Early warning processing methods, devices, and computer-readable storage media for mechanical equipment

Abstract

This application discloses a method for early warning processing of mechanical equipment, comprising: acquiring target working data of the target mechanical equipment within a preset time period; based on the construction objects and construction elements corresponding to each unit time period, performing predictive processing through virtual body models of machine groups corresponding to different construction objects to obtain a first benchmark value of the target mechanical equipment's early warning indicator for operation of the construction objects corresponding to each unit time period; obtaining a test statistic of the target mechanical equipment's operation of the construction objects corresponding to each unit time period based on the deviation between the first benchmark value and the first actual value of the target mechanical equipment's early warning indicator for operation of the construction objects corresponding to each unit time period; and, when the target test statistic is greater than or equal to a preset corresponding abnormal alarm threshold, executing an abnormal alarm prompt for the early warning indicator under the construction object corresponding to the target unit time period. This method enables rapid and effective abnormal early warning, improving the accuracy of early warnings.

Description

Technical Field

[0001] This invention relates to the field of mechanical equipment technology, and in particular to a method, apparatus, and computer-readable storage medium for early warning processing of mechanical equipment. Background Technology

[0002] With technological advancements and development, the stability and reliability of mechanical equipment are gradually improving. However, due to the uncertainties of the construction process, the objects being constructed, and the construction environment, mechanical equipment often encounters anomalies during construction. Therefore, to improve the efficiency and service life of mechanical equipment, timely and accurate early warning of anomalies is necessary. In conceiving and implementing this application, the inventors discovered at least the following problems: existing early warning methods for mechanical equipment can only compare data under all construction objects simultaneously, failing to provide rapid and effective anomaly warnings, and their accuracy rate is low.

[0003] The preceding description is intended to provide general background information and does not necessarily constitute prior art. Summary of the Invention

[0004] The purpose of this application is to provide a method, device, and computer-readable storage medium for early warning processing of mechanical equipment, which can quickly and effectively provide early warning of abnormalities and improve the accuracy of early warning.

[0005] To achieve the above objectives:

[0006] In a first aspect, embodiments of this application provide a method for early warning processing of mechanical equipment, the method comprising:

[0007] S101. Obtain the target work data of the target mechanical equipment within a preset time period. The target work data includes the first actual value of the construction object, construction element and warning indicator corresponding to each unit time period of the target mechanical equipment within the preset time period.

[0008] S102. Based on the construction objects and construction elements corresponding to the target mechanical equipment in each unit time period, predictive processing is performed through the trained virtual body models of the machine group corresponding to different construction objects to obtain the first benchmark value of the warning indicator for the target mechanical equipment during the operation of the construction objects corresponding to the target mechanical equipment in each unit time period.

[0009] S103. Based on the deviation between the first benchmark value and the first actual value of the warning indicator for the target mechanical equipment during the operation of the construction object corresponding to each unit time period, obtain the inspection statistics for the target mechanical equipment during the operation of the construction object corresponding to each unit time period.

[0010] S104. When the target test statistic is greater than or equal to the preset corresponding abnormal alarm threshold, execute the abnormal alarm prompt for the construction object under the target unit time period, where the target unit time period is the unit time period corresponding to the target test statistic.

[0011] Optionally, the target work data may also include the second actual value of the target mechanical equipment's warning indicator within a preset time period;

[0012] The method further includes:

[0013] The detection method checks whether the second actual value of the warning indicator of the target mechanical equipment within a preset time period is less than or equal to the preset first abnormal warning threshold, or greater than or equal to the preset second abnormal warning threshold. The first abnormal warning threshold and the second abnormal warning threshold are determined based on the distribution of the warning indicators of the same type of mechanical equipment group within a historical preset time period.

[0014] If so, proceed to step S102.

[0015] Optionally, the method further includes:

[0016] Obtain the historical actual values ​​of the warning indicators for each piece of machinery in a group of the same type of machinery within a preset historical time period;

[0017] The distribution of the warning indicators of each piece of machinery and equipment is determined by fitting the historical actual values ​​of the warning indicators of the machinery and equipment group within the historical preset time period.

[0018] Based on the distribution of indicators to be warned in the mechanical equipment cluster within a historical preset time period, determine the first abnormal warning threshold and / or the second abnormal warning threshold.

[0019] Optionally, before step S102, the method further includes:

[0020] Obtain historical working data of each piece of machinery in a group of the same type of machinery within a preset historical time period. The historical working data includes the third actual value of the construction object, construction element and warning indicator corresponding to each unit time period of each piece of machinery within the preset historical time period.

[0021] Using the indicators to be warned as the dependent variable and the construction elements as the independent variables, virtual models of machine groups corresponding to different construction objects are constructed and trained based on historical work data.

[0022] Optionally, the target work data also includes the fourth actual value of the IPO parameter and the indicator to be warned for the target machinery and equipment completing at least one sub-action within the target unit time period, respectively. The IPO parameter is used to describe the function and construction process information of the target machinery and equipment in completing the corresponding sub-action; the method further includes:

[0023] Based on the IPO parameters corresponding to the completion of at least one action by the target mechanical equipment within the target unit time period, the second benchmark value of the warning indicator corresponding to the completion of at least one action by the target mechanical equipment within the target unit time period is obtained by predicting through the sub-action regression model corresponding to different sub-actions that has been trained.

[0024] Based on the deviation between the fourth actual value and the second benchmark value of the warning indicator corresponding to each of the target mechanical equipment completing at least one action within the target unit time period, the abnormal probability corresponding to each of the at least one action is obtained.

[0025] Identify target sub-actions with an abnormal probability greater than or equal to a preset probability threshold, and issue warnings for the target sub-actions.

[0026] Optionally, the method further includes:

[0027] Obtain historical working data of each piece of machinery in a group of the same type of machinery within a preset historical time period. The historical working data includes the IPO parameters and the fifth actual value of the warning indicator corresponding to each piece of machinery completing at least one action.

[0028] Using the indicator to be warned as the dependent variable and the IPO parameter as the independent variable, regression models for different sub-actions are constructed and trained based on historical working data.

[0029] Optionally, the method further includes:

[0030] Feature extraction is performed on the IPO parameters corresponding to the target sub-actions to obtain the IPO feature parameters corresponding to the target sub-actions;

[0031] The root cause localization algorithm is used to identify the IPO feature parameters corresponding to the target sub-actions, and to determine and output the abnormal feature parameters.

[0032] Optionally, the method further includes:

[0033] Based on abnormal feature parameters, combined with the equipment data and historical fault data of the target mechanical equipment, fault diagnosis is performed to determine and output corresponding maintenance suggestions.

[0034] Secondly, embodiments of this application provide a warning processing device for mechanical equipment, comprising: a processor and a memory storing a computer program, wherein when the processor runs the computer program, the steps of the aforementioned warning processing method for mechanical equipment are implemented.

[0035] Thirdly, embodiments of this application provide a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the aforementioned early warning processing method for mechanical equipment.

[0036] The present application provides a method, apparatus, and computer-readable storage medium for early warning processing of mechanical equipment. The method includes: acquiring target work data of the target mechanical equipment within a preset time period, the target work data including the construction object, construction element, and first actual value of the indicator to be warned corresponding to each unit time period of the target mechanical equipment within the preset time period; based on the construction object and construction element corresponding to the target mechanical equipment in each unit time period, performing prediction processing through a trained virtual machine group model corresponding to different construction objects to obtain a first benchmark value of the indicator to be warned when the target mechanical equipment is operating in each unit time period; based on the deviation between the first benchmark value and the first actual value of the indicator to be warned when the target mechanical equipment is operating in each unit time period, obtaining a test statistic of the construction object corresponding to the target mechanical equipment in each unit time period; and when the target test statistic is greater than or equal to a preset corresponding abnormal alarm threshold, executing an abnormal alarm prompt for the indicator to be warned under the construction object corresponding to the target unit time period, where the target unit time period is the unit time period corresponding to the target test statistic. In this way, by combining virtual swarm modeling with statistical hypothesis testing, mechanical equipment can be compared under the same construction conditions to identify potential warning indicators, thereby enabling rapid and effective anomaly warnings and improving the accuracy of warnings. Attached Figure Description

[0037] Figure 1 A flowchart illustrating the early warning processing method for mechanical equipment provided in an embodiment of the present invention;

[0038] Figure 2 A schematic diagram of the architecture of the early warning processing method for mechanical equipment provided in an embodiment of the present invention;

[0039] Figure 3 This is a schematic diagram of the process of the early warning processing method for mechanical equipment provided in an embodiment of the present invention;

[0040] Figure 4 This is a schematic diagram of the daily fuel consumption rate distribution of the fleet in an embodiment of the present invention;

[0041] Figure 5 This is a schematic diagram comparing the virtual body of the machine cluster and the physical hole of a single machine in an embodiment of the present invention;

[0042] Figure 6 This is a schematic diagram illustrating the principle of deviation calculation in an embodiment of the present invention;

[0043] Figure 7This is a schematic diagram of action anomaly analysis in an embodiment of the present invention;

[0044] Figure 8 This is a schematic diagram illustrating the push of maintenance suggestions in an embodiment of the present invention. Detailed Implementation

[0045] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0046] See Figure 1 This application provides a method for early warning processing of mechanical equipment. This method can be executed by a device for early warning processing of mechanical equipment, which can be implemented using software and / or hardware. Specifically, the device can be mechanical equipment, electronic equipment, a server, etc. In this embodiment, the server is used as the executing entity for the method. The method for early warning processing of mechanical equipment provided in this embodiment includes:

[0047] Step S101: Obtain the target working data of the target machinery and equipment within a preset time period. The target working data includes the first actual values ​​of the construction object, construction elements, and warning indicators corresponding to each unit time period of the target machinery and equipment within the preset time period.

[0048] Optionally, the target machinery and equipment is equipment that requires abnormal early warning and diagnosis, including but not limited to engineering machinery and equipment such as rotary drilling rigs, agricultural machinery and equipment, hoisting machinery and equipment, and mining vehicles. Optionally, the target operating data can be collected by the target machinery and equipment through sensors or other devices installed on itself, such as obtaining fuel consumption rate by collecting oil level data through a liquid level sensor, and then transmitting it to the server via a wireless communication network.

[0049] It is understood that the preset time period is a pre-set time period for sampling target working data. Specifically, it can be set based on information such as the working hours of the target machinery, for example, it can be set to 7:00 AM to 6:00 PM within a day, or 8:00 AM to 10:00 PM within a week. The unit time period can be set based on actual needs; in this embodiment, a unit time period of 1 hour is used as an example. The warning indicator can be used to indicate the working performance or operating parameters of the machinery within a unit time period, and can be set based on actual needs, including but not limited to fuel consumption rate, work efficiency, etc. Furthermore, for different types of machinery, the same warning indicator can represent different meanings. For example, taking work efficiency as an example, if the target machinery is a rotary drilling rig, the corresponding work efficiency is the number of meters drilled; if the target machinery is a concrete pump truck, the corresponding work efficiency is the volume of concrete pumped, etc.

[0050] It should be noted that the corresponding construction objects and construction elements may differ depending on the type of target machinery and equipment and the construction requirements. Taking a rotary drilling rig as an example, the construction object may be rock, soil, soft soil, etc., and the construction elements may include at least one of the following: geological type, pile diameter, and hole depth. Within a preset time period, the target machinery and equipment may have one or more construction objects; no specific limitation is made here. In this embodiment, we take the example of the target machinery and equipment corresponding to only one construction object within a unit time period. Furthermore, if the construction elements change within a unit time period, the average construction elements within that unit time period are obtained.

[0051] Step S102: Based on the construction objects and construction elements corresponding to the target mechanical equipment in each unit time period, predictive processing is performed through the trained virtual body models of the machine group corresponding to different construction objects to obtain the first benchmark value of the warning indicator when the target mechanical equipment is operating in each unit time period.

[0052] Specifically, for the construction object and construction elements corresponding to the target mechanical equipment in each unit time period, a virtual body model of the machine group corresponding to the construction object can be determined first based on the construction object in that unit time period. Then, the construction elements corresponding to that unit time period are input into the virtual body model of the machine group corresponding to the construction object for prediction processing, and the first benchmark value of the warning index of the target mechanical equipment when it is operating on the construction object in that unit time period is obtained. It should be noted that the first benchmark value is used to characterize the value of the warning index predicted by the virtual body model of the machine group when the target mechanical equipment is operating on a construction object.

[0053] The virtual machine cluster model can be used to find the correlation between the warning indicators and construction elements under the same construction object, facilitating comparative analysis of the warning indicators under the same construction object, and further verifying the preliminary warning based on distribution estimation, thus eliminating the influence of the construction object on the warning indicators. It can be understood that if the target machinery is in operation throughout a preset time period, the number of first benchmark values ​​obtained is the same as the number of unit time periods contained in the preset time period. Taking a preset time period from 8:00 AM to 6:00 PM as an example, if the target machinery is working on object A from 8:00 AM to 12:00 PM, and on object B from 12:00 PM to 6:00 PM, by inputting the construction elements corresponding to each hour from 8:00 AM to 12:00 PM into the virtual body model of the machinery cluster for object A, the first baseline values ​​of the warning indicators corresponding to each hour of the target machinery cluster from 8:00 AM to 12:00 PM can be obtained, i.e., 4 first baseline values. Similarly, by inputting the construction elements corresponding to each hour from 12:00 PM to 6:00 PM into the virtual body model of the machinery cluster for object B, the first baseline values ​​of the warning indicators corresponding to each hour of the target machinery cluster from 12:00 PM to 6:00 PM can be obtained, i.e., 6 first baseline values. Furthermore, the virtual body model of the machinery cluster can be built based on algorithms such as multiple linear regression, logistic regression, support vector regression, generalized linear regression, random forest, and neural networks, without specific limitations here.

[0054] Step S103: Based on the deviation between the first benchmark value and the first actual value of the warning indicator of the target mechanical equipment during the operation of the construction object corresponding to each unit time period, obtain the inspection statistics of the target mechanical equipment during the operation of the construction object corresponding to each unit time period.

[0055] It is understandable that, since each unit time period corresponds to a first actual value and a first benchmark value of the indicator to be warned, meaning the number of first actual values ​​and first benchmark values ​​is equal to the number of unit time periods, a deviation vector D composed of multiple deviations can be obtained based on the deviation between the first actual value and the first benchmark value of the indicator to be warned corresponding to the target machinery in each unit time period. Combined with the standard error of the virtual body model of the machine group corresponding to different construction objects, the test statistic of the target machinery during the operation of the construction object corresponding to each unit time period can be calculated based on the statistical hypothesis test of the deviation. Among them, the test statistic S of the target machinery during the operation of the construction object corresponding to each unit time period... + The calculation formula is as follows:

[0056] and

[0057] Where n′ is the non-zero logarithm, i.e., sign(D) i The number of ) Di d represents the deviation in the deviation vector D, that is, the deviation between the first baseline value and the first actual value of the indicator to be warned corresponding to the i-th unit time period. i Let represent the standard error of the virtual machine group model corresponding to the construction object in the i-th unit time period, and b be a multiple of the standard error, usually a constant, ranging from 0 to 3. The standard error of the virtual machine group model characterizes the prediction accuracy of the model; the smaller the standard error, the higher the prediction accuracy. It should be noted that when calculating the verification statistics of the target machinery during the operation of the construction object in any unit time period within the preset time period, it is necessary to consider the deviation between the first baseline value and the first actual value of the warning indicator corresponding to other unit time periods preceding that unit time period. Furthermore, since the warning indicator of the target machinery usually shows anomalies in subsequent unit time periods after anomalies in earlier unit time periods, it is not necessary to obtain the verification statistics of the target machinery during the operation of the construction object in the first few unit time periods within the preset time period, or the verification statistics of the target machinery during the operation of the construction object in the first few unit time periods within the preset time period can be set to preset values.

[0058] Taking a preset time period of 6:00 AM to 6:00 PM within a day, with each time period measured in hours, and the target machinery being a rotary drilling rig, as an example, based on the first actual values ​​of the construction object, construction elements, and warning indicators corresponding to each unit time period within the aforementioned time period, the first benchmark value of the warning indicators for the rotary drilling rig during operation of the construction object corresponding to each unit time period can be obtained. Combined with the first actual value of the warning indicators, a deviation vector composed of the deviation between the first benchmark value and the first actual value of the warning indicators corresponding to different consecutive unit time periods can be obtained. For example, the deviation vector composed of the deviation between the first benchmark value and the first actual value of the warning indicators corresponding to each unit time period from 6:00 AM to 1:00 PM, the deviation vector composed of the deviation between the first benchmark value and the first actual value of the warning indicators corresponding to each unit time period from 6:00 AM to 2:00 PM, the deviation vector composed of the deviation between the first benchmark value and the first actual value of the warning indicators corresponding to each unit time period from 6:00 AM to 3:00 PM, etc. At this point, if it is necessary to calculate the inspection statistic for the construction object operation corresponding to the unit time period at 2 PM, it can be obtained based on the deviation vector composed of the deviation between the first benchmark value and the first actual value of the indicator to be warned for each unit time period from 6 AM to 1 PM, using the aforementioned formula for calculating the inspection statistic. Similarly, if it is necessary to calculate the inspection statistic for the construction object operation corresponding to the unit time period at 3 PM, it can be obtained based on the deviation vector composed of the deviation between the first benchmark value and the first actual value of the indicator to be warned for each unit time period from 6 AM to 2 PM, using the aforementioned formula for calculating the inspection statistic.

[0059] Step S104: When the target test statistic is greater than or equal to the preset corresponding abnormal alarm threshold, execute the abnormal alarm prompt for the construction object under the target unit time period, where the target unit time period is the unit time period corresponding to the target test statistic.

[0060] The corresponding anomaly alarm thresholds may differ for different time periods, and each time period's corresponding anomaly alarm threshold can be obtained based on the bias vector used when calculating the test statistic for that time period. Here, for any given time period, the corresponding test statistic will follow a binomial distribution with parameters n′ and 0.5. By combining the set significance level 'a', the corresponding anomaly alarm threshold C can be determined. It is understandable that, in the process of calculating the inspection statistics of the target mechanical equipment during the operation of the construction object in any unit time period within the preset time period, it is necessary to combine the deviation between the first benchmark value and the first actual value of the indicator to be warned corresponding to other unit time periods before that unit time period. That is, the number of deviations may affect the size of the abnormal alarm threshold.

[0061] It is understandable that, for the inspection statistics of the target machinery and equipment during the operation of the corresponding construction object in each unit time period, it can be determined whether the inspection statistics are greater than or equal to the preset corresponding abnormal alarm threshold. If there is a target inspection statistic greater than or equal to the preset corresponding abnormal alarm threshold, it is considered that the indicator to be warned is abnormal within the unit time period corresponding to the target inspection statistic, i.e., the target unit time period. Therefore, an abnormal alarm prompt can be executed for the indicator to be warned under the construction object corresponding to the target unit time period. It should be noted that executing the abnormal alarm prompt for the indicator to be warned under the construction object corresponding to the target unit time period can output an abnormal alarm prompt message, which is used to indicate that the indicator to be warned is behaving abnormally under the construction object corresponding to the target unit time period. In addition, if the inspection statistics of the target machinery and equipment during the operation of the construction object in a unit time period are less than the preset corresponding abnormal alarm threshold, it means that the indicator to be warned is behaving normally under the construction object corresponding to that unit time period.

[0062] In summary, the early warning processing method for mechanical equipment provided in the above embodiments combines virtual body modeling of machine groups with statistical hypothesis testing, enabling mechanical equipment to compare the indicators to be warned under the same construction object, thus enabling rapid and effective early warning of anomalies and improving the accuracy of early warning.

[0063] In one embodiment, the target work data also includes a second actual value of the target mechanical equipment's warning indicator within a preset time period;

[0064] The method further includes:

[0065] The detection method checks whether the second actual value of the warning indicator of the target mechanical equipment within a preset time period is less than or equal to the preset first abnormal warning threshold, or greater than or equal to the preset second abnormal warning threshold. The first abnormal warning threshold and the second abnormal warning threshold are determined based on the distribution of the warning indicators of the same type of mechanical equipment group within a historical preset time period.

[0066] If so, proceed to step S102.

[0067] Specifically, after obtaining the second actual value of the indicator to be warned for the target mechanical equipment within a preset time period, an initial abnormal warning detection can be performed on the indicator to be warned. That is, it can be detected whether the second actual value of the indicator to be warned for the target mechanical equipment within a preset time period is less than or equal to the preset first abnormal warning threshold, or greater than or equal to the preset second abnormal warning threshold. If the second actual value is less than or equal to the preset first abnormal warning threshold, or greater than or equal to the preset second abnormal warning threshold, it indicates that an initial abnormal warning has been detected for the indicator to be warned. Then, step S102 is executed to perform false alarm filtering. Otherwise, the processing can be terminated.

[0068] The first anomaly warning threshold is lower than the second anomaly warning threshold, and both thresholds are determined based on the distribution of warning indicators for the same type of machinery group within a historical preset time period. For example, if the preset time period is 24 hours and the warning indicator is fuel consumption rate, the second actual value of the warning indicator for the target machinery within the preset time period is the total fuel consumption rate of the target machinery within 24 hours. By combining the first and second anomaly warning thresholds determined by the historical distribution of total fuel consumption rates for the same type of machinery group within 24 hours, it can be determined whether the total fuel consumption rate of the target machinery within 24 hours is less than or equal to the preset first anomaly warning threshold, or greater than or equal to the preset second anomaly warning threshold. Thus, by performing initial anomaly warning detection based on the distribution of warning indicators for the same type of machinery group within a historical preset time period, initial anomaly warnings are selected for further processing, improving the efficiency of warning processing.

[0069] In one embodiment, the method further includes:

[0070] Obtain the historical actual values ​​of the warning indicators for each piece of machinery in a group of the same type of machinery within a preset historical time period;

[0071] The distribution of the warning indicators of each piece of machinery and equipment is determined by fitting the historical actual values ​​of the warning indicators of the machinery and equipment group within the historical preset time period.

[0072] Based on the distribution of indicators to be warned in the mechanical equipment cluster within a historical preset time period, determine the first abnormal warning threshold and / or the second abnormal warning threshold.

[0073] It is understandable that the historical actual values ​​of the warning indicators for each piece of machinery in a group of the same type of machinery within a preset historical time period can reflect the normal and common usage of the warning indicators during the operation of each piece of machinery. Therefore, the distribution of the historical actual values ​​of the warning indicators for each piece of machinery within a preset historical time period can be fitted first to determine the distribution of the warning indicators for the group of machinery within the preset historical time period. Then, based on the distribution of the warning indicators for the group of machinery within the preset historical time period, the first abnormal warning threshold and / or the second abnormal warning threshold can be determined. Initial abnormal warning detection can then be performed on the warning indicators based on the first abnormal warning threshold and / or the second abnormal warning threshold, thereby improving the accuracy of the initial abnormal warning.

[0074] In one embodiment, before step S102, the method further includes:

[0075] Obtain historical working data of each piece of machinery in a group of the same type of machinery within a preset historical time period. The historical working data includes the third actual value of the construction object, construction element and warning indicator corresponding to each unit time period of each piece of machinery within the preset historical time period.

[0076] Using the indicators to be warned as the dependent variable and the construction elements as the independent variables, virtual models of machine groups corresponding to different construction objects are constructed and trained based on historical work data.

[0077] Optionally, to find the correlation between the warning indicators and construction elements under the same construction object, facilitate comparative analysis of the warning indicators under the same construction object, further verify the above-mentioned abnormal warnings, and eliminate the influence of the construction object on the warning indicators, the warning indicators can be used as the dependent variable and the construction elements as the independent variables. Based on the historical working data of each piece of machinery in the same type of machinery group within a preset historical time period, virtual machinery group models corresponding to different construction objects can be constructed and trained. In other words, for each type of construction object, a corresponding virtual machinery group model is established to facilitate subsequent comparative analysis of the warning indicators under the same construction object.

[0078] In one embodiment, the target work data further includes the fourth actual value of the IPO parameter and the indicator to be warned, corresponding to the target mechanical equipment completing at least one sub-action within the target unit time period. The IPO parameter is used to describe the function and construction process information of the target mechanical equipment in completing the corresponding sub-action. The method further includes:

[0079] Based on the IPO parameters corresponding to the completion of at least one action by the target mechanical equipment within the target unit time period, the second benchmark value of the warning indicator corresponding to the completion of at least one action by the target mechanical equipment within the target unit time period is obtained by predicting through the sub-action regression model corresponding to different sub-actions that has been trained.

[0080] Based on the deviation between the fourth actual value and the second benchmark value of the warning indicator corresponding to each of the target mechanical equipment completing at least one action within the target unit time period, the abnormal probability corresponding to each of the at least one action is obtained.

[0081] Identify target sub-actions with an abnormal probability greater than or equal to a preset probability threshold, and issue warnings for the target sub-actions.

[0082] Optionally, when the indicator to be warned exhibits abnormal behavior under the construction object corresponding to the target unit time period, since the target machinery and equipment will perform multiple sub-actions during operation under the target unit time period, it is necessary to further determine the specific abnormal sub-actions. Therefore, based on the IPO parameters corresponding to each of the target machinery and equipment completing at least one sub-action within the target unit time period, prediction processing can be performed using a pre-trained regression model corresponding to each sub-action to obtain the second baseline value of the indicator to be warned corresponding to each of the target machinery and equipment completing at least one sub-action within the target unit time period. Then, based on the deviation between the fourth actual value of the indicator to be warned corresponding to each of the target machinery and equipment completing at least one sub-action within the target unit time period and the second baseline value, the anomaly probability corresponding to each of the at least one sub-action can be obtained. IPO stands for Input-Process-Output, and IPO parameters can specifically describe the functions of the machinery and equipment and the information and data of the construction process.

[0083] Among them, it can be based on the formula Calculate the deviation between the fourth actual value and the second benchmark value of the indicator to be warned for at least one action, and according to the formula... Calculate the anomalous probability corresponding to at least one action. Here, Z i This represents the deviation corresponding to the i-th sub-action. σ represents the probability of an anomaly corresponding to the i-th sub-action. i This represents the standard deviation of the regression model for the i-th sub-action. It can be understood that since the target machinery may repeat the same sub-action multiple times within the target time period, multiple deviations may be obtained for that sub-action. In this case, the average of these multiple deviations can be used to calculate the corresponding anomaly probability.

[0084] It should be noted that the corresponding sub-actions may differ for different types of target machinery. For example, taking a rotary drilling rig as the target machinery, the corresponding sub-actions may include lowering, drilling, lifting, rotating, soil dumping, and traveling. The preset probability threshold can be set according to actual needs, such as 65% or 70%, without specific limitations here. When the abnormal probability of a target sub-action is greater than or equal to the preset probability threshold, it indicates that the target sub-action may be abnormal, and an early warning is issued for the target sub-action, such as outputting an early warning message that may include information such as the abnormal probability of the target sub-action. In this way, the abnormal probability of each sub-action during construction can be comprehensively identified, facilitating the source analysis of abnormal alarms and improving the quality and efficiency of early warnings.

[0085] In one embodiment, the method further includes:

[0086] Obtain historical working data of each piece of machinery in a group of the same type of machinery within a preset historical time period. The historical working data includes the IPO parameters and the fifth actual value of the warning indicator corresponding to each piece of machinery completing at least one action.

[0087] Using the indicator to be warned as the dependent variable and the IPO parameter as the independent variable, regression models for different sub-actions are constructed and trained based on historical working data.

[0088] Optionally, to find the correlation between the warning indicator and the IPO parameter under the same sub-action, and to facilitate comparative analysis of the warning indicator under the same sub-action, the warning indicator can be used as the dependent variable and the IPO parameter as the independent variable. Based on the historical working data of each piece of machinery in the same type of machinery group within a preset historical time period, sub-action regression models corresponding to different sub-actions can be constructed and trained. That is, for each sub-action, a corresponding sub-action regression model is established to facilitate subsequent comparative analysis of the warning indicator under the same sub-action. The IPO parameters for different sub-actions may vary. For example, for a walking action, the corresponding IPO parameters may include distance and duration; for lifting and lowering actions, the corresponding IPO parameters may include distance, duration, motor pressure, etc.

[0089] In one embodiment, the method further includes:

[0090] Feature extraction is performed on the IPO parameters corresponding to the target sub-actions to obtain the IPO feature parameters corresponding to the target sub-actions;

[0091] The root cause localization algorithm is used to identify the IPO feature parameters corresponding to the target sub-actions, and to determine and output the abnormal feature parameters.

[0092] It is understandable that after identifying an abnormal target sub-action, it may be necessary to further determine the abnormal IPO parameters within the target sub-action to pinpoint the cause of the anomaly. Therefore, feature extraction can be performed on the IPO parameters corresponding to the target sub-action to obtain the corresponding IPO feature parameters. Next, a root cause localization algorithm is used to identify the IPO feature parameters corresponding to the target sub-action, determining and outputting the abnormal feature parameters. Here, the root cause localization algorithm includes, but is not limited to, random walk, PageRank, Pearson correlation coefficient, and depth-first search methods. Specifically, the root cause localization algorithm can first determine the correlation between the indicator to be warned and the IPO feature parameters, and then identify the abnormal feature parameters from the IPO feature parameters based on the correlation information. Furthermore, when outputting the abnormal feature parameters, a reference baseline value can also be output, which can be obtained based on the historical operating data of each piece of machinery in a group of machines of the same type.

[0093] For example, taking a rotary drilling rig as the target mechanical equipment, drilling as the target sub-action, and fuel consumption rate as the indicator to be warned, the following steps are taken: First, the fuel consumption rate of the drilling action and related IPO parameters such as engine gear, torque, engine mode, main pump pressure, power head A / B port pressure, system flow rate, and system pressure loss are extracted. Then, the time-domain and frequency-domain characteristics of the IPO parameters, as well as fluctuation and impact indicators, are extracted. Finally, an abnormal characteristic parameter is identified from the extracted IPO feature parameters using a root cause localization algorithm, and a reference benchmark value is provided. For example, an abnormal characteristic parameter might be that the engine gear is at 7th gear, while the reference benchmark value is 5th gear; or an abnormal characteristic parameter might be that the system pressure difference is large, while the reference benchmark value is 20 bar less. In this way, by obtaining the abnormal characteristic parameters from the IPO feature parameters corresponding to the target sub-action, abnormal alarm source tracing analysis can be achieved, improving the efficiency and quality of early warning.

[0094] In one embodiment, the method further includes:

[0095] Based on abnormal feature parameters, combined with the equipment data and historical fault data of the target mechanical equipment, fault diagnosis is performed to determine and output corresponding maintenance suggestions.

[0096] Specifically, after determining the abnormal characteristic parameters, machine learning or deep learning algorithms can be used to analyze the equipment data and historical fault data of the target machinery. Expert experience can also be incorporated to diagnose the faults in the target machinery. Once the corresponding faults are identified, appropriate maintenance recommendations can be determined and output. The equipment data can include operating condition data and maintenance data. In this way, by recommending corresponding maintenance suggestions, reliance on the technical skills of service personnel can be reduced, thus lowering maintenance service costs.

[0097] Based on the same inventive concept as the foregoing embodiments, the foregoing embodiments will be described in detail below through a specific example. The mechanical equipment in this embodiment can be engineering machinery equipment.

[0098] In existing technologies, when mechanical equipment encounters malfunctions and requires troubleshooting, service personnel typically obtain firsthand information by questioning the machine operator. However, this method is time-consuming, and the information received lacks sufficient accuracy, versatility, and granularity to support comprehensive and accurate judgments, hindering the identification of the root cause of the malfunction and compromising the quality and efficiency of subsequent problem handling. Furthermore, if mechanical equipment malfunctions are not detected promptly, it can lead to passive customer complaints, negatively impacting product reputation and service efficiency.

[0099] To address the above problems, this embodiment provides a method for early warning processing of mechanical equipment, see reference. Figure 2 and Figure 3 The early warning processing method for mechanical equipment provided in this embodiment mainly includes the following processes:

[0100] (I) Distribution Estimation

[0101] The distribution estimation includes two parts: historical distribution estimation based on the cluster and anomaly warning based on the distribution.

[0102] (1) Estimation based on the historical distribution of the cluster.

[0103] Based on the identified warning indicators for mechanical equipment (including but not limited to fuel consumption and work efficiency), the daily distribution of warning indicators for a group of mechanical equipment of the same type is estimated. This distribution can be used for initial warnings of the warning indicators. Furthermore, the distribution estimation can employ statistical distributions such as normal distribution, exponential distribution, and Weibull distribution. The distribution can also be updated periodically, triggered by regular events or events such as improvements in the model or quality of the mechanical equipment.

[0104] (2) Distribution-based anomaly warning.

[0105] Based on the estimated historical distribution of the machinery and equipment fleet, early warning rules, i.e., abnormal early warning thresholds, can be determined for the indicators to be warned regarding the machinery and equipment. If the historical distribution approximates a normal distribution, the early warning rule can adopt the principle of the normal distribution mean m±nσ, that is, the second abnormal early warning threshold is m+nσ and the first abnormal early warning threshold is m-nσ, where m is the normal distribution mean, σ is the standard deviation of the normal distribution, and n is generally taken as 2 or 3. If the historical distribution is other distributions, the second abnormal early warning threshold can be determined based on quantiles, for example, the 99th or 99.5th quantile can be used as the second abnormal early warning threshold. In addition, the determination of the abnormal early warning threshold can also be combined with the business meaning of the indicator to be warned. After determining the abnormal early warning threshold, anomaly detection can be performed on the target working data of the machinery and equipment within a preset time period, that is, to detect whether the actual value of the indicator to be warned by the machinery and equipment within the preset time period is greater than the second abnormal early warning threshold or less than the first abnormal early warning threshold. If so, it is determined that the machinery and equipment has an initial warning anomaly.

[0106] Taking the fuel consumption rate (also known as oil consumption rate) as the indicator to be warned, the daily fuel consumption rate distribution of the rig group can be calculated based on historical daily fuel consumption rate data of the same type of rotary drilling rig group, such as... Figure 4 As shown, based on the daily fuel consumption rate distribution of the fleet, the daily fuel consumption rate can be warned using the m±3σ principle. That is, when the daily fuel consumption rate is greater than or equal to m+3σ or less than or equal to m-3σ, an abnormal warning is issued.

[0107] (II) False Alarm Filtering

[0108] (1) Constructing a virtual machine cluster model

[0109] Constructing a virtual cluster model is crucial for false alarm filtering. The indicator to be warned, y, is used. i The dependent variable is the index of the construction object, i.e., construction elements x1, x2, ..., x3. p Using y as the independent variable, a regression model is established as a virtual entity model of the cluster: i = f(x1,x2,…,x) p The purpose of establishing this regression model is to find the correlation between the indicators to be warned and the indicators of the construction object under the same construction object, so as to facilitate comparative analysis of the indicators to be warned under the same construction object, and to further verify the above-mentioned abnormal warnings, eliminating the influence of the construction object on the indicators to be warned. The methods for establishing the regression model include, but are not limited to: multiple linear regression, logistic regression, support vector regression, generalized linear regression, random forest, neural networks, etc. It should be noted that the indicators to be warned here are the average indicators to be warned over the hourly period.

[0110] For example, taking a rotary drilling rig as an example, since the main function of a rotary drilling rig is drilling operations, and the construction object is rock and soil, the fuel consumption rate can be the dependent variable, and construction factors such as geology (e.g., clay, sand, strongly weathered, moderately weathered, slightly weathered, etc.), borehole diameter, and borehole depth can be used as independent variables. A regression algorithm is then used to construct a virtual model of the rig cluster, specifically: y i = f(x1,x2,x3). Where x1,x2,x3 are the geological features, borehole diameter, and borehole depth, respectively, and y i Let f be the fuel consumption rate, and f be the virtual entity model of the fleet. For example... Figure 5 The diagram shown is a comparison between a virtual swarm object and a physical hole of a single machine. The virtual swarm object is generated based on the virtual swarm object model.

[0111] (2) Deviation calculation based on cluster virtual body model

[0112] After establishing the virtual model of the machine cluster, the deviation between the actual value of the warning indicator for a single machine entity (i.e., a single piece of machinery) and the baseline value predicted by the virtual model can be calculated. The magnitude of this deviation indicates the degree of deviation between the actual value of the warning indicator for a single machine entity and the baseline value predicted by the virtual model for the same construction object. The deviation can be expressed by the formula... Perform calculations, y i This is the actual value for a single machine. As the baseline value for the aircraft group, the principle of deviation calculation is as follows: Figure 6 As shown.

[0113] Taking the calculation of the fuel consumption rate deviation of rotary drilling rigs as an example, firstly, the actual geological conditions of a single hole (x1), hole diameter (x2), and hole depth (x3) are input into the virtual body model of the rig group to obtain its fuel consumption rate benchmark value. Then, subtract the reference value of the fuel consumption rate from the actual fuel consumption rate of a single hole to obtain the deviation value D. i ,and

[0114] (3) Determination of deviations and anomalies based on sign test

[0115] After calculating the bias, the anomaly is further verified through bias-based statistical hypothesis testing. The specific approach and steps are as follows:

[0116] Test the null hypothesis: The data does not have a trend; alternative hypothesis: The data does have a trend.

[0117] The basic idea is: if the data is normal, the data distribution is consistent; if the distribution is inconsistent and the p-value is less than a pre-set threshold, the null hypothesis is rejected, the indicator to be warned is considered abnormal, and an alarm is issued. The specific steps are as follows:

[0118] 1) Construct the test statistic S + :

[0119] and

[0120] Where n′ is the non-zero logarithm, i.e., sign(D) i The number of ) , d i denoted as the standard error of the virtual body model of the machine group corresponding to the construction object in the i-th unit time period, where b is a multiple of the standard error and is usually a constant, with a value between 0 and 3.

[0121] At this point, under the original assumptions, S + It follows a binomial distribution with parameters n′ and 0.5.

[0122] 2) Calculate the alarm threshold C for abnormal alarms, and Where 'a' represents the significance level set.

[0123] 3) Determine whether the abnormal warning is a false alarm, i.e., if S + If the result is >C, it is considered that the indicator to be warned is abnormal under the construction object, and it is determined to be an abnormal alarm, and an abnormal action analysis is performed; otherwise, it is determined to be a false alarm, and the indicator to be warned is considered to be normal under the construction object.

[0124] (III) Analysis of abnormal movements

[0125] First, the work data of the mechanical equipment within a preset time period is broken down into sub-actions. Generally, for mechanical equipment to complete a specified construction function, it can be broken down into different sub-actions along the time dimension.

[0126] Next, for each of the decomposed sub-actions, calculate the indicators to be warned within that sub-action. and IPO parameters Where i = 1, 2, ..., r, and r is the number of different sub-actions.

[0127] Then, use the action to prepare for early warning indicators. IPO parameters are the dependent variable. Using as the independent variable, a sub-action regression model is established as follows:

[0128]

[0129] Here, f represents the sub-action regression model, which can be established using linear regression, Bayesian methods, neural networks, or ensemble algorithms. IPO parameters can specifically describe the functions of the mechanical equipment and information and data related to the construction process.

[0130] Finally, the IPO parameters within each sub-action are input into the corresponding sub-action model to calculate the deviation degree Z of the indicator to be warned within the sub-action. i And calculate the anomaly probability. Warnings are issued for actions with an anomaly probability exceeding a specified threshold (e.g., 0.995). The formulas for calculating the deviation degree (specifically, the deviation itself) and the anomaly probability are as follows:

[0131]

[0132]

[0133] Here, Z i This represents the deviation corresponding to the i-th sub-action. σ represents the probability of an anomaly corresponding to the i-th sub-action. i It represents the standard deviation of the sub-action regression model corresponding to the i-th sub-action.

[0134] Taking rotary drilling rigs as the mechanical equipment and fuel consumption rate as the indicator to be warned, after the fuel consumption rate warning judgment calculation is completed, the fuel consumption rate anomaly analysis of sub-actions is carried out through sub-action regression model to determine whether the sub-action is abnormal and push the anomaly probability of the sub-action. The specific process is as follows:

[0135] First, based on historical data, and after constraining the borehole diameter, depth, and geological conditions, the fuel consumption rate of each action can be determined. As the dependent variable, it is established based on different response variables. The action-based regression model is as follows. For the action-based regression model, such as... Figure 7As shown, the main single actions are first distinguished based on the segmentation of rotary drilling operations and the identification tags of action slices, such as lowering, drilling, lifting, slewing, soil throwing, and traveling. Then, a multivariate regression algorithm is used to construct a sub-action regression model for predicting the baseline of fuel consumption rate. In constructing the model, the fuel consumption rate of a single action is the dependent variable, and other variables of the single action are the independent variables, such as traveling action (distance, duration, etc.), lifting / lowering action (distance, duration, motor pressure, main winch wire rope tension, etc.), drilling action (pressure ratio, engine torque, engine speed, power head AB port pressure, drill bit position, drill rod speed, etc.), slewing action (angle, duration, etc.), and soil throwing action (frequency, duration, etc.).

[0136] Next, for work data identified as having abnormal daily fuel consumption rates, a sub-action regression model is input to calculate the fuel consumption rate of each sub-action. Based on the prediction results, the probability of abnormality for each sub-action is obtained, and then the abnormal action is located by the magnitude of the probability of abnormality. For example... Figure 7 As shown, if the abnormal probability threshold of the sub-action is set to 65%, then warnings will be issued for sub-actions such as lowering (70%), drilling (75%), and throwing soil (68%).

[0137] (iv) Locating the root causes of IPO parameter anomalies

[0138] For the sub-actions of the early warning, the IPO parameters and their characteristics are calculated within the sub-action to determine the abnormal characteristic parameters.

[0139] (1) IPO parameter definition and feature calculation

[0140] Taking action as the dimension, feature parameters of IPO parameters corresponding to different actions are extracted. Feature extraction refers to common feature extraction methods, such as time domain, frequency domain, time-frequency features, and features strongly related to business, such as fluctuation indicators, impact indicators, and response indicators.

[0141] (2) Locating Abnormal Parameters in IPOs

[0142] Here, the Pearson correlation coefficient method can be used to explore the correlation between the indicators to be warned and the IPO parameters. Then, the differences between the characteristic parameters of a single piece of machinery and the characteristic parameters of the group of machines are compared, and the characteristic parameters or parameter groups with high differences are output. At the same time, the baseline value can also be output. The root cause localization algorithm used includes, but is not limited to, random walk, PageRank, Pearson correlation coefficient, depth-first search and other algorithms.

[0143] Taking the root cause analysis of fuel consumption rate of rotary drilling rigs as an example: If, after the above steps, the sub-actions with a high probability of abnormality are identified as lifting, drilling, and soil ejection, then, taking the drilling action as an example, the fuel consumption rate of the drilling action, as well as relevant IPO parameters such as engine gear, torque, engine mode, main pump pressure, power head A / B port pressure, system flow rate, and system pressure loss, are extracted. Furthermore, the time-domain and frequency-domain characteristics of the IPO parameters, as well as fluctuation impact indicators, are extracted. Finally, through the root cause analysis algorithm, abnormal characteristic parameters are identified, and reference benchmark values ​​are provided. For example, an abnormal characteristic parameter might be that the engine gear is at 7th gear, while the reference benchmark value is 5th gear; or an abnormal characteristic parameter might be that the system pressure difference is large, while the reference benchmark value is 20 bar less, etc.

[0144] (V) Repair suggestion push

[0145] After identifying the abnormal characteristic parameters at the root cause, current equipment parameters and historical fault data can be analyzed using rule-based algorithms, machine learning algorithms, or deep learning algorithms. Expert experience can also be incorporated into the analysis to ultimately provide corresponding repair recommendations. Figure 8 As shown.

[0146] In summary, the early warning processing method for mechanical equipment provided in this embodiment has the following advantages compared with existing methods: (1) By combining virtual body modeling of machine groups with statistical hypothesis testing, engineering machinery equipment can be compared under the same construction object, and can more quickly and effectively identify the real early warning, thereby improving the accuracy of early warning; (2) A complete set of early warning identification and root cause localization methods is provided, which can comprehensively identify each sub-action of the construction process and conduct fault source analysis, thereby quickly improving service efficiency and quality; (3) Through maintenance suggestion recommendations, expert knowledge is accumulated, reducing the dependence on individual technical skills of service personnel and reducing maintenance service costs.

[0147] Based on the same inventive concept as the foregoing embodiments, the foregoing embodiments will be described in detail below through a specific example. In this embodiment, the mechanical equipment is a rotary drilling rig, the indicator to be warned is the fuel consumption rate, and the rotary drilling rig only operates during the working hours of one day (i.e., from 8:00 am to 6:00 pm).

[0148] Assuming that the time interval is measured in hours, the construction object A corresponding to a rotary drilling rig in the t-th unit time interval is recorded. t Construction Element B t And the actual value of fuel consumption rate C t Meanwhile, the total fuel consumption rate C of the rotary drilling rig was also recorded for one day. 总 .

[0149] First, conduct abnormal early warning detection on the rotary drilling rig.

[0150] Based on the daily fuel consumption rate distribution of rotary drilling rigs of the same type, the m±3σ principle is used for early warning of daily fuel consumption rate. If the total fuel consumption rate C of the rotary drilling rig in a day... 总 If the fuel consumption rate is greater than m+3σ or less than m-3σ, the rotary drilling rig is considered to have an abnormality, and false alarm filtering is then implemented. If the total fuel consumption rate C of the rotary drilling rig in a day... 总 If the fuel consumption rate is less than m+3σ and greater than m-3σ, then the fuel consumption rate of the rotary drilling rig is considered normal.

[0151] Next, false alarm filtering is performed on the rotary drilling rig.

[0152] Specifically, the construction elements B corresponding to the rotary drilling rig in the t-th unit time period are sequentially assigned. t Input construction object A t The corresponding virtual model of the drilling rig is used to obtain the baseline value C of the fuel consumption rate of the rotary drilling rig in the t-th unit time period predicted by the virtual model of the drilling rig. t ', and calculate the actual value C of the fuel consumption rate corresponding to the rotary drilling rig in the t-th unit time period. t Deviation D between t Where D1 represents the deviation between the actual value and the benchmark value of the fuel consumption rate of the rotary drilling rig in the first unit time period (i.e., 8:00-9:00 AM), D2 represents the deviation between the actual value and the benchmark value of the fuel consumption rate of the rotary drilling rig in the second unit time period (i.e., 9:00-10:00 AM), and so on, thus obtaining the deviation vector D composed of the deviations between the actual value and the benchmark value of the fuel consumption rate of the rotary drilling rig in the above 10 unit time periods, that is, D = (D1, D2, D3, D4, D5, D6, D7, D8, D9, D...). 10 When it is necessary to verify whether the fuel consumption rate is abnormal under the construction object corresponding to a certain unit time period, the test statistic corresponding to that unit time period can be calculated based on the deviation of other time periods before that unit time period. For example, if it is necessary to verify whether the fuel consumption rate is abnormal under the construction object corresponding to the 8th unit time period (i.e., 3:00 pm to 4:00 pm), the deviation vector D' composed of the deviation between the actual value of the fuel consumption rate and the benchmark value corresponding to 8:00 am to 4:00 pm is obtained, i.e., D' = (D1, D2, D3, D4, D5, D6, D7, D8). Based on the calculation formula of the test statistic and the abnormal alarm threshold in the aforementioned embodiment, the test statistic and the corresponding abnormal alarm threshold corresponding to that unit time period are obtained. Then, it is determined whether the test statistic corresponding to that unit time period is greater than the corresponding abnormal alarm threshold. If so, it is determined that the fuel consumption rate is abnormal under the construction object corresponding to that unit time period; otherwise, it is determined to be a false alarm.

[0153] Next, we conducted motion anomaly analysis and root cause anomaly localization of the IPO parameters of the rotary drilling rig to determine the abnormal characteristic parameters.

[0154] Specifically, if the fuel consumption rate is determined to be abnormal under the construction object corresponding to a certain unit time period, then the IPO parameters corresponding to each sub-action i completed by the rotary drilling rig within that unit time period are sequentially input into the sub-action regression model corresponding to sub-action i, to obtain the baseline value C of the fuel consumption rate of the rotary drilling rig when completing sub-action i, as predicted by the sub-action regression model. i ', and calculate the actual value C of the fuel consumption rate corresponding to the rotary drilling rig completing the sub-action i. i Deviation D between i In this context, D'1' represents the deviation between the actual value and the benchmark value of the fuel consumption rate of the rotary drilling rig when completing the first sub-action, D'2 represents the deviation between the actual value and the benchmark value of the fuel consumption rate of the rotary drilling rig when completing the second sub-action, and so on. Based on the deviation between the actual value and the benchmark value of the fuel consumption rate of the rotary drilling rig when completing each sub-action, and using the calculation formula for the anomaly probability in the aforementioned embodiment, the anomaly probability corresponding to each sub-action is obtained, and the target sub-action with an anomaly probability greater than or equal to a preset probability threshold is determined. Next, feature extraction is performed on the IPO parameters corresponding to the target sub-action to obtain the IPO feature parameters corresponding to the target sub-action, and the IPO feature parameters corresponding to the target sub-action are identified using a root cause localization algorithm to determine and output the anomaly feature parameters.

[0155] Finally, based on the determined abnormal characteristic parameters, corresponding maintenance suggestions are identified and output to facilitate the repair of abnormal faults.

[0156] Based on the same inventive concept as the foregoing embodiments, the present invention provides a warning processing device for mechanical equipment, including a memory and a processor. The memory stores a computer program, and when the computer program is executed by the processor, it implements the steps of the warning processing method for mechanical equipment in any of the above embodiments.

[0157] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the early warning processing method for mechanical equipment in any of the above embodiments.

[0158] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0159] In this document, the terms “comprising,” “including,” or any other variations thereof are intended to cover non-exclusive inclusion, which includes not only the elements listed but also other elements not expressly listed.

[0160] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method of early warning processing of a mechanical device, characterized by, The method includes the following steps: S101. Obtain the target work data of the target mechanical equipment within a preset time period. The target work data includes the first actual value of the construction object, construction element and warning indicator corresponding to each unit time period of the target mechanical equipment within the preset time period. S102. Based on the construction objects and construction elements corresponding to the target mechanical equipment in each unit time period, predictive processing is performed through the trained virtual body models of the machine group corresponding to different construction objects to obtain the first benchmark value of the warning indicator for the target mechanical equipment during the operation of the construction objects corresponding to the target mechanical equipment in each unit time period. S103. Based on the deviation between the first benchmark value and the first actual value of the warning indicator for the target mechanical equipment during the operation of the construction object corresponding to each unit time period, obtain the inspection statistics for the target mechanical equipment during the operation of the construction object corresponding to each unit time period. S104. When the target test statistic is greater than or equal to the preset corresponding abnormal alarm threshold, execute the abnormal alarm prompt for the construction object under the target unit time period, where the target unit time period is the unit time period corresponding to the target test statistic.

2. The method of claim 1, wherein, The target work data also includes the second actual value of the warning indicators for the target machinery and equipment within a preset time period; The method further includes: The detection method checks whether the second actual value of the warning indicator of the target mechanical equipment within a preset time period is less than or equal to the preset first abnormal warning threshold, or greater than or equal to the preset second abnormal warning threshold. The first abnormal warning threshold and the second abnormal warning threshold are determined based on the distribution of the warning indicators of the same type of mechanical equipment group within a historical preset time period. If so, proceed to step S102.

3. The method of claim 2, wherein, The method further includes: Obtain the historical actual values ​​of the warning indicators for each piece of machinery in a group of the same type of machinery within a preset historical time period; The distribution of the warning indicators of each piece of machinery and equipment is determined by fitting the historical actual values ​​of the warning indicators of the machinery and equipment group within the historical preset time period. Based on the distribution of indicators to be warned in the mechanical equipment cluster within a historical preset time period, determine the first abnormal warning threshold and / or the second abnormal warning threshold.

4. The method according to any one of claims 1 to 3, characterized in that, Before step S102, the method further includes: Obtain historical working data of each piece of machinery in a group of the same type of machinery within a preset historical time period. The historical working data includes the third actual value of the construction object, construction element and warning indicator corresponding to each unit time period of each piece of machinery within the preset historical time period. Using the indicators to be warned as the dependent variable and the construction elements as the independent variables, virtual models of machine groups corresponding to different construction objects are constructed and trained based on historical work data.

5. The method according to any one of claims 1 to 3, characterized in that, The target work data also includes the IPO parameters and the fourth actual values ​​of the warning indicators corresponding to the target mechanical equipment completing at least one action within the target unit time period. The IPO parameters are used to describe the function and construction process information of the target mechanical equipment in completing the corresponding action. The method further includes: Based on the IPO parameters corresponding to the completion of at least one action by the target mechanical equipment within the target unit time period, the second benchmark value of the warning indicator corresponding to the completion of at least one action by the target mechanical equipment within the target unit time period is obtained by predicting through the sub-action regression model corresponding to different sub-actions that has been trained. Based on the deviation between the fourth actual value and the second benchmark value of the warning indicator corresponding to each of the target mechanical equipment completing at least one action within the target unit time period, the abnormal probability corresponding to each of the at least one action is obtained. Identify target sub-actions with an abnormal probability greater than or equal to a preset probability threshold, and issue warnings for the target sub-actions.

6. The method of claim 5, wherein, The method further includes: Obtain historical working data of each piece of machinery in a group of the same type of machinery within a preset historical time period. The historical working data includes the IPO parameters and the fifth actual value of the warning indicator corresponding to each piece of machinery completing at least one action. Using the indicator to be warned as the dependent variable and the IPO parameter as the independent variable, regression models for different sub-actions are constructed and trained based on historical working data.

7. The method of claim 5, wherein, The method further includes: Feature extraction is performed on the IPO parameters corresponding to the target sub-actions to obtain the IPO feature parameters corresponding to the target sub-actions; The root cause localization algorithm is used to identify the IPO feature parameters corresponding to the target sub-actions, and to determine and output the abnormal feature parameters.

8. The method of claim 7, wherein, The method further includes: Based on abnormal feature parameters, combined with the equipment data and historical fault data of the target mechanical equipment, fault diagnosis is performed to determine and output corresponding maintenance suggestions.

9. An early warning processing device for a mechanical plant, characterized in that include: The processor and the memory storing a computer program implement the steps of the early warning processing method for the mechanical equipment according to any one of claims 1 to 8 when the processor runs the computer program.

10. A computer-readable storage medium, characterized in that, The device contains a computer program that, when executed by a processor, implements the steps of the early warning processing method for mechanical equipment as described in any one of claims 1 to 8.

Images