Machine learning device and machine learning method for learning a correlation between shipping control information and operational alarm information for an object

Abstract

Machine learning device (2) which learns a correlation between shipping control information (X1n) obtained by checking an object during its shipping and operational alarm information (X2n) issued during the operation of the object, and which includes: - a condition monitoring unit (21) that monitors the shipping control information (X1n) and the operational alarm information (X2n) as data input from an environment (1), and - a learning unit (22) which generates a learning model based on the shipping control information (X1n) and operational alarm information (X2n) monitored by the condition monitoring unit (21), wherein the object includes a motor the environment (1) comprises an engine control device (11) that controls the engine during its shipment and outputs the shipping control information (X1n), and an engine control device (12) that gives an alarm during the operation of the engine and outputs the operating alarm information (X2n), The shipping control information (X1n) includes control element results associated with a motor model and motor inspection date, an insulation resistance value, an earth resistance value, a current value and / or a switching impulse voltage for the motor, and The operating alarm information (X2n) includes an overcurrent alarm, a noise alarm and / or an overload alarm for the motor.

Description

Background of the invention 1. Field of the invention

[0001] The present invention relates to a machine learning device and a machine learning method for learning the correlation between shipping control information and operational alarm information for an object. 2. Description of the relevant state of the art

[0002] Traditionally, electric motors (motors: objects) are used in a variety of electrical machines and devices, including, for example, machine tools and robots, which are controlled by computer numerical control (CNC) devices. Such motors are typically subjected to shipping inspection when products, such as machine tools or robots, are dispatched to end users. This shipping inspection is performed not only when shipping products that use motors, but also, for example, when shipping the motors themselves, both of which are then delivered to end users. In this description, the information obtained by inspecting a motor or a product using the motor during dispatch is referred to below as shipping inspection information.

[0003] After delivery to the users, the products (motors), such as machine tools or robots, are used (operated) in actual operating environments. However, alarms may occur during operation of the machine tools or similar equipment. These alarms are triggered by various factors, including electrical problems (e.g., motor insulation or servo amplifier failure) and mechanical (structural) problems (e.g., bearing failure or metal fatigue). In this description, the information relating to an alarm triggered during the operation of a motor or a product using the motor is referred to below as operating alarm information.

[0004] In this description, the target for learning the correlation between shipping control information and operational alarm information through machine learning is predominantly a motor (a product using the motor) as an example of an object, although the object is not limited to a motor and various objects can be used whose shipping control information and operational alarm information can be captured, which include, for example, a servo amplifier that performs a servo drive of a motor.

[0005] The correlation described above between shipping control information and operational alarm information for an object (motor) has traditionally received little attention, and even when the shipping control information and operational alarm information for the motor do exhibit a correlation, individuals may have to expend considerable effort to analyze (organize) this correlation. Furthermore, because the operating conditions (e.g., ambient temperature or humidity, runtime, and load size during operation) of each individual motor (including motors used in a variety of electrical machines and devices) vary significantly for each actual application, it is difficult for individuals to understand the actual operating conditions and clarify the correlation between shipping control information and operational alarm information for that motor.

[0006] Furthermore, the published Japanese patent JP H07 - 174 616 A, for example, discloses a control device that checks the object, such as a compact motor, for anomalies based on vibrations of an object or a noise emitted by the object.The control device uses a sensor to detect vibrations of the object or a noise emitted by the object, performs envelope rectification and Fourier transformation, calculates a certainty factor for the presence of an anomaly of the object using a processing unit, further calculates a certainty factor for the presence of an anomaly of the object by inputting a feature vector signal, representing the features of Fourier transform signals, into a neural network, performs a fuzzy calculation based on the certainty factors calculated by both the processing unit and the neural network, and determines the presence or absence of an anomaly of the object.

[0007] Furthermore, for example, the Japanese national publication of international patent application JP 2007-528985A discloses a technique for optically measuring a structure formed on a semiconductor wafer using a machine learning system. In other words, in a method for inspecting a structure formed on a semiconductor wafer, a first diffraction signal measured using a measuring instrument is obtained, and at least one parameter characterizing the structure's profile is received as input to obtain a second diffraction signal generated using the machine learning system.The first diffraction signal and the second diffraction signal are compared and, if they match within the range of matching criteria, the shape of the structure is obtained based on the profile or at least one parameter used by the machine learning system to generate the second diffraction signal.

[0008] Furthermore, for example, the published Japanese patent publication JP 2004 - 354 250 A discloses a defect detection device capable of processing an image obtained by capturing an object in order to classify defects efficiently and accurately, wherein a defect detection device using a neural network provided according to each specific defect to be classified that occurs in the manufacturing process of an object, enables each neural network to learn for each defect to be classified and determines the presence or absence of a defect in the object for each defect to be classified by each learned neural network.The neural network includes a neural network for image defocus errors to classify defocus errors that occur image by image in the exposure process of the object, whereby the image size is implemented as preprocessing input into the neural network according to the input layer size.

[0009] Further state of the art is known from DE 199 14 865 A1 and WO 2005 / 003 911 A2.

[0010] Some patent specifications of the associated prior art, as described above, conventionally disclose techniques for calculating a certainty factor for the presence of an anomaly in a motor using a neural network, or techniques for testing a semiconductor wafer or the like based on a machine learning model.

[0011] However, no special attention has been paid to the correlation between shipping control information and operational alarm information for an engine, and although the shipping control information and the operational alarm information for the engine do show a certain correlation, as described above, it is difficult for people to analyze this correlation because they may have to expend an enormous amount of effort.

[0012] Furthermore, since the operating conditions of each individual engine vary considerably for each actual operating environment, it can be very difficult for individuals to understand the actual operating conditions and to clarify the correlation between shipping control information and operational alarm information for that engine, where the correlation between shipping control information and operational alarm information for an engine, taking such actual operating conditions into account, is assessed solely based on the experience or intuition of qualified engineers or users who have used the engine in the operating environment for an extended period of time.

[0013] In view of the conventional problems described above, one objective of the present invention is to obtain a correlation between shipping control information and operational alarm information for an object. Obtaining such a correlation leads to solutions for various problems, such as improvements to shipping control elements, structural enhancement, lifetime determination, and improved object quality. Summary of the invention

[0014] According to a first aspect of the present invention, a machine learning device is provided which learns a correlation between shipping control information obtained by checking an object during its shipment and operational alarm information issued during the operation of the object, and which comprises a state monitoring unit which monitors the shipping control information and the operational alarm information, and a learning unit which generates a learning model based on the shipping control information and operational alarm information monitored by the state monitoring unit.

[0015] The learning unit can generate a distribution correlation between the shipping control information and the operational alarm information as a learning model. The machine learning device can further include an output application unit that, based on the learning model generated by the learning unit, outputs a control element upon shipment of the object, which influences an alarm triggered during the object's operation. The learning unit can generate the learning model by clustering the shipping control information and the operational alarm information.The learning unit can generate the learning model through hierarchical clustering, in which the shipping control information and the operational alarm information are computationally processed in a hierarchical structure, and non-hierarchical clustering, in which the shipping control information and the operational alarm information are computationally processed based on a distance between nodes until a predetermined number is reached.

[0016] The machine learning device can further comprise a neural network. The machine learning device can be connected to at least one other machine learning device and exchange or share the learning model generated by the learning unit of the machine learning device with the at least one other machine learning device. The machine learning device can reside in a first cloud server, and the other machine learning device can reside in a second cloud server that differs from the first cloud server. The object can comprise a motor, and the shipping control information can comprise control results from control elements that are associated with a model of the motor and a control data point of the motor.The object can include a servo amplifier that performs a servo drive of a motor, and the shipping control information can include control results from control elements associated with a model of the servo amplifier and a control date of the servo amplifier. The shipping control information can include an insulation resistance value, an earth resistance value, a current value, and / or a switching impulse voltage for the object, and the operating alarm information can include an overcurrent alarm, a noise alarm, and / or an overload alarm for the object.

[0017] According to a second aspect of the present invention, a machine learning method is provided for learning a correlation between shipping control information obtained by controlling an object during its shipment and operational alarm information issued during the operation of the object, comprising monitoring the shipping control information and the operational alarm information and generating a learning model based on the monitored shipping control information and operational alarm information.

[0018] Generating the learning model can involve creating a distribution correlation between the shipping control information and the operational alarm information as the learning model. The machine learning process can further include outputting a control element upon shipment of the object, which influences an alarm triggered during the object's operation, based on the generated learning model. Generating the learning model can involve creating it by clustering the shipping control information and the operational alarm information. Brief description of the drawings

[0019] The present invention is more clearly understood with reference to the following associated drawings. Fig. 1 is a block diagram that schematically represents an embodiment of a machine learning device according to the present invention; Fig. 2 is a graphical representation to illustrate an example of computational processing, which refers to the in Fig. 1 machine learning device shown is used; and Fig. 3 is a representation to illustrate another exemplary computing process, which refers to the one in Fig. The machine learning device shown in section 1 is used. Detailed description

[0020] The following describes in detail embodiments of a machine learning device and a machine learning method according to the present invention with reference to the accompanying drawings. The following uses a motor as an example object for learning the correlation between shipping control information and operational alarm information by machine learning. However, the object is not limited to a motor, and various objects can be used whose shipping control information and operational alarm information can be acquired. These objects may include, for example, a servo amplifier that drives a servo motor.

[0021] Fig. Figure 1 is a block diagram schematically illustrating an embodiment of a machine learning device according to the present invention. A machine learning device 2 according to this embodiment comprises, as shown in Figure 1, a block diagram schematically representing an embodiment of a machine learning device according to the present invention. Fig. Figure 1 shows a condition monitoring unit 21, a learning unit 22, and an output application unit 23. The condition monitoring unit 21, for example, monitors shipping control information (X1n) obtained by a check during the shipping of a motor (object) and operating alarm information (X2n) output during the operation of the motor as data input from an environment 1.

[0022] Environment 1 includes, for example, a motor control unit 11, which monitors a motor during its shipment and outputs shipment control information (X1n), and a motor control unit 12, which raises an alarm during motor operation and outputs operational alarm information (X2n). The shipment control information (X1n) and operational alarm information (X2n) monitored by the condition monitoring unit 21 are fed into the learning unit 22, which generates a distribution correlation between the shipment control information (X1n) and the operational alarm information (X2n) as a learning model.

[0023] The output application unit 23, based on the learning model generated by the learning unit 22, outputs a control element, for example, when a motor is shipped, which influences an alarm triggered during the motor's operation. The output application unit 23 can be located outside the machine learning device 2, and its output, based on the learning model generated by the learning unit 22, is not limited to the aforementioned control element when a motor is shipped, which influences an alarm triggered during the motor's operation.

[0024] Learning Unit 22 applies "unsupervised learning" to the shipping control information (X1n) and the operational alarm information (X2n) to generate a learning model, for example, through clustering. In other words, Learning Unit 22 generates a learning model through hierarchical clustering, where the shipping control information (X1n) and the operational alarm information (X2n) are computationally processed in a hierarchical structure, or non-hierarchical clustering, where the shipping control information (X1n) and the operational alarm information (X2n) are computationally processed based on the distance between nodes until a predefined number is reached.

[0025] Fig. 2 is a graphical representation to illustrate an example of computational processing, which refers to the in Fig. 1 machine learning device is used, i.e. to illustrate non-hierarchical clustering (clustering based on the K-means method). Fig. 3 is a representation to illustrate another exemplary computing process, which refers to the one in Fig. Figure 1 shows a machine learning device applied to an autoencoder, and demonstrates an exemplary neural network applied to an autoencoder for dimensional compression in hierarchical clustering.

[0026] The machine learning device (machine learning method) according to this embodiment uses "unsupervised learning" to perform clustering and therefore extracts, for example, a useful rule, a knowledge representation, and a determination criterion from input data of the shipping control information (X1n) and operational alarm information (X2n) through analysis, outputs the determination results, and acquires knowledge (machine learning). Machine learning device 2 uses, for example, a neural network; however, in the actual implementation of machine learning device 2, a general-purpose computer or processor can be used, whereby the use of general-purpose computing on graphics processing units (GPGPU) or large PC clusters enables higher processing speeds.

[0027] In unsupervised learning, large amounts of input data are fed exclusively into the machine learning device 2 to learn the distribution of the input data. Unlike supervised learning, the input data can be compressed, classified, and shaped without corresponding teacher output data. This enables, for example, the clustering of features in input datasets with similar features. Based on the resulting data, an arbitrary standard can be defined, and outputs can be assigned to this standard for optimization purposes, thereby predicting a future output. In this description, "unsupervised learning" refers to a broad definition that includes, for example, a hybrid approach between supervised and unsupervised learning, which is referred to as "semi-supervised learning."

[0028] More precisely, the machine learning device 2 (learning unit 22), as described in Fig. Figure 2 shows that the dispatch control information (X1n) and operational alarm information (X2n) are received and, based on the distance between nodes (the distance between points in Fig. 2) For example, perform clustering until a predetermined number (e.g., k = 3) is reached to obtain clusters A, B, and C. In other words, clusters A, B, and C can be obtained based on the correlation between shipping control information (X1n) and operational alarm information (X2n) for a motor (object).

[0029] As in Fig. As shown in Figure 3, a neural network 3 can be used as an autoencoder for dimensional compression if, for example, input shipping control information (X1n) and operational alarm information (X2n) for a motor are hierarchically arranged and grouped into clusters. Shipping control information (X1n) and operational alarm information (X2n) for a motor are used, for example, as input X (input data Xn) of the neural network (autoencoder) 3, and output data Yn, hierarchically grouped into clusters, are output as output Y. The process described in Figure 3 demonstrates this. Fig. 2. Non-hierarchical clustering shown and the one in Fig. The hierarchical clustering shown in the three examples is merely an illustration, and of course various known methods for “unsupervised learning” can be applied to the machine learning device 2 according to this embodiment.

[0030] The machine learning device 2 according to this embodiment can, for example, be located in a manufacturer's server, as described in more detail below. However, the machine learning device 2 is connectable to at least one other machine learning device and can exchange or share a learning model generated by the learning unit 22 of each machine learning device 2 with the at least one other machine learning device. The servers equipped with these machine learning devices 2 can, for example, be run as other cloud servers that can be accessed via a communication link, such as the internet.

[0031] In this way, the machine learning device (machine learning method) according to this embodiment can perform a correlation (e.g., between clusters A, B, and C in Fig.2) Correlations can be obtained between shipping control information and operational alarm information for an object (motor). Examples of the correlation obtained can include various correlations, such as that between the surge voltage in the shipping control information and a noise alarm (an alarm due to a communication error) in the operational alarm information, as well as that between the ground resistance in the shipping control information and a communication error alarm in the operational alarm information. Obtaining a correlation between shipping control information and operational alarm information for an object can lead to solutions for various problems, such as improvements to shipping control elements, structural enhancements, service life determination, and improvement of the object's quality.

[0032] An application example of the machine learning device (machine learning method) according to this embodiment is described below. First, the motor control device 11 performs a shipping check of a motor during its shipment to obtain information (X1n) relating to the check performed during shipping. Examples of the shipping check information (X1n) for each motor may include the type (model), the inspection date, the insulation resistance value, the winding resistance value, the back EMF value, the current value, and the shaft friction torque value during operation of that motor, wherein the shipping check information (X1n) is recorded in a storage device (e.g., a non-volatile storage device such as a hard disk drive or flash memory).

[0033] After the motors are shipped, the operating alarm information (X2n) is recorded during the operation of each motor. Examples of the operating alarm information (X2n) for each motor may include the details of an alarm, the time elapsed until the alarm occurred, and the speed, torque, current, and temperature at the time the alarm occurred. This operating alarm information (X2n) is recorded in a storage device. Examples of alarm types may include an overcurrent alarm, a noise alarm (an alarm given when noise is present at or above a predetermined level, or an alarm given when a communication error occurs), an overload alarm (an alarm given when the motor overheats), and an excessive movement error.

[0034] The shipping control information (X1n) of each individual motor is stored, for example, on a server (server storage device) belonging to the manufacturer of the motor or of a manufacturer of a product using the motor. The operational alarm information (X2n) of each individual motor is temporarily stored, for example, in a storage device of the control unit 12, which controls the motor, and copied and stored from the control unit 12 to the manufacturer's server by the manufacturer's service technician. Alternatively, the operational alarm information (X2n) of each individual motor can be configured, for example, to be uploaded directly from the motor control unit 12 to the manufacturer's server via a communication link, such as the internet.Therefore, the machine learning device 2, located for example on the manufacturer's server, can perform the learning (processing) using the shipping control information (X1n) and operational alarm information (X2n) of each individual motor as input.

[0035] The processing by the machine learning device, located, for example, on the manufacturer's server, according to this embodiment, is described below. First, a correlation between (X1n) and (X2n) is generated as a learning model based on the shipping control information (X1n) and operating alarm information (X2n) for a motor. Shipping control elements (a1, a2, a3, a4, ...) belonging to the shipping control information (X1n) for the motor, which influence the operating alarm information (X2n), are then output from the learning model.

[0036] For faults that trigger specific alarms (e.g., a motor's ground fault resulting in an overcurrent alarm, and motor demagnetization resulting in an overheat alarm), the fault occurrence rate can be predicted based on information (shipping control information) related to the shipping control elements of motors that are still operating without problems. This prediction can be made using the shipping control information (X1n) for the motor that triggers an alarm. In other words, the fault occurrence rate of a specific motor during operation can be predicted based on the correlation between the shipping control information (X1n) and the operating alarm information (X2n) for that motor.The effects achieved based on the correlation between shipping control information and operational alarm information for a motor (object) are not limited to such a prediction of the occurrence rate of motor failures, and various effects can of course be achieved.

[0037] Regarding, for example, the demagnetization of a motor accompanied by an overheating alarm, the cause can be assessed and immediate action taken as follows: if the counter-electromotive force in the shipping control information (shipping control) is affected, the choice of magnet with regard to its design or magnetization can be considered responsible; if the winding resistance in the shipping control information is affected, the winding during manufacturing can be considered responsible; or if the shaft friction resistance value in the shipping control information is affected, an increase in friction resistance due to damage inflicted on the bearing during manufacturing can be considered responsible.

[0038] If the occurrence rate of a specific alarm increases after a particular shipping date, a component replacement point, a manufacturing facility, or similar factors could be considered responsible. Similarly, if the alarm occurrence rate increases for a specific model, factors specific to that model's design could be identified as responsible. In any case, confirming the correlation between shipping control information and operational alarm information allows for predicting the failure rate for each individual motor operating within the area of ​​operation and even enables effective targeting of a failure prevention measure when implementing such a measure.

[0039] Although a motor was used as the example object in the preceding description, a servo amplifier that performs a servo drive of a motor is given below as an example object. In this case, the shipping control information (X1n) of the servo amplifier includes the model of the servo amplifier and the control results of control elements associated with the control date of the servo amplifier. More specifically, the shipping control information (X1n) of the servo amplifier includes, for example, the insulation resistance value, the earth resistance value, the current value, and the switching impulse voltage for the servo amplifier, and the operating alarm information (X2n) of the servo amplifier includes, for example, an overcurrent alarm, a noise alarm, and an overload alarm (overheating of the amplifier) ​​for the servo amplifier.Examples of correlations obtained by the machine learning device (machine learning method) according to this embodiment can include various correlations, such as that between the surge voltage in the shipping control information and a communication error alarm (noise alarm) in the operational alarm information, as well as that between the ground resistance in the shipping control information and a communication error alarm in the operational alarm information, as in the case where a motor is assumed to be the object. Obtaining a correlation between shipping control information and operational alarm information for an object leads, as described above, to solutions for various problems, such as improvements to shipping control elements, structural enhancement, lifetime determination, and improvement of the object's quality.

[0040] The machine learning device and the machine learning method according to the present invention have the effect of obtaining a correlation between shipping control information and operational alarm information for an object.

Claims

[1] Machine learning device (2) which learns a correlation between shipping control information (X1n) obtained by checking an object during its shipping and operational alarm information (X2n) issued during the operation of the object, and which includes: - a condition monitoring unit (21) that monitors the shipping control information (X1n) and the operational alarm information (X2n) as data input from an environment (1), and - a learning unit (22) which generates a learning model based on the shipping control information (X1n) and operational alarm information (X2n) monitored by the condition monitoring unit (21), wherein the object includes a motor the environment (1) comprises an engine control device (11) that controls the engine during its shipment and outputs the shipping control information (X1n), and an engine control device (12) that gives an alarm during the operation of the engine and outputs the operating alarm information (X2n), The shipping control information (X1n) includes control element results associated with a motor model and motor inspection date, an insulation resistance value, an earth resistance value, a current value and / or a switching impulse voltage for the motor, and The operating alarm information (X2n) includes an overcurrent alarm, a noise alarm and / or an overload alarm for the motor. [2] Machine learning device (2) which learns a correlation between shipping control information (X1n) obtained by checking an object during its shipping and operational alarm information (X2n) issued during the operation of the object, and which includes: - a condition monitoring unit (21) that monitors the shipping control information (X1n) and the operational alarm information (X2n) as data input from an environment (1), and - a learning unit (22) which generates a learning model based on the shipping control information (X1n) and operational alarm information (X2n) monitored by the condition monitoring unit (21), wherein the object includes a servo amplifier that performs a servo drive of a motor, the environment (1) comprises an engine control device (11) that controls the engine during its shipment and outputs the shipping control information (X1n), and an engine control device (12) that gives an alarm during the operation of the engine and outputs the operating alarm information (X2n), The shipping control information (X1n) includes control results from control elements associated with a model of the servo amplifier and a control date of the servo amplifier, an insulation resistance value, an earth resistance value, a current value and / or a switching impulse voltage for the servo amplifier, and The operating alarm information (X2n) includes an overcurrent alarm, a noise alarm and / or an overload alarm for the servo amplifier. [3] Machine learning device according to claim 1 or 2, wherein the learning unit (22) generates a distribution correlation between the shipping control information (X1n) and the operational alarm information (X2n) as a learning model. [4] Machine learning device according to one of claims 1 to 3, further comprising: - an output application unit (23) which, based on the learning model generated by the learning unit (22), outputs a control element when the object is dispatched, which affects an alarm given during the operation of the object. [5] Machine learning device according to one of claims 1 to 4, wherein the learning unit (22) generates the learning model by clustering the shipping control information (X1n) and the operational alarm information (X2n). [6] Machine learning device according to claim 5, wherein the learning unit (22) generates the learning model by hierarchical clustering, in which the shipping control information (X1n) and the operational alarm information (X2n) are computationally processed in a hierarchical structure, and non-hierarchical clustering, in which the shipping control information (X1n) and the operational alarm information (X12) are computationally processed based on a distance between nodes until a predetermined number is reached. [7] Machine learning device according to any one of claims 1 to 6, wherein the machine learning device further comprises a neural network. [8] Machine learning device according to any one of claims 1 to 7, wherein the machine learning device is connectable to at least one other machine learning device and exchanges or shares the learning model generated by the learning unit (22) of the machine learning device with the at least one other machine learning device. [9] Machine learning device according to claim 8, wherein - the machine learning device is located in a first cloud server and - the other machine learning device is located in a second cloud server that differs from the first cloud server. [10] Machine learning method for learning a correlation between shipping control information (X1n) obtained by checking an object during its shipment and operational alarm information (X2n) issued during the operation of the object, comprising: - Monitoring the shipping control information (X1n) and the operational alarm information (X2n) as data input from an environment (1) and - Generating a learning model based on the monitored shipping control information (X1n) and operational alarm information (X2n), wherein the object includes a motor the environment (1) comprises an engine control device (11) that controls the engine during its shipment and outputs the shipping control information (X1n), and an engine control device (12) that gives an alarm during the operation of the engine and outputs the operating alarm information (X2n), The shipping control information (X1n) includes control element results associated with a motor model and motor inspection date, an insulation resistance value, an earth resistance value, a current value and / or a switching impulse voltage for the motor, and The operating alarm information (X2n) includes an overcurrent alarm, a noise alarm and / or an overload alarm for the motor. [11] Machine learning method for learning a correlation between shipping control information (X1n) obtained by checking an object during its shipment and operational alarm information (X2n) issued during the operation of the object, comprising: - Monitoring the shipping control information (X1n) and the operational alarm information (X2n) as data input from an environment (1) and - Generating a learning model based on the monitored shipping control information (X1n) and operational alarm information (X2n), wherein the object includes a servo amplifier that performs a servo drive of a motor, the environment (1) comprises an engine control device (11) that controls the engine during its shipment and outputs the shipping control information (X1n), and an engine control device (12) that gives an alarm during the operation of the engine and outputs the operating alarm information (X2n), The shipping control information (X1n) includes control results from control elements associated with a model of the servo amplifier and a control date of the servo amplifier, an insulation resistance value, an earth resistance value, a current value and / or a switching impulse voltage for the servo amplifier, and The operating alarm information (X2n) includes an overcurrent alarm, a noise alarm and / or an overload alarm for the servo amplifier. [12] Machine learning method according to claim 10 or 11, wherein generating the learning model comprises generating a distribution correlation between the shipping control information (X1n) and the operational alarm information (X2n) as a learning model. [13] Machine learning method according to any one of claims 10 to 12, further comprising: - Outputting a control element upon dispatch of the object, which influences an alarm given during the operation of the object, based on the generated learning model. [14] Machine learning method according to one of claims 10 to 13, wherein the generation of the learning model comprises generation by clustering the shipping control information (X1n) and the operational alarm information (X2n).

Images