemented to the training data set, that is, the training sample set includes multiple types of samples (including normal samples and various types of abnormal samples), and at this time, the Gaussian mixture model is used as the diagnosis model, and the Gaussian mixture model can further monitor the types of abnormalities that may occur to the equipment, so as to perform targeted maintenance on the equipment.

[0069] In some embodiments, the state monitoring method further includes: generating one or more types of abnormality samples based on alarm data and supplementing the abnormality samples to the training sample set to update the training sample set, wherein the alarm data is generated in response to one or more of the plurality of second devices having an abnormality; and re-determining the diagnosis model based on the types of samples in the updated training sample set.

[0070] In some embodiments, re-determining the diagnosis model based on the types of samples in the updated training sample set includes: in response to the number of at least one type of abnormality sample in the one or more types of abnormality samples being greater than a first threshold, re-determining the diagnosis model based on the types of samples in the updated training sample set. Re-determining the diagnosis model in the case where the number of the same type of abnormality sample is sufficient can improve the effectiveness and accuracy of the determined diagnosis model.

[0071] The state monitoring method of the present disclosure considers that the early use stage of the equipment only includes normal samples, and abnormality samples are generated only after the use time of the equipment accumulates, so that a single Gaussian model is initially used as the diagnosis model, the single Gaussian model is used to identify abnormal states to generate abnormality samples, and then a Gaussian mixture model is used as the diagnosis model, and the Gaussian mixture model can identify various types of abnormality samples. Moreover, the Gaussian mixture model can be continuously updated with the generation of various new types of abnormality samples, thereby improving the accuracy of monitoring the state of the equipment.

[0072] In step S106, the training sample set is used to train the diagnosis model to determine the values of the parameters of the diagnosis model.

[0073] In the case where the training sample set only includes normal samples, the diagnosis model is a single Gaussian model, and the parameters of the diagnosis model include a second threshold, a mean of the single Gaussian model, and a covariance matrix. The second threshold represents a false alarm margin, and a diagnosis boundary is established by setting the false alarm margin, that is, a certain proportion of the probability density values of the training samples are allowed to be lower than the second threshold, that is, a certain proportion of the training samples are located outside the diagnosis boundary.

[0074] FIG. 2 shows a schematic diagram of a diagnostic boundary of a single Gaussian model according to some embodiments of the present disclosure. FIG. 2 shows a diagnostic boundary 21 of a single Gaussian model. As shown in FIG. 2, the single Gaussian model is able to identify abnormal samples. There are also a certain number of normal samples outside the diagnostic boundary 21, which indicates that the single Gaussian model sets a certain false positive margin.

[0075] Let all the normal samples {x1, x2, x3, …, xm} in the training sample set W form a sample matrix X, where m is the number of normal samples. Then the parameters of the single Gaussian model are determined based on the training sample set W as follows: m} form a sample matrix X, where m is the number of normal samples. Then the parameters of the single Gaussian model are determined based on the training sample set W as follows:

[0076] The mean value μ of all the normal samples in the sample matrix X is determined according to formula (1);

[0077] The covariance matrix S is determined according to formula (2);

[0078] The probability density function of the single Gaussian model is determined according to the mean value μ and the covariance matrix S, i.e. formula (3).

[0079] The threshold parameter of the diagnostic boundary is determined by setting the false positive margin. For example, the false positive margin can be set in the range of 5% to 10%. For example, when the false positive margin is set to 5%, the probability density values of 5% of the training samples are lower than the second threshold ε.

[0080] The state monitoring method further comprises training the redetermined diagnostic model using the updated training sample set to determine the values of the parameters of the redetermined diagnostic model.

[0081] In the case where the training sample set includes normal samples and abnormal samples, the diagnostic model is a Gaussian mixture model, and the parameters of the diagnostic model include the second threshold, the weight, the mean value, and the covariance matrix of each single Gaussian model included in the Gaussian mixture model.

[0082] Let the training sample set include M types of samples (including normal samples). Then the probability density function of the Gaussian mixture model is obtained by weighted summation of the probability density functions of the M single Gaussian models, and the probability density function of the Gaussian mixture model can be represented by formula (4).

[0083] where M is the order of the Gaussian mixture model, X represents the samples in the training sample set, p i (x) represents the probability density function of the i-th single Gaussian model, ω i represents the weight of the i-th single Gaussian model, and

[0084] The probability density function of the ith single Gaussian model can be represented by equation (5).

[0085] Therefore, the Gaussian mixture model is described by three parameters, i.e., the mean, the covariance matrix and the weight, namely the parameters of the Gaussian mixture model λ = {ω i , μ i , S i}, i = 1, 2, 3, …, M. The converged parameters λ are determined after the parameters λ are estimated by an algorithm such as expectation-maximization (EM).

[0086] The Gaussian mixture model also sets a second threshold to represent the false alarm margin. FIG. 3A shows a schematic diagram of a diagnostic boundary of a second-order Gaussian mixture model according to some embodiments of the present disclosure. When the training sample set includes two types of samples, the diagnostic boundary 31 is as shown in FIG. 3A. FIG. 3B shows a schematic diagram of a diagnostic boundary of a third-order Gaussian mixture model according to some embodiments of the present disclosure. When the training sample set includes three types of samples, the diagnostic boundary 32 is as shown in FIG. 3B.

[0087] As shown in FIGS. 3A and 3B, the Gaussian mixture model can identify normal or known faults (known abnormalities), new and abnormal faults (unknown abnormalities). That is, samples falling within the diagnostic boundary are considered normal or known faults, and samples outside the diagnostic boundary are considered new and abnormal faults. After identifying the new and abnormal fault samples, corresponding maintenance is performed and the abnormal type of the new and abnormal fault is determined, and after a sufficient number of new and abnormal fault samples are accumulated, they are supplemented to the training sample set, so that the order of the Gaussian mixture model and the values of the parameters thereof can be re-determined based on the updated training sample set.

[0088] In step S108, the diagnostic model and the values of the parameters thereof are sent to a plurality of second devices, so that each second device in the plurality of second devices determines whether its own state is abnormal by using the diagnostic model and the values of the parameters thereof, wherein the second devices are engineering mechanical devices of the same model as the first devices.

[0089] The state monitoring method further includes sending the re-determined diagnostic model and the values of the parameters thereof to a plurality of second devices, so that each second device in the plurality of second devices re-determines whether its own state is abnormal by using the re-determined diagnostic model and the values of the parameters thereof.

[0090] The second devices can be the same as the first devices. That is, after the training sample set is generated based on the state data of the plurality of first devices, the diagnostic model and the values of the parameters thereof are determined, the diagnostic model and the values of the parameters thereof are sent to the plurality of first devices, so that each first device in the plurality of first devices determines whether its own state is abnormal.

[0091] The second device can be selected from the first devices. For example, part of the first devices in the plurality of first devices are selected as the second device.

[0092] The second device can also be a construction machinery device that is not selected from the plurality of first devices but has the same model as the first device.

[0093] Sending the diagnostic model and the values of the parameters of the diagnostic model to the second device enables the second device to monitor its own state, which can reduce resource pressure, has low requirements on the network, and enables the second device to directly monitor its own state based on its own state data without the need to transmit intermediate data, which can improve monitoring efficiency and also improve the accuracy of monitoring.

[0094] In some embodiments, the state monitoring method of the present disclosure monitors the state of the construction machinery device based on the MQTT protocol. That is, the construction machinery device sends its own running data to the server, and the server processes, forwards, etc. these running data for state monitoring.

[0095] In some embodiments, sending the diagnostic model and the values of the parameters of the diagnostic model to the plurality of second devices includes sending the diagnostic model and the values of the parameters of the diagnostic model to the server, so that the server forwards the diagnostic model and the values of the parameters of the diagnostic model to the plurality of second devices based on the MQTT protocol.

[0096] The present disclosure determines a diagnostic model based on the state data of the plurality of first devices and trains the diagnostic model. After determining the values of the parameters of the diagnostic model, the diagnostic model and the values of the parameters of the diagnostic model are sent to the plurality of second devices, so that the second devices can use the diagnostic model and the values of the parameters of the diagnostic model to monitor the state of the second devices. The present disclosure enables the construction machinery device to monitor its own state, which can improve the efficiency of state monitoring, save resources, and enable the construction machinery device to monitor its own state in real time, thereby improving the accuracy of state monitoring.

[0097] FIG. 4 shows a structural schematic diagram of a state monitoring apparatus according to some embodiments of the present disclosure. As shown in FIG. 4, the state monitoring apparatus 40 includes modules 410-440.

[0098] The first generation module 410 is configured to generate a training sample set based on the state data of the plurality of first devices, wherein the plurality of first devices have the same model and the first device is a construction machinery device.

[0099] The first determination module 420 is configured to determine a diagnostic model based on the types of the samples in the training sample set.

[0100] The first training module 430 is configured to train the diagnostic model using the training sample set to determine the values of the parameters of the diagnostic model.

[0101] The first sending module 440 is configured to send the diagnostic model and the values of the parameters of the diagnostic model to the plurality of second devices, so that each of the plurality of second devices determines whether the state of the second device is abnormal by using the diagnostic model and the values of the parameters of the diagnostic model, wherein the second device is a same type of construction machinery device as the first device.

[0102] In some embodiments, the state monitoring apparatus 40 further comprises modules 450-480.

[0103] The second generating module 450 is configured to generate one or more types of abnormal samples based on alarm data of the plurality of second devices, and supplement the abnormal samples to the training sample set to update the training sample set, wherein the alarm data is generated in response to one or more of the plurality of devices being abnormal.

[0104] The second determining module 460 is configured to re-determine the diagnostic model based on the types of the samples in the updated training sample set.

[0105] The second training module 470 is configured to train the re-determined diagnostic model by using the updated training sample set to determine the values of the parameters of the re-determined diagnostic model.

[0106] The second sending module 480 is configured to send the re-determined diagnostic model and the parameters of the re-determined diagnostic model to the plurality of second devices, so that each of the plurality of second devices re-determines whether the state of the second device is abnormal by using the re-determined diagnostic model and the parameters of the re-determined diagnostic model.

[0107] In some embodiments, the second training module 470 is configured to re-determine the diagnostic model based on the types of the samples in the updated training sample set in response to a number of at least one type of abnormal sample in the one or more types of abnormal samples being greater than a first threshold value.

[0108] In some embodiments, in a case where the training sample set only comprises normal samples, the diagnostic model is a single Gaussian model, and the parameters of the diagnostic model comprise a second threshold value, a mean value of the single Gaussian model, and a covariance matrix.

[0109] In some embodiments, in a case where the training sample set comprises normal samples and abnormal samples, the diagnostic model is a Gaussian mixture model, and the parameters of the diagnostic model comprise a second threshold value, a weight, a mean value, and a covariance matrix of each single Gaussian model included in the Gaussian mixture model.

[0110] In some embodiments, the first generating module 410 is configured to select a plurality of feature parameters from the state data of the plurality of first devices; and generate the training sample set based on the state data corresponding to the plurality of feature parameters.

[0111] In some embodiments, the first sending module 440 is configured to send the diagnostic model and the values of the parameters thereof to the server, so that the server forwards the diagnostic model and the values of the parameters thereof to the plurality of second devices based on a Message Queue Telemetry Transport (MQTT) protocol.

[0112] FIG. 5 shows a structural schematic diagram of a condition monitoring system according to some embodiments of the present disclosure. As shown in FIG. 5, the condition monitoring system 50 includes a condition monitoring device 510 as described above; a second device 520 configured to receive the diagnostic model and the parameters thereof sent by the condition monitoring device, and determine whether the state of the second device is abnormal by using the diagnostic model and the parameters thereof.

[0113] In some embodiments, the second device is configured to generate a plurality of test samples based on the state data of the second device; input the plurality of test samples into the diagnostic model to determine a probability density value corresponding to each of the plurality of test samples; and determine that the state of the second device is unknown abnormal in a case where the probability density value corresponding to each of the plurality of test samples is less than a second threshold value.

[0114] The second device acquires the state data of the second device in real time, extracts the feature parameters corresponding to the training samples, and generates a plurality of test samples, which are input into the diagnostic model.

[0115] In some embodiments, to prevent false positives due to the false positive margin of the diagnostic model, the plurality of test samples are acquired continuously, i.e., a plurality of continuous state data corresponding to the feature parameters are used as the plurality of test samples. In a case where the probability density value corresponding to each of the plurality of test samples is less than the second threshold value, i.e., the plurality of test samples all fall outside the diagnostic boundary of the diagnostic model, it is determined that the second device has an unknown abnormality (a new abnormality).

[0116] In some embodiments, the second device is configured to generate first alarm data based on the state data of the second device in a case where it is determined that the second device has an unknown abnormality. After a user or a maintenance personnel receives the first alarm data, the second device can be maintained by remote / on-site maintenance, and a maintenance record is generated, which includes information such as an abnormality type and a solution measure. The second device can send the maintenance record to the condition monitoring device, so that the corresponding abnormality type of the abnormality sample is marked based on the maintenance record, for subsequent use in updating the training sample set.

[0117] In some embodiments, the second device is configured to determine that the state of the second device is normal or has a known abnormality in a case where there is at least one test sample in the plurality of test samples whose corresponding probability density value is not less than the second threshold value.

[0118] In some embodiments, in the case that the training model is a single Gaussian model, the second device is configured to determine that the state of the second device is normal in the case that there is at least one test sample in the plurality of test samples corresponding to a probability density value not less than the second threshold.

[0119] Since the single Gaussian model only has two identification results, i.e., normal or abnormal, in the case that the training model is a single Gaussian model, in the case that there is at least one test sample in the plurality of test samples corresponding to a probability density value not less than the second threshold, it is considered that the state of the second device is normal.

[0120] And in the case that the training model is a Gaussian mixture model, in the case that there is at least one test sample in the plurality of test samples corresponding to a probability density value not less than the second threshold, it is considered that the state of the second device is a known state, i.e., normal or existing known abnormality.

[0121] In some embodiments, in the case that the diagnosis model is a Gaussian mixture model, the second device is configured to input the plurality of test samples into each single Gaussian model included in the Gaussian mixture model, and determine a state monitoring result and a probability density value corresponding to each single Gaussian model, wherein for each single Gaussian model, the state monitoring result corresponding to the single Gaussian model is normal or existing known abnormality and indicates a type of known abnormality corresponding to the single Gaussian model; and determine the state monitoring result corresponding to the second device according to the probability density value corresponding to each single Gaussian model.

[0122] In the case that the diagnosis model is a Gaussian mixture model, inputting the plurality of test samples into each single Gaussian model included in the Gaussian mixture model can further determine that the state of the second device is normal, or the abnormal type of the existing known abnormality.

[0123] In some embodiments, in the case that the diagnosis model is a Gaussian mixture model, the second device is configured to determine the state monitoring result of the single Gaussian model corresponding to the maximum value in the probability density value corresponding to each single Gaussian model as the state monitoring result corresponding to the second device.

[0124] In some embodiments, in the case that the diagnosis model is a Gaussian mixture model, the second device is configured to generate second alarm data based on the state data of the second device in the case that the second device has a known abnormality. After the user or the maintenance personnel receives the second alarm data, the second device can be maintained based on historical solutions through remote / on-site maintenance.

[0125] In some embodiments, the state monitoring system 50 further comprises a server 530 configured to receive the diagnosis model and its parameters sent by the state monitoring device, and forward the diagnosis model and its parameters to the plurality of second devices based on the MQTT protocol.

[0126] The server 530 is responsible for receiving and distributing status messages from various construction machinery devices. It communicates with the construction machinery devices through the MQTT protocol, receives the data sent by the construction machinery devices, and transmits the data to the relevant devices or apparatuses.

[0127] In some embodiments, the status monitoring system 50 further comprises a monitoring device 540 configured to receive the data sent by the second device and forward the data sent by the second device to the status monitoring apparatus.

[0128] The monitoring device is an application terminal connected to the server 530 through the MQTT protocol, and can be distributed in different geographical locations, such as offices, construction sites or other remote locations. The carrier can be various devices such as mobile phones, tablets, computers, etc., which establish a connection with the server 530 through wireless or wired means. Any monitoring device can flexibly choose to subscribe to the data sent by the second device, monitor and display the received construction machinery device status messages in real time, and provide data filtering, searching, chart display, database storage, data post-processing, etc.

[0129] In some embodiments, the server is configured to receive the data sent by the second device based on the MQTT protocol and forward the data sent by the second device to the monitoring device.

[0130] In some embodiments, the second device is configured to send maintenance data to the server, wherein the maintenance data is generated in response to maintenance of the second device based on the alarm data of the second device and completion.

[0131] In some embodiments, the server is configured to receive the data sent by the monitoring device based on the MQTT protocol, and forward the data sent by the monitoring device to one or more second devices according to the message identification of the data sent by the monitoring device, so that the monitoring device configures or controls one or more second devices.

[0132] Considering that the number of construction machinery devices and monitoring devices is large, and different monitoring devices are responsible for monitoring different construction machinery devices, classification and filtering are required when transmitting data between the construction machinery devices and the monitoring devices, so that the communication messages can be reasonably organized and managed. Therefore, the present disclosure provides a method for setting the topic structure of the message identification based on the MQTT protocol.

[0133] The data sent by at least one of the second device and the monitoring device has a message identification, and the message identification includes a plurality of hierarchical sub-identifications.

[0134] In some embodiments, the plurality of hierarchical sub-identifications includes the type of the device, the model of the device, the customer, the number of the device, and the topic of the message.

[0135] In some embodiments, the message identification includes a multi-layer topic structure, each layer of the topic structure representing a sub-identification, and the name format of each layer of the topic structure is as follows.

[0136] 1. Machine type layer (i.e., the type of equipment): The name format is "{machine_type}", and different types of construction machinery use different machine type topics, such as excavators, cranes, etc.

[0137] 2. Model layer (i.e., the model of equipment): The name format is "{machine_type} / {model}", which is used to distinguish different models of equipment under the same machine type.

[0138] 3. Customer layer (i.e., the customer of equipment): The name format is "{machine_type} / {model} / {customer_id}", which is used to identify the customer to whom the equipment is sold.

[0139] 4. Device layer (i.e., the number of equipment): The name format is "{machine_type} / {model} / {customer_id} / {device_id}", which is used to uniquely identify a specific equipment instance.

[0140] 5. Topic layer (i.e., the topic of the message), which is used to classify different types of application messages for equipment. The topic includes the following five types.

[0141] 1) Device status topic: The name format is "{machine_type} / {model} / {customer_id} / {device_id} / status", which is used to report the running parameters, sensor data, and other status information of the equipment.

[0142] 2) Device alarm topic: The name format is "{machine_type} / {model} / {customer_id} / {device_id} / alarm", which is used to publish alarm information of the equipment.

[0143] 3) Device control topic: The name format is "{machine_type} / {model} / {customer_id} / {device_id} / control", which is used to control the equipment to perform corresponding actions, such as starting, stopping, etc.

[0144] 4) Device maintenance topic: The name format is "{machine_type} / {model} / {customer_id} / {device_id} / maintenance", which is used to publish maintenance records, repair records, and other information of the equipment.

[0145] 5) Device configuration topic: the name format is "{layer name} / configuration", which is used to manage the configuration information of the device, including parameter setting, software upgrade, firmware update, etc. The "layer name" represents the name format of any level described in the 1-4 layers. When configuring the device, one-to-many batch configuration can be performed according to the specific model, type or customer, or one-to-one configuration can be performed for a specific device.

[0146] Table 1 shows the publishing and subscribing relationship of the message topics according to some embodiments of the present disclosure.

[0147] Table 1

[0148] In summary, the message identifier first classifies the construction machinery according to the model, type, customer and device. Then, under the device level, different message topics are used to classify different types of messages such as device status, alarm, control and maintenance, so as to realize reasonable grouping and effective isolation of messages. In addition, configuration topics are provided under each level of the model, type, customer and device. When configuring the target device, one-to-many batch configuration can be performed according to the model, type or customer, or one-to-one configuration can be performed for a specific device.

[0149] In some embodiments, the message identifier further includes a root layer, which is set in the layer before the model layer and is represented by the name format "engineering_machinery" for identifying the construction machinery. At this time, the complete message identifier includes a six-layer topic structure, which is respectively: root layer, model layer, type layer, customer layer, device layer and topic layer.

[0150] Figure 6 shows a schematic diagram of the topic structure of the message identifier according to some embodiments of the present disclosure. As shown in Figure 6, the message identifier includes a root layer, a model layer, a type layer, a customer layer, a device layer and a topic layer, wherein the root layer is used to identify the construction machinery; the model layer is used to identify the type of the device, such as cranes and excavators; the type layer is used to identify the type of the device, such as cranes of type XC and cranes of type G2; the customer layer is used to identify the customer, such as the customer of the cranes of type XC is customer A; the device layer is used to represent a specific device, such as the device 1 of the cranes of type XC of customer A; and the topic layer is used to identify the topic of the message, such as the message topics related to the device 3 of the cranes of type XC of customer B are status, alarm, control, maintenance and configuration.

[0151] In some embodiments, the topic of the message included in the message identifier of the data sent by the second device is any one of status, alarm and maintenance, and the topic of the information included in the message identifier of the data sent by the monitoring device is any one of control and configuration.

[0152] In some embodiments, the server is configured to receive data sent by a plurality of second devices; determine, according to the subscription information of the monitoring device, data subscribed by the monitoring device from the data sent by the plurality of second devices; and send the data subscribed by the monitoring device to the monitoring device, so that the monitoring device monitors the subscribed device.

[0153] In some embodiments, the server is configured to: determine a message identifier matching the subscription information of the monitoring device; and send data corresponding to the message identifier to the monitoring device.

[0154] In some embodiments, the server is configured to: for any level in the plurality of levels, if a sub-identifier of the level of the message identifier matches a value of the level of the subscription information of the monitoring device, determine that the message identifier matches the subscription information of the monitoring device.

[0155] In some embodiments, the value of one or more levels in the subscription information is a first wildcard, and for any level in the plurality of levels, if the value of the level in the subscription information is the first wildcard, the value of the level in the subscription information matches any sub-identifier.

[0156] In some embodiments, the last level in the subscription information is a second wildcard, wherein the second wildcard indicates that the level at which the second wildcard of the subscription information is located and the sub-levels of the second wildcard match any sub-identifier.

[0157] For example, the subscription message of the monitoring device can be set through a topic filter. The monitoring device sets a topic filter to the server based on the six-layer topic structure described above to subscribe to the messages of the target topic. The topic filter subscribed by the monitoring device is valid until the monitoring device actively cancels the subscription or the session is terminated. The topic filter can be a complete six-layer structure topic name, or a wildcard can be used to replace a specified level in the topic name.

[0158] The wildcard includes a first wildcard and a second wildcard. The first wildcard is represented by “+” for example, and is used only for single-layer wildcard matching a single topic level. The single-layer wildcard can be used at any level of the topic filter, including the first and last levels. It can be used in multiple levels in the topic filter, or together with a multi-layer wildcard.

[0159] The second wildcard is represented by “#” for example, and is used for multi-layer wildcard matching any level in the topic name. The multi-layer wildcard indicates its parent level and any number of sub-levels. The multi-layer wildcard must be the last character and must not be preceded by a character or preceded by a character “ / ”.

[0160] The construction machinery equipment publishes the equipment state message to the server through the respective equipment state topic (message identifier). The construction machinery equipment collects state information such as operating parameters and sensor data, including key parameters such as the working time, voltage, temperature, and oil pressure of the equipment; and binds these data to its equipment state topic (message identifier) and sends them to the server through the MQTT protocol.

[0161] After receiving the message sent by the construction machinery equipment, the server checks the equipment state topic (message identifier), matches it with the subscription information (topic filter) of the monitoring equipment, and then distributes the application message to all monitoring equipment matching the topic filter. The entire topic is matched only when each level of the topic filter matches the corresponding level of the equipment state topic.

[0162] For each level, a single-layer wildcard matches any character. A multi-layer wildcard matches all levels from this level and the following levels. When there is no wildcard, the strings must be exactly the same to be considered a match (including case, spaces, other special characters, etc.).

[0163] Once the server receives the equipment state message, it will deliver the message to the corresponding monitoring equipment according to the subscription relationship, ensuring that the monitoring equipment can timely obtain the running state data of the equipment.

[0164] After receiving the subscribed information, the monitoring equipment extracts the state information of the equipment, realizing the monitoring of the equipment running state. After receiving the state message of the equipment, the monitoring equipment can parse, process, and display the data. The monitoring equipment can perform operations such as chart display, real-time monitoring, alarm notification, database storage, data post-processing, etc. according to the needs, to realize the monitoring and analysis of the equipment running state.

[0165] Figure 7 shows a schematic diagram of a state monitoring system according to some other embodiments of the present disclosure. As shown in Figure 7, the state monitoring system includes an MQTT server 710, various construction machinery equipment 720, various monitoring equipment 730, a fault mode library 740, and a state monitoring device 750.

[0166] The state monitoring device 750 is responsible for developing and training models for construction machinery anomaly detection and diagnosis. It receives state data from construction machinery equipment and uses machine learning or other related technologies for model training to identify and predict abnormal types.

[0167] The fault mode library 740 is a database that stores different construction machinery fault mode samples. It contains various possible fault modes and corresponding feature samples, providing a training sample set for the establishment of a diagnosis model.

[0168] The MQTT server 710 receives the messages sent by the various engineering machinery devices 720 based on the MQTT protocol, and distributes the data sent by the engineering machinery devices to the corresponding monitoring devices according to the message identification and the topic filter of the monitoring device 730. The monitoring device monitors the engineering machinery device based on the received subscribed message. And the monitoring device can extract the state data of the engineering machinery device and store it into the fault mode library 740, so that the state monitoring device 750 can generate a training sample set of various types of devices according to the fault mode library to determine the diagnostic model and the value of its parameters, and send the diagnostic model and the value of its parameters to the corresponding type of device through the MQTT server 710 to realize self-monitoring of the device on its own state.

[0169] FIG. 8 shows a flowchart of state monitoring by the monitoring device according to some embodiments of the present disclosure. As shown in FIG. 8, the state monitoring by the monitoring device includes steps S820-S880.

[0170] In S820, the monitoring device sets a topic filter to the MQTT server to subscribe to the messages of the target topic based on the six-layer topic structure of the message identification of the engineering machinery device (based on the MQTT protocol).

[0171] In S840, the engineering machinery device publishes the device state message to the MQTT server through the respective device state topic.

[0172] In S860, the MQTT server receives the message sent by the engineering machinery device, checks the message identification, matches it with the topic filter of the monitoring device, and then distributes the message to all monitoring devices matching the topic filter.

[0173] In S880, the monitoring device receives the subscribed topic message, extracts the state information from the message, and realizes the monitoring of the running state of the engineering machinery.

[0174] FIGS. 9A-9C show schematic diagrams of the monitoring device monitoring the state of the device according to some embodiments of the present disclosure.

[0175] As shown in FIG. 9A, the monitoring device monitors the device state parameters of a specific device. The monitoring device subscribes to “engineering_machinery / excavator / B1 / customerB / device001 / status” to receive the state message of the excavator device 001 of customer B, model B1, and monitors the state of the device.

[0176] As shown in FIG. 9B, the monitoring device monitors the status parameters of all devices of a specific model under a specific machine type. The monitoring device subscribes to “engineering_machinery / crane / XC / + / + / status” to receive the status messages of all devices of a specific model under a specific machine type, i.e., the status monitoring of all devices of the crane of model XC.

[0177] As shown in FIG. 9C, the monitoring device monitors the status parameters of all engineering machinery devices used by a specific customer. The monitoring device subscribes to “engineering_machinery / + / + / customerA / + / status” to receive the status messages of all engineering machinery devices used by a specific customer, i.e., the status monitoring of all engineering machinery devices used by customer A.

[0178] In addition to FIGS. 9A-9C, a variety of different monitoring scenarios can be met by flexible combination of wildcards at each level, such as monitoring the status parameters of all devices of a specific machine type purchased by a specific customer, and the like. Here, no longer one by one example.

[0179] Further, the engineering machinery device subscribes to the configuration topics at each level, so that the monitoring device can perform one-to-many batch configuration according to the machine type, model or customer, or one-to-one configuration for a certain device.

[0180] FIGS. 10A-10B show a schematic diagram of the monitoring device configuring the device according to some embodiments of the present disclosure.

[0181] As shown in FIG. 10A, for a specific device, the monitoring device configures the parameters in a one-to-one manner. The engineering machinery device subscribes to the configuration topics at each level, and the message identifier of the monitoring device is the topic “engineering_machinery / excavator / B1 / customerB / device001 / configuration”, i.e., the configuration parameters are issued to the excavator device 001 of model B1 of customer B.

[0182] As shown in FIG. 10B, for all devices of a specific model, the monitoring device configures the parameters in a one-to-many manner. The engineering machinery device subscribes to the configuration topics at each level, and the message identifier of the monitoring device is “engineering_machinery / crane / XC / configuration”, i.e., the configuration parameters are issued to the crane of model XC.

[0183] Similar to the configuration of the device, the monitoring device can configure the device in a one-to-one and one-to-many manner, which is not described here.

[0184] FIG. 11 shows a flowchart of a state monitoring method according to some embodiments of the present disclosure. The state monitoring method is used to monitor a specific model of construction machinery, i.e., for different models of construction machinery, a corresponding abnormal diagnosis model is established respectively. In order to avoid the shortcomings of centralized fault diagnosis, the model parameters are published to each construction machinery under the model, and the abnormal diagnosis is implemented by each construction machinery in a distributed manner. As shown in FIG. 11, for any model of equipment, the state monitoring method includes steps S1-S6.5.

[0185] In step S1, a plurality of working parameters (such as vibration signals of each component, hydraulic valve port pressure, engine temperature, etc.) are selected from the operating state parameters of the construction machinery as characteristic samples for discriminating abnormal states.

[0186] In step S2, in the early stage of use of the construction machinery of the model, only normal samples are in the fault mode library, and only normal samples are used to establish a diagnosis model, including steps S2.1-S2.4.

[0187] In step S2.1, the monitoring device subscribes to the device state topic of the model “engineering_machinery / {machine_type} / {model} / + / + / status”, receives the operating state data of all construction machinery of the model under normal operating state, extracts the characteristic samples, and stores them in the fault mode library.

[0188] In step S2.2, the state monitoring device establishes a diagnosis model based on the sample set under normal operating state using a single Gaussian model, and establishes a diagnosis boundary by setting a false alarm margin (second threshold).

[0189] In step S2.3, the state monitoring device publishes the trained model and the values of its parameters to the configuration topic of the model “engineering_machinery / {machine_type} / {model} / configuration”. Since the diagnosis model is a single Gaussian model, in fact, the state monitoring device can only publish the values of the model parameters to enable the construction machinery to determine the diagnosis model and the values of its parameters.

[0190] In step S2.4, all construction machinery of the model subscribes to the configuration topic and receives the values of the model parameters.

[0191] In step S3.1, each engineering machinery device under the model runs the distributed single-Gaussian model, acquires the running parameters in real time and extracts the test sample features, and inputs the test sample features into the single-Gaussian model to determine whether the state of the engineering machinery device is abnormal. To avoid false positives due to the false positive margin set in S2.2, a plurality of continuous test samples are used to determine the state of the engineering machinery device. If the probability density values of the test sample features of the continuous test samples are greater than the second threshold value, it is determined that the state of the engineering machinery device is normal, and step S3 is repeated, that is, the state of the engineering machinery device is continuously monitored, and in the case where the current state monitoring result is normal, the state monitoring of the engineering machinery device is maintained; if the probability density values of the test sample features of the continuous test samples are all less than the second threshold value, it is determined that the state of the engineering machinery device is abnormal, and step S4 is performed.

[0192] FIG. 12 shows the sample probability density of the engineering machinery device in the normal state according to some embodiments of the present disclosure.

[0193] FIG. 13 shows the sample probability density of the engineering machinery device in the abnormal state according to some embodiments of the present disclosure. As shown in FIGS. 12 and 13, the sample probability densities of the engineering machinery device in the normal state and the abnormal state are different, and the samples in the abnormal state have obviously small probability density values of a plurality of continuous samples, that is, less than the second threshold value.

[0194] In step S4, as the engineering machinery device ages, the engineering machinery device will gradually have new and different failures that have not occurred before, and generate corresponding abnormal samples. At this time, the abnormal samples are updated to the fault mode library, including steps S4.1-S4.5.

[0195] In step S4.1, when the engineering machinery device identifies a new abnormality, an abnormality alarm is issued, and before the abnormality alarm is cleared, the sample features of the abnormality are continuously published to the device alarm topic “engineering_machinery / {machine_type} / {model} / {customer_id} / {device_id} / alarm”.

[0196] In step S4.2, the monitoring device subscribes to the device alarm topic “engineering_machinery / {machine_type} / {model} / + / + / alarm” of the model through a single-layer wildcard, to receive the abnormal sample features of all engineering machinery devices of the model. Since the specific abnormal type is not determined at this time, the abnormal type is temporarily marked as an unknown abnormality, and is stored in the fault mode library.

[0197] In step S4.3, the user / driver removes the fault alarm of the equipment by remote or on-site means after monitoring the alarm signal; the engineer is assigned to locate the anomaly; after the anomaly is solved, the maintenance record is input into the engineering machinery equipment, which contains information such as anomaly type and solving measures.

[0198] In step S4.4, the engineering machinery equipment publishes the maintenance record to the equipment maintenance topic "engineering_machinery / {machine_type} / {model} / {customer_id} / {device_id} / maintenance".

[0199] In step S4.5, the monitoring equipment subscribes to the equipment maintenance topic "engineering_machinery / {machine_type} / {model} / + / + / maintenance" of the model through single-layer wildcard, to receive the maintenance records of all engineering machinery equipment of the model. After receiving the maintenance record, the monitoring equipment extracts and records the corresponding solving measures, and extracts the anomaly type information and re-labels the samples marked as unknown anomalies in the fault mode library as the anomaly type.

[0200] In step S5, when the number of samples of a certain anomaly type in the fault mode library accumulates to a certain number, the state monitoring device is triggered to re-determine the diagnostic model and the value of the parameter, and publish the new diagnostic model and the value of the parameter to each engineering machinery equipment, including steps S5.1-S5.5.

[0201] In step S5.1, the fault mode library detects the number of samples of each anomaly type that has not been added to the training, and when the number of anomaly samples of a certain anomaly type reaches a certain number (exceeds the first threshold), step S5.2 is executed; otherwise, step S5.5 is executed.

[0202] In step S5.2, the fault mode library sends a trigger instruction to the state monitoring device, and the state monitoring device establishes a Gaussian mixture model based on normal samples and anomaly samples of different anomaly types, and establishes a diagnostic boundary by setting a false alarm margin (second threshold).

[0203] In step S5.3, the state monitoring device re-publishes the value of the re-determined model parameter to the configuration topic of the model "engineering_machinery / {machine_type} / {model} / configuration".

[0204] In step S5.4, all engineering machinery equipment of the model subscribes to the configuration topic and receives the value of the parameter of the diagnostic model.

[0205] In step S5.5, each engineering machinery device judges the type of its own diagnosis model. If it is a single Gaussian model, S3 is executed. If it is a Gaussian mixture model, S6 is executed.

[0206] In step S6, the engineering machinery device runs the re-determined Gaussian mixture model to monitor its own state in real time, including S6.1-S6.5.

[0207] In step S6.1, the engineering machinery device acquires running parameters in real time and extracts test sample features, and inputs the test sample features into the Gaussian mixture model. When judging faults, a plurality of samples are used for comprehensive judgment. If the probability density values of a plurality of consecutive test sample features are greater than the second threshold value, it is determined that the state of the engineering machinery device is normal or there is a known abnormality, and step S6.2 is executed. If the probability density values of a plurality of consecutive test sample features are all less than the second threshold value, it is determined that the state of the engineering machinery device is an unknown abnormality, and S4 is executed.

[0208] In step S6.2, the test sample features are respectively input into each single Gaussian model constituting the Gaussian mixture model, the probability density values of the test sample features under different abnormality types (including normal and a plurality of known abnormality types) are calculated, and the type with the highest probability density value is selected as the output abnormality type.

[0209] In step S6.3, if the abnormality type diagnosed by the engineering machinery device is normal, step S6.1 is executed, that is, the state monitoring of the engineering machinery device is maintained. If the abnormality type is a known abnormality, step S6.4 is executed.

[0210] In step S6.4, the engineering machinery device issues a fault alarm, and at the same time, publishes the abnormality type to the device alarm topic “engineering_machinery / {machine_type} / {model} / {customer_id} / {device_id} / alarm”.

[0211] In step S6.5, the monitoring device subscribes to the device alarm topic “engineering_machinery / {machine_type} / {model} / + / + / alarm” of the model through a single layer wildcard, so as to receive the abnormality type and match the corresponding solution measure to push to the user. The user can quickly repair the abnormality according to the abnormality type and the solution measure. After the abnormality is repaired, S6.1 is returned, and the state monitoring of the engineering machinery device is maintained.

[0212] In the aspect of the operation state monitoring of the engineering machinery device, the disclosure formulates the MQTT six-layer topic structure suitable for the engineering machinery, and provides the monitoring method and system of the operation state of the engineering machinery on the basis, so that the flexible monitoring and data acquisition and parameter configuration of the specific model, type, customer or device can be realized according to different classification requirements.

[0213] In the aspect of the abnormal state monitoring of the engineering machinery device, the disclosure provides a distributed engineering machinery device abnormal state monitoring method with model adaptive dynamic updating, solves the problem that the diagnostic model is difficult to establish in the early stage of the use of the engineering machinery device due to the lack of enough abnormal samples, and the diagnostic model is used in a distributed manner, and the model parameters are updated from the central server, so that the problems of high load and network congestion caused by centralized abnormal diagnosis are solved, the efficiency of the state monitoring is improved, the resources are saved, the engineering machinery device can perform real-time monitoring in the self-state monitoring, and the accuracy of the state monitoring is improved.

[0214] The state monitoring apparatus in the embodiments of the disclosure can be implemented by various computing devices or computer systems, which are described below in combination with FIG. 14 and FIG. 15.

[0215] FIG. 14 shows a structural schematic diagram of a state monitoring apparatus according to some other embodiments of the disclosure. As shown in FIG. 14, the apparatus 140 of this embodiment includes a memory 1410 and a processor 1420 coupled to the memory 1410, and the processor 1420 is configured to execute the state monitoring method in any some embodiments of the disclosure based on the instructions stored in the memory 1410.

[0216] The memory 1410 may, for example, include a system memory, a fixed nonvolatile storage medium, etc. The system memory, for example, stores an operating system, an application program, a Boot Loader, a database, and other programs, etc.

[0217] Fig. 15 shows a structural schematic diagram of a state monitoring device according to yet some embodiments of the present disclosure. As shown in Fig. 15, the device 150 of this embodiment includes a memory 1510 and a processor 1520, which are similar to the memory 1410 and the processor 1420 respectively. It can also include an input / output interface 1530, a network interface 1540, a storage interface 1550, etc. These interfaces 1530, 1540, 1550 and the memory 1510 and the processor 1520 can be connected through a bus 1560, for example. Among them, the input / output interface 1530 provides a connection interface for display, mouse, keyboard, touch screen and other input / output devices. The network interface 1540 provides a connection interface for various networking devices, which can be connected to a database server or a cloud storage server, etc. The storage interface 1550 provides a connection interface for external storage devices such as SD card and U disk.

[0218] Embodiments of the present disclosure also provide a computer readable storage medium having stored thereon a computer program, characterized in that the program, when executed by a processor, implements any of the preceding state monitoring methods.

[0219] Embodiments of the present disclosure also provide a computer program product comprising instructions which, when executed by a processor, cause the processor to carry out any of the preceding state monitoring methods.

[0220] Those skilled in the art should understand that the embodiments of the present disclosure can be provided as a method, a system, or a computer program product. Therefore, the present disclosure can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present disclosure can take the form of a computer program product implemented on one or more computer-usable non-transitory storage media (including, but not limited to, disk storage, CD-ROMs, optical storage devices, etc.) having computer-usable program code contained therein.

[0221] The present disclosure is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present disclosure. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, as well as combinations of flows and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing apparatus to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing apparatus produce the functions specified in the flowchart and / or block diagram one or more flows and / or blocks.

[0222] These computer program instructions can also be stored in a computer readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer readable memory produce an article of manufacture including instructions which implement the function specified in the flowchart or flowsheets and / or block or blocks of the block diagrams.

[0223] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart or flowsheets and / or block or blocks of the block diagrams.

[0224] The above description is merely illustrative of the application, and is not to be taken in a limiting sense. It is contemplated that various modifications and changes can be made by those skilled in the art without departing from the spirit and scope of the application.

Claims

1. A state monitoring method, comprising: generating a training sample set based on state data of a plurality of first devices, wherein the plurality of first devices are of a same model, and the first devices are engineering machinery devices; determining a diagnosis model based on types of samples in the training sample set; training the diagnosis model using the training sample set to determine values of parameters of the diagnosis model; sending the diagnosis model and the values of the parameters to a plurality of second devices, so that each second device of the plurality of second devices determines whether an abnormality exists in a state of the second device using the diagnosis model and the parameters.

2. The state monitoring method of claim 1, further comprising: generating one or more types of abnormal samples based on alarm data and supplementing the abnormal samples to the training sample set to update the training sample set, wherein the alarm data is generated in response to one or more second devices of the plurality of second devices having an abnormality; redetermining a diagnosis model based on types of samples in the updated training sample set; training the redetermined diagnosis model using the updated training sample set to determine values of parameters of the redetermined diagnosis model; sending the redetermined diagnosis model and the parameters to the plurality of second devices, so that each second device of the plurality of second devices redetermines whether an abnormality exists in a state of the second device using the redetermined diagnosis model and the parameters.

3. The condition monitoring method of claim 1 or 2, wherein, The redetermining a diagnosis model based on types of samples in the updated training sample set comprises: in response to a number of at least one type of abnormal sample of the one or more types of abnormal samples being greater than a first threshold, redetermining the diagnosis model based on types of samples in the updated training sample set.

4. The state monitoring method of any one of claims 1 to 3, wherein, in a case where the training sample set only includes normal samples, the diagnosis model is a single Gaussian model, and the parameters of the diagnosis model include a second threshold, a mean of the single Gaussian model, and a covariance matrix; or in a case where the training sample set includes normal samples and abnormal samples, the diagnosis model is a Gaussian mixture model, and the parameters of the diagnosis model include a second threshold, a weight, a mean, and a covariance matrix of each single Gaussian model included in the Gaussian mixture model.

5. The condition monitoring method according to any one of claims 1 to 4, wherein, The generating a training sample set based on state data of a plurality of first devices comprises: selecting a plurality of feature parameters from the state data of the plurality of first devices; generating the training sample set based on state data corresponding to the plurality of feature parameters.

6. The condition monitoring method of any one of claims 1 to 5, wherein, The sending the diagnosis model and the values of the parameters to the plurality of second devices comprises: sending the diagnosis model and the values of the parameters to a server, so that the server forwards the diagnosis model and the values of the parameters to the plurality of second devices based on a message queue telemetry transport (MQTT) protocol.

7. A state monitoring apparatus, comprising: The first generation module is configured to generate a training sample set based on state data of a plurality of first devices, wherein the plurality of first devices are of the same model and the first devices are engineering mechanical devices; The first determination module is configured to determine a diagnosis model based on types of samples in the training sample set; The first training module is configured to train the diagnosis model using the training sample set to determine values of parameters of the diagnosis model; The first sending module is configured to send the diagnosis model and the values of the parameters to the plurality of second devices, so that each second device in the plurality of second devices determines whether its state is abnormal using the diagnosis model and the parameters, wherein the second devices are engineering mechanical devices of the same model as the first devices.

8. A state monitoring apparatus comprising: a processor; and a memory coupled to the processor for storing instructions that, when executed by the processor, cause the processor to perform the state monitoring method of any one of claims 1 to 6.

9. A state monitoring system comprising: the state monitoring apparatus of claim 7 or 8; a second device configured to receive a diagnosis model and parameters sent by the state monitoring apparatus and determine whether its state is abnormal using the diagnosis model and the parameters. The second device is configured to:

10. The condition monitoring system of claim 9, wherein, generate a plurality of test samples based on state data of the second device; input the plurality of test samples into the diagnosis model to determine a probability density value corresponding to each test sample in the plurality of test samples; determine that the state of the second device is unknown abnormal if the probability density value corresponding to each test sample in the plurality of test samples is less than a second threshold, or determine that the state of the second device is normal or known abnormal if there is at least one test sample in the plurality of test samples whose corresponding probability density value is not less than the second threshold. The second device is configured to:

11. A condition monitoring system according to claim 9 or 10, wherein, generate first alarm data based on the state data of the second device if it is determined that the second device has unknown abnormality; or generate second alarm data based on the state data of the second device if the second device has known abnormality.

12. The state monitoring system of claim 10 or 11, wherein the second device is configured to determine that the state of the second device is normal if there is at least one test sample in the plurality of test samples whose corresponding probability density value is not less than the second threshold when the training model is a single Gaussian model; or ​ The second device is configured to, in the case that the diagnostic model is a Gaussian mixture model, input the plurality of test samples into each single Gaussian model included in the Gaussian mixture model, and determine a state monitoring result and a probability density value corresponding to each single Gaussian model, wherein for each single Gaussian model, the state monitoring result corresponding to the single Gaussian model is normal or there is a known abnormality and indicates a type of the known abnormality, the type of the known abnormality corresponding to the single Gaussian model, and the state monitoring result corresponding to the second device is determined according to the probability density value corresponding to each single Gaussian model.

13. The state monitoring system according to any one of claims 9 to 12, further comprising: a server configured to receive the diagnostic model and its parameters sent by the state monitoring device, and forward the diagnostic model and its parameters to the plurality of second devices based on an MQTT protocol, or configured to receive data sent by the second devices based on the MQTT protocol, and forward the data sent by the second devices to a monitoring device.

14. The state monitoring system according to any one of claims 9 to 13, further comprising: a monitoring device configured to receive data sent by the second devices, and forward the data sent by the second devices to the state monitoring device.

15. The condition monitoring system of any one of claims 12 to 14, wherein, The second device is configured to: send maintenance data to the server, wherein the maintenance data is generated in response to maintenance of the second device according to alarm data of the second device and after completion.

16. The condition monitoring system of any one of claims 13 to 15, wherein, The server is configured to: receive data sent by the monitoring device based on an MQTT protocol, and forward the data sent by the monitoring device to one or more second devices according to message identification of the data sent by the monitoring device, so as to configure or control the one or more second devices by the monitoring device.

17. The condition monitoring system of any one of claims 13 to 16, wherein, The data sent by at least one of the second devices and the monitoring device has message identification, and the message identification includes a plurality of hierarchical sub-identifications.

18. The state monitoring system according to claim 17, wherein the plurality of hierarchical sub-identifications include type of device, model of device, customer, number of device, and topic of message; or the plurality of hierarchical sub-identifications include type of device, model of device, customer, number of device, and topic of message, and the topic of message included in the message identification of the data sent by the second device is any one of state, alarm, and maintenance, and the topic of message included in the message identification of the data sent by the monitoring device is any one of control and configuration.

19. The condition monitoring system of any one of claims 13 to 18, wherein, The server is configured to: receive data sent by a plurality of second devices; determine data subscribed by the monitoring device from the data sent by the plurality of second devices according to subscription information of the monitoring device; send the data subscribed by the monitoring device to the monitoring device, so as to enable the monitoring device to monitor the subscribed devices.

20. The state monitoring system according to any one of claims 17 to 19, wherein The server is configured to determine a message identification matching the subscription information of the monitoring device, wherein for any level in the plurality of levels, the message identification is determined to match the subscription information of the monitoring device if a sub-identification of the level of the message identification matches a value of the level of the subscription information of the monitoring device; send data corresponding to the message identification to the monitoring device; or The server is configured to determine one or more second devices matching a message identification of the data sent by the monitoring device; send the data sent by the monitoring device to the one or more second devices.

21. The condition monitoring system of claim 20, wherein, a value of one or more levels in the subscription information is a first wildcard, and for any level in the plurality of levels, the value of the level in the subscription information matches any sub-identification if the value of the level in the subscription information is the first wildcard; or a last level in the subscription information is a second wildcard, wherein the second wildcard indicates that a level of the second wildcard in the subscription information and sub-levels of the second wildcard match any sub-identification.

Images