Frequency converter and drive system

Abstract

The invention relates to a frequency converter (1) for actuating an electric motor (2), comprising: - sensor elements (3 to 8), which are designed to generate sensor data (x1 to xn), and - a control unit (9), which is designed to control the operation of the frequency converter (1) depending on the sensor data (x1 to xn), - wherein the control unit (9) is designed to calculate at least one feature (y1 to ym) from the sensor data (x1 to xn).

Description

[0001] Frequency converter and drive system

[0002] The invention is based on the object of providing a frequency converter and a drive system that enable a statement about their condition as reliably as possible.

[0003] The frequency converter is used to control an electric motor. The frequency converter has conventional sensor elements or sensors that are designed to generate sensor data. The frequency converter further has a control unit, for example in the form of a microprocessor-based controller, that is designed to control the operation of the frequency converter depending on the sensor data. The control unit is designed to calculate at least one feature from the sensor data. A feature is in particular a scalar value that is obtained from the sensor data and characterizes it. For example, a feature is a maximum value, a minimum value, a temporal average, an energy within certain frequency bands, etc. of the sensor data.

[0004] In one embodiment, the at least one feature forms an input variable of an artificial intelligence (AI) unit separate from the frequency converter.

[0005] In one embodiment, the frequency converter has at least one, in particular all, of the following sensor elements: a current sensor for detecting a drive current output by the frequency converter, a speed sensor for detecting a motor speed of an electric motor controlled by the frequency converter, a position sensor for detecting a position of a mechanical load moved by the electric motor, a sensor for detecting an operating time of the frequency converter, a voltage sensor for detecting a drive voltage generated by the frequency converter, a temperature sensor for detecting an operating temperature of the frequency converter, and a temperature sensor connection for connecting an external temperature sensor, in particular a motor temperature sensor.

[0006] In one embodiment, the control unit is configured to calculate at least one, in particular all, of the following features from the sensor data: a frequency spectrum of the drive current, and / or a frequency spectrum of the motor speed, and / or a frequency spectrum of the load position, and / or a frequency spectrum of a following error, and / or an average torque generated by the electric motor, and / or an average motor speed, and / or an average electrical or mechanical power, and / or a variance of the torque generated by the electric motor, and / or a variance of the motor speed, a number of operating hours, and / or a number of motor revolutions at a torque greater than a threshold torque, and / or a maximum speed, and / or a maximum torque generated by the electric motor, and / or a maximum electrical power, and / or a maximum operating temperature,and / or a maximum tracking error.,

[0007] In one embodiment, the control unit is configured to calculate the at least one feature from the sensor data discontinuously over time, i.e., not continuously or continuously. This makes it possible to generate small amounts of data with a high information content, since the at least one feature is only generated when characteristic and meaningful conditions exist. During other periods, no calculation is performed, so that the computational effort and data volume can be minimized.

[0008] In one embodiment, the control unit is designed to calculate the at least one feature from the sensor data only in time intervals during which the frequency converter controls the electric motor to execute a predetermined test drive profile or measurement profile. The at least one feature is therefore determined exclusively under comparable or identical conditions.

[0009] In one embodiment, the predefined test drive profile consists of a predefined sequence of accelerations and movements at a constant speed. For example, the predefined test drive profile can be composed as follows: from a standstill, the vehicle accelerates to a maximum speed, maintains the maximum speed constant for a predefined period of time, and then decelerates to a standstill.

[0010] In one embodiment, the control unit is configured to calculate the at least one feature exclusively from sensor data relating to a torque generated by the electric motor. According to the invention, it was recognized that the torque is highly characteristic of the state of the electric drive system, so it may be sufficient to evaluate only the torque.

[0011] In one embodiment, the control unit is designed to calculate the at least one feature only from sensor data for which predetermined criteria are met, which ensure the information content and comparability. Typical selection criteria can be that the speed lies within a specified range during the measurement, or that the torque permanently exceeds a minimum value. The drive system has a frequency converter described above and a Kl unit formed separately from the frequency converter, wherein the at least one feature generated by means of the frequency converter is intended to form an input variable of the Kl unit.

[0012] In one embodiment, the AI ​​unit is configured to perform predictive maintenance, i.e., to determine a state of the drive system based on the at least one feature. The state can be, for example, error-free, requiring maintenance, faulty, etc.

[0013] To determine the condition of a drive system, for example, to determine whether a component of the drive system is defective, AI or machine learning is used according to the invention. In these methods, unlike deep learning, the sensor data or their measurement series are not used directly. Instead, the input to the AI ​​unit is formed by scalar quantities calculated from the sensor data, or so-called features. These features are, for example, properties of a measurement or a single value. During the AI ​​training process, a wide variety of features are calculated and used. Part of the learning process involves selecting the best features. Simple features include, for example, the largest value, the average, the number of values ​​above 80%, etc. Linked features include, for example, the average of a product of two values, the maximum of the sum of two values, the largest value, etc.Statistical features include, for example, the width and / or shape of a distribution curve. Frequency-domain features include, for example, a frequency component at a specific frequency, a lowest frequency, the evenness of a frequency distribution, etc.

[0014] Features are particularly meaningful when they are technically motivated. Speed-correlated errors occur, for example, in the case of positional damage. A nonlinearity in the power transmission around zero indicates, for example, a loose timing belt. The combination of domain knowledge in the selection of features and K1 is optimal for machine learning.

[0015] It is usually assumed that the data required to calculate the features is available at the location where the AI ​​unit is located, typically in a so-called cloud. However, this is generally not the case with drive systems, as certain data is not available in the cloud or is not stored. A frequency converter generates a large amount of sensor data at a high sampling rate. The resulting data volumes cannot usually be transferred to a cloud in real time. Calculating a feature in the frequency domain is therefore not possible, as the associated sensor data is not available with the necessary temporal resolution. Furthermore, transferring sensor data to the cloud is very complex.

[0016] The invention further recognized that sensor data generated by targeted excitation of the frequency converter is significantly more meaningful than sensor data acquired during normal operation. An excitation caused by the frequency converter can, for example, be torque noise. This means that a better forecast probability can be achieved based on a short learning phase than during a long period of real-world operation. Furthermore, feature-based evaluation requires significantly less computing power than sensor data-based evaluation.

[0017] According to the invention, features are stored in the frequency converter or its control unit. According to the invention, features are determined that are meaningful for the state of the frequency converter or the drive system. Furthermore, suitable excitations can be generated using the frequency converter, such as torque noise, current surges, characteristic movements or superimposed movements, etc., which serve to generate meaningful sensor data, which are then used to generate characteristic features.

[0018] Some features are described below as examples.

[0019] 1. Features based on a frequency analysis of current oscillations, speed oscillations, a following error and / or a position.

[0020] For this purpose, a rotational speed is determined using an excitation containing torque noise, for example, and the determined rotational speed is transformed into the frequency domain. The spectral components in different frequency ranges are then summed.

[0021] Such features make it possible, for example, to determine changes in a natural frequency. Nonlinear changes in the natural frequency indicate, for example, looseness in a drive train, such as a loose timing belt. A correlation between a high-amplitude vibration and a speed indicates shocks caused by defective gears.

[0022] 2. Features based on an average torque, speed, or electrical power during a constant mechanical movement. For example, the corresponding static torque can be determined when a constant speed is reached. This can be used to determine friction, bearing condition, capacitor condition, etc.

[0023] 3. Features based on torque and speed variance.

[0024] A wide variance or distribution of the sensor data may indicate local friction due to jamming and / or contamination and failure of the sensor elements.

[0025] 4. Features that represent the inertia of a drive train.

[0026] Mass inertia can be calculated when the drive is accelerating and the friction and process forces are known. Mass inertia, especially its variation, is a feature that provides information about the dynamic properties of the drive system.

[0027] 5. Features that indicate a longevity.

[0028] The number of operating hours, the number of revolutions at a torque greater than a threshold, etc. provide information about the remaining service life of many components of the drive system, since aging depends on the stress on a component.

[0029] 6. Features that affect maximum values ​​of the sensor data.

[0030] Maximum values ​​of speed, torque, position, following error, power, temperature, acceleration, etc. are typically constant in a cyclic process. A change in these maximum values ​​can be inferred from a change in the state of the drive system.

[0031] The invention enables optimal preparation for machine learning and predictive maintenance in a drive system, as standardized data sets are provided in the form of the features in the frequency converter. Furthermore, the meaningfulness of the data for all types of AI is improved through optimized excitation.

[0032] According to the invention, information is provided in the frequency converter in the form of meaningful features. Due to the increased information content of the generated features or data, it is possible to make reliable statements about the condition of the drive system using simple decision trees. The invention is described in detail below with reference to the drawings. Herein:

[0033] Fig. 1 an electric drive system with a frequency converter and a Kl unit for determining the condition of the electric drive system.

[0034] Fig. 1 shows an electric drive system 100 with a frequency converter 1 and a control unit 10 for determining the state of the electric drive system 100.

[0035] The frequency converter 1 is conventionally used to control an electric motor 2. The frequency converter 1 has sensor elements 3 to 8, which are designed to generate corresponding sensor data x1 to xn, where n is a natural number greater than 1. The sensor data x1 to xn are each embodied, for example, as a time-discrete sequence of digital values.

[0036] The frequency converter 1 has a control unit 9 which is designed to control the operation of the frequency converter 1 in a conventional manner depending on the sensor data x1 to xn.

[0037] The control unit 9 is designed to calculate m features y1 to ym from the sensor data x1 to xn, where m is a natural number greater than or equal to 1.

[0038] The frequency converter 1 can, for example, have the following sensor elements: a current sensor 3 for detecting a drive current, a speed sensor 4 for detecting a motor speed, a position sensor 5 for detecting a load position, a sensor 6 for detecting an operating time of the frequency converter 1, a voltage sensor 7 for detecting a drive voltage, and a temperature sensor 8 for detecting an operating temperature of the frequency converter 1.

[0039] The features y1 to ym are intended to form input variables of the Kl unit 10, which is separate from the frequency converter 1.

[0040] The control unit 9 is designed to calculate at least the following features y1 to ym from the sensor data: a frequency spectrum of the drive current, and / or a frequency spectrum of the motor speed, and / or a frequency spectrum of the load position, and / or a frequency spectrum of a following error, and / or an average torque generated by the electric motor 2, and / or an average motor speed, and / or an average electrical or mechanical power, and / or a variance of the torque generated by the electric motor 2, and / or a variance of the motor speed, and / or a number of operating hours, and / or a number of motor revolutions at a torque greater than a threshold torque, and / or a maximum speed, and / or a maximum torque generated by the electric motor 2, and / or a maximum electrical power, and / or a maximum operating temperature,and / or a maximum tracking error.,

[0041] The Kl unit 10 is designed to determine a state in the form of state variables Z1 to Zk of the drive system 100 based on the features y1 to ym, where k is a natural number greater than or equal to 1.

Claims

Patent claims 1. Frequency converter (1) for controlling an electric motor (2), comprising: sensor elements (3 to 8) which are designed to generate sensor data (x1 to xn), and a control unit (9) which is designed to control the operation of the frequency converter (1) as a function of the sensor data (x1 to xn), characterized in that the control unit (9) is designed to calculate at least one feature (y1 to ym) from the sensor data (x1 to xn).

2. Frequency converter (1) according to claim 1, characterized in that the at least one feature (y1 to ym) is provided to form an input variable of a Kl unit (10) separate from the frequency converter.

3. Frequency converter (1) according to one of the preceding claims, characterized in that the frequency converter (1) has at least one of the following sensor elements: a current sensor (3) for detecting a drive current, a speed sensor (4) for detecting a motor speed, a position sensor (5) for detecting a load position, a sensor (6) for detecting an operating time of the frequency converter (1), a voltage sensor (7) for detecting a drive voltage, a temperature sensor (8) for detecting an operating temperature of the frequency converter (1), and a temperature sensor connection for connecting an external temperature sensor, in particular a motor temperature sensor.

4. Frequency converter (1) according to claim 3, characterized in that the control unit (9) is designed to calculate at least one of the following features (y1 to ym) from the sensor data: a frequency spectrum of the drive current, and / or a frequency spectrum of the motor speed, and / or a frequency spectrum of the load position, and / or a frequency spectrum of a lag error, an average torque generated by the electric motor (2), and / or an average motor speed, and / or an average electrical or mechanical power, a variance of the torque generated by the electric motor (2), and / or a variance of the engine speed, a number of operating hours, and / or a number of engine revolutions at a torque that is greater than a threshold torque, and a maximum speed, and / or a maximum torque generated by the electric motor (2), and / or a maximum electrical power, and / or a maximum operating temperature, and / or a maximum lag error.

5. Frequency converter (1) according to one of the preceding claims, characterized in that the control unit (9) is designed to calculate the at least one feature (y1 to ym) discontinuously from the sensor data (x1 to xn).

6. Frequency converter (1) according to one of the preceding claims, characterized in that the control unit (9) is designed to calculate the at least one feature (y1 to ym) from the sensor data (x1 to xn) only as long as the frequency converter (1) controls the electric motor (2) to execute a predetermined test drive profile.

7. Frequency converter (1) according to claim 6, characterized in that the predetermined test drive profile consists of a predetermined sequence of accelerations and movements at constant speed.

8. Frequency converter (1) according to one of the preceding claims, characterized in that the control unit (9) is designed to calculate the at least one feature (y1 to ym) exclusively from sensor data (x1 to xn) relating to a torque generated by the electric motor (2).

9. Frequency converter (1) according to one of the preceding claims, characterized in that the control unit (9) is designed to calculate the at least one feature (y1 to ym) exclusively from selected sensor data (x1 to xn), wherein for the selected sensor data (x1 to xn) operating parameter values, in particular in the form of position values, speed values and torque values, lie within specified limits.

10. Drive system (100), comprising: a frequency converter (1) according to one of the preceding claims, and a Kl unit (10) formed separately from the frequency converter (1), wherein the at least one feature (y1 to ym) generated by means of the frequency converter (1) is provided to form an input variable of the Kl unit (10).

11. Drive system (100) according to claim 10, characterized in that the Kl unit (10) is designed to determine a state (Z1 to Zk) of the drive system (100) based on the at least one feature.

Images