Fault detection method and system applied to motor

Abstract

The invention discloses a fault detection method and system applied to a motor, and the method comprises the steps: obtaining the operation data and external environment data of the motor in real time through sensors disposed inside and outside the motor; performing normalization processing on the operation data and the external environment data, inputting the operation data and the external environment data into a first full-connection neural network model and a second full-connection neural network model, and generating a first vector and a second vector respectively; splicing the first vector and the second vector, performing normalization processing to generate a third vector, and inputting the third vector into a third full-connection neural network model to generate a fourth vector; and inputting the fourth vector into the activation function, generating a detection result, judging the state of the motor according to the detection result, and executing corresponding operation. The method is higher in robustness, universality and accuracy, and fault detection is carried out from the operation condition of the motor and the external environment condition by using a double-tower framework.

Description

Technical Field

[0001] The present invention relates to the technical field of motor fault detection, and in particular to a fault detection method and system applied to a motor. Background Art

[0002] The motor is a device that provides power and is widely used in various fields such as industrial production and agricultural production. During the operation of the motor, some factors often cause the motor to malfunction. Large motor equipment is expensive. If the motor malfunctions and corresponding measures are not discovered in time and taken, it will inevitably lead to further damage to the motor and even cause production accidents.

[0003] However, existing related motor fault detection methods often only perform fault detection based on the motor's own operating conditions, ignoring the impact of the motor's external environmental conditions on the motor, and do not jointly consider the factors affecting the motor's own operating conditions and external environmental conditions. Summary of the invention

[0004] In view of the above-mentioned technical deficiencies, the purpose of the present invention is to provide a fault detection method and system for motors, aiming to solve the problem that the existing related motor fault detection methods often only perform fault detection based on the operating conditions of the motor itself, ignoring the impact of the external environmental conditions of the motor on the motor, and failing to jointly consider the factors affecting the operating conditions of the motor itself and the external environmental conditions.

[0005] In view of the above problems, the present application provides a fault detection method and system for a motor.

[0006] In a first aspect disclosed in the present application, a fault detection method for a motor is provided, the method comprising the following steps: Step 1: Acquire the operation data of the motor in real time through the internal sensor set inside the motor, wherein the operation data includes vibration amplitude, vibration frequency, vibration acceleration, rotor temperature, stator temperature, speed, load and current; Step 2: acquiring external environmental data of the motor in real time through an external sensor disposed outside the motor, the external environmental data including ambient temperature, air humidity, dust mass concentration, and dust particle concentration; Step 3: normalize the operation data and input it into the first fully connected neural network model to generate a first vector, and normalize the external environment data and input it into the second fully connected neural network model to generate a second vector; Step 4: concatenate the first vector and the second vector and perform normalization processing to generate a third vector, and input the third vector into a third fully connected neural network model to generate a fourth vector; Step 5: Input the fourth vector into the activation function to generate a detection result, determine the state of the motor according to the detection result, and perform corresponding operations.

[0007] Preferably, the internal sensors in step 1 include a vibration sensor, a temperature sensor, a rotation speed sensor, a load sensor and a current sensor.

[0008] Preferably, the external sensors in step 2 include a temperature sensor, a humidity sensor and a dust concentration sensor.

[0009] Preferably, the step 3 specifically includes the following steps: Step 3.1: Arrange the operation data in a predetermined order to generate an operation data vector T, and perform maximum and minimum value normalization on the operation data vector T using formula (1) to generate a standard operation data vector T', where T min The vector representing the theoretical minimum value of the motor operation data is generated by arranging it in the same order as the operation data vector T. max The vector representing the theoretical maximum value of the motor operation data is generated by arranging it in the same order as the operation data vector T: T'=(TT min ) / (T max -T min ) formula (1); Step 3.2: Input the standard running data vector T' into the first fully connected neural network model, and calculate and generate the first vector using formula (2), where V represents the first vector, ReLU represents the ReLU activation function, and W 1 and W 2 represents the weight matrix of the first fully connected neural network model, b 1 and b 2 Represents the offset vector of the first fully connected neural network model: V=W 2 ReLU(W 1 T'+b 1 )+b 2 Formula (2); Step 3.3: Arrange the external environment data in order to generate an external environment data vector E, and perform maximum and minimum value normalization on the external environment data vector E using formula (3) to generate a standard external environment data vector E', where E min The vector representing the theoretical minimum value of the motor external environment data is generated by arranging it in the same order as the external environment data vector E. max The vector representing the theoretical maximum value of the motor's external environment data is generated by arranging it in the same order as the external environment data vector E: E'=(EE min ) / (Emax -E min ) formula (3); Step 3.4: Input the standard external environment data vector E' into the second fully connected neural network model, and calculate and generate the second vector using formula (4), where O represents the second vector, ReLU represents the ReLU activation function, and W 3 and W 4 represents the weight matrix of the second fully connected neural network model, b 3 and b 4 Represents the offset vector of the second fully connected neural network model: O=W 4 ReLU(W 3 E'+b 3 )+b 4 Formula (4).

[0010] Preferably, the step 4 specifically includes the following steps: Step 4.1: Perform a vector concatenation operation on the first vector and the second vector to generate a concatenated vector, and perform layer normalization on the concatenated vector using formula (5) to generate a third vector, where C i represents the i-th component of the third vector, P i represents the i-th component of the splicing vector, γ represents the scaling factor, which is used to control the scale after normalization, β represents the offset factor, which is used to control the displacement after normalization, μ represents the average value of all components in the third vector, and σ represents the standard deviation of all components in the third vector: C i =γ(P i -μ) / σ+β (5); Step 4.2: Input the third vector into the third fully connected neural network model, and calculate and generate the fourth vector using formula (6), where L represents the fourth vector, C represents the third vector, ReLU represents the ReLU activation function, and W 5 and W 6 represents the weight matrix of the third fully connected neural network model, b 5 and b 6 Represents the offset vector of the third fully connected neural network model: L=W 6 ReLU(W 5 C+b 5 )+b 6 Formula (6).

[0011] Preferably, the step 5 specifically includes the following steps: Step 5.1: Input the fourth vector into the softmax activation function to generate a detection result, wherein the fourth vector and the detection result are both vectors of dimension 3, and each component of the detection result is the probability of various states of the motor; Step 5.2: If the first component of the test result is the maximum value among all the components of the test result, it is determined that the motor operation state is normal. If the second component of the test result is the maximum value among all the components of the test result, it is determined that the motor operation state is abnormal and a warning signal is issued. If the third component of the test result is the maximum value among all the components of the test result, it is determined that the motor operation state fails, the motor is controlled to cut off power and stop running, and a fault signal is issued.

[0012] A second aspect disclosed in the present application provides a fault detection system applied to a motor, the system being used for the above-mentioned fault detection method applied to a motor, the system comprising: An internal acquisition module, the internal acquisition module is used to obtain the operation data of the motor in real time through an internal sensor arranged inside the motor, the operation data including vibration amplitude, vibration frequency, vibration acceleration, rotor temperature, stator temperature, speed, load and current; An external acquisition module, the external acquisition module is used to obtain external environmental data of the motor in real time through an external sensor arranged outside the motor, the external environmental data including ambient temperature, air humidity, dust mass concentration and dust particle concentration; A dual-tower module, wherein the dual-tower module is used to input the operation data into a first fully connected neural network model after normalization processing to generate a first vector, and input the external environment data into a second fully connected neural network model after normalization processing to generate a second vector; An integration module, wherein the integration module is used to concatenate the first vector and the second vector and perform normalization processing to generate a third vector, and input the third vector into a third fully connected neural network model to generate a fourth vector; A determination module is used to input the fourth vector into the activation function, generate a detection result, determine the state of the motor according to the detection result, and perform a corresponding operation.

[0013] The beneficial effects of the present invention are: (1) The fault detection robustness, universality and accuracy of the detection method of the present invention are higher. It not only performs fault detection based on the operating conditions of the motor itself, but also considers the impact of the external environmental conditions of the motor on the motor, as well as the trigger threshold of fault detection under different external environmental conditions.

[0014] (2) A dual-tower architecture is used to calculate the motor's own operating conditions and the factors affecting the external environmental conditions separately, which reduces the amount of calculation while avoiding data confusion and making the detection more accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0016] Figure 1 The figure is an overall flow chart of a fault detection method applied to a motor.

[0017] Figure 2 The figure is an overall structural diagram of a fault detection system applied to a motor. DETAILED DESCRIPTION

[0018] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0019] like Figure 1 As shown, an embodiment of the present application provides a fault detection method applied to a motor, the method comprising the following steps: Step 1: Acquire the operation data of the motor in real time through the internal sensor set inside the motor, wherein the operation data includes vibration amplitude, vibration frequency, vibration acceleration, rotor temperature, stator temperature, speed, load and current; Step 2: acquiring external environmental data of the motor in real time through an external sensor disposed outside the motor, the external environmental data including ambient temperature, air humidity, dust mass concentration, and dust particle concentration; Step 3: normalize the operation data and input it into the first fully connected neural network model to generate a first vector, and normalize the external environment data and input it into the second fully connected neural network model to generate a second vector; Step 4: concatenate the first vector and the second vector and perform normalization processing to generate a third vector, and input the third vector into a third fully connected neural network model to generate a fourth vector; Step 5: Input the fourth vector into the activation function to generate a detection result, determine the state of the motor according to the detection result, and perform corresponding operations.

[0020] Furthermore, the internal sensors in step 1 include a vibration sensor, a temperature sensor, a rotation speed sensor, a load sensor and a current sensor.

[0021] Furthermore, the external sensors in step 2 include a temperature sensor, a humidity sensor and a dust concentration sensor.

[0022] Furthermore, the step 3 specifically includes the following steps: Step 3.1: Arrange the operation data in a predetermined order to generate an operation data vector T, and perform maximum and minimum value normalization on the operation data vector T using formula (1) to generate a standard operation data vector T', where T min The vector representing the theoretical minimum value of the motor operation data is generated by arranging it in the same order as the operation data vector T. max The vector representing the theoretical maximum value of the motor operation data is generated by arranging it in the same order as the operation data vector T: T'=(TT min ) / (T max -T min ) formula (1); Step 3.2: Input the standard running data vector T' into the first fully connected neural network model, and calculate and generate the first vector using formula (2), where V represents the first vector, ReLU represents the ReLU activation function, and W 1 and W 2 represents the weight matrix of the first fully connected neural network model, b 1 and b 2 Represents the offset vector of the first fully connected neural network model: V=W 2 ReLU(W 1 T'+b 1 )+b 2 Formula (2); Step 3.3: Arrange the external environment data in order to generate an external environment data vector E, and perform maximum and minimum value normalization on the external environment data vector E using formula (3) to generate a standard external environment data vector E', where E min The vector representing the theoretical minimum value of the motor external environment data is generated by arranging it in the same order as the external environment data vector E. max The vector representing the theoretical maximum value of the motor's external environment data is generated by arranging it in the same order as the external environment data vector E: E'=(EE min ) / (E max -E min ) formula (3); Step 3.4: Input the standard external environment data vector E' into the second fully connected neural network model, and calculate and generate the second vector using formula (4), where O represents the second vector, ReLU represents the ReLU activation function, and W 3 and W 4 represents the weight matrix of the second fully connected neural network model, b 3 and b 4 Represents the offset vector of the second fully connected neural network model: O=W 4 ReLU(W 3 E'+b 3 )+b 4 Formula (4).

[0023] Specifically, maximum and minimum value normalization is also called linear normalization. It scales the value to the interval [0,1] through a linear transformation. It is simple to calculate and is the simplest and fastest normalization method. It can help the model converge faster during training and make the model more accurate when performing fault detection. It also prevents numerical instability or gradient vanishing and gradient exploding problems caused by large differences in numerical scales. Using a dual-tower architecture consisting of a first fully connected neural network model and a second fully connected neural network model, the operating conditions of the motor itself and the factors affecting external environmental conditions are calculated separately, which can reduce the amount of calculation while avoiding data confusion caused by mixing data together.

[0024] Furthermore, the step 4 specifically includes the following steps: Step 4.1: Perform a vector concatenation operation on the first vector and the second vector to generate a concatenated vector, and perform layer normalization on the concatenated vector using formula (5) to generate a third vector, where C i represents the i-th component of the third vector, P i represents the i-th component of the splicing vector, γ represents the scaling factor, which is used to control the scale after normalization, β represents the offset factor, which is used to control the displacement after normalization, μ represents the average value of all components in the third vector, and σ represents the standard deviation of all components in the third vector: C i =γ(P i -μ) / σ+β (5); Step 4.2: Input the third vector into the third fully connected neural network model, and calculate and generate the fourth vector using formula (6), where L represents the fourth vector, C represents the third vector, ReLU represents the ReLU activation function, and W 5 and W 6 represents the weight matrix of the third fully connected neural network model, b 5 and b 6Represents the offset vector of the third fully connected neural network model: L=W 6 ReLU(W 5 C+b 5 )+b 6 Formula (6).

[0025] Specifically, by normalizing the concatenated vectors, the value of each component of the third vector is kept within a stable range to prevent the gradient from disappearing or exploding due to excessive or small values. ReLU is the rectified linear unit activation function. For the weight matrix W 1 , W 2 , W 3 , W 4 , W 5 and W 6 and the offset vector b 1 、b 2 、b 3 、b 4 、b 5 and b 6 , by selecting from a Gaussian distribution with standard deviation 0.01 and mean 0 N Random sampling is performed in (0, 0.01) to initialize the parameters in each weight matrix. These weight matrices and offset vectors are the goals of subsequent model training and are also the core parameters for fault detection.

[0026] Furthermore, the step 5 specifically includes the following steps: Step 5.1: Input the fourth vector into the softmax activation function to generate a detection result, wherein the fourth vector and the detection result are both vectors of dimension 3, and each component of the detection result is the probability of various states of the motor; Step 5.2: If the first component of the test result is the maximum value among all the components of the test result, it is determined that the motor operation state is normal. If the second component of the test result is the maximum value among all the components of the test result, it is determined that the motor operation state is abnormal and a warning signal is issued. If the third component of the test result is the maximum value among all the components of the test result, it is determined that the motor operation state fails, the motor is controlled to cut off power and stop running, and a fault signal is issued.

[0027] Specifically, the first fully connected neural network model, the second fully connected neural network model and the third fully connected neural network model need to be trained. The data set for model training adopts a three-class sample balancing strategy. The sample labels of normal motor operation, abnormal motor operation and motor operation failure are set to [1, 0, 0] respectively. T , [0,1,0] T and [0,0,1]T The loss function uses the multivariate cross entropy loss function. The ratio of the training set, test set, and validation set is 8:1:1. The Adam optimizer is used for model training. The initial learning rate is set to 0.001 and the batch size is set to 128.

[0028] In summary, the fault detection method for a motor provided in the embodiment of the present application has the following technical effects: (1) The fault detection robustness, universality and accuracy of the detection method of the present invention are higher. It not only performs fault detection based on the operating conditions of the motor itself, but also considers the impact of the external environmental conditions of the motor on the motor, as well as the trigger threshold of fault detection under different external environmental conditions.

[0029] (2) A dual-tower architecture is used to calculate the motor's own operating conditions and the factors affecting the external environmental conditions separately, which reduces the amount of calculation while avoiding data confusion and making the detection more accurate.

[0030] Based on the same inventive concept as the fault detection method applied to a motor in the aforementioned embodiment, Figure 2 As shown, the present application provides a fault detection system applied to a motor, the system comprising: An internal acquisition module, the internal acquisition module is used to obtain the operation data of the motor in real time through an internal sensor arranged inside the motor, the operation data including vibration amplitude, vibration frequency, vibration acceleration, rotor temperature, stator temperature, speed, load and current; An external acquisition module, the external acquisition module is used to obtain external environmental data of the motor in real time through an external sensor arranged outside the motor, the external environmental data including ambient temperature, air humidity, dust mass concentration and dust particle concentration; A dual-tower module, wherein the dual-tower module is used to input the operation data into a first fully connected neural network model after normalization processing to generate a first vector, and input the external environment data into a second fully connected neural network model after normalization processing to generate a second vector; An integration module, wherein the integration module is used to concatenate the first vector and the second vector and perform normalization processing to generate a third vector, and input the third vector into a third fully connected neural network model to generate a fourth vector; A determination module is used to input the fourth vector into the activation function, generate a detection result, determine the state of the motor according to the detection result, and perform a corresponding operation.

[0031] Through the above detailed description of a fault detection method applied to a motor in this specification, those skilled in the art can clearly know a fault detection system applied to a motor in this embodiment. Since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part description.

[0032] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, 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.

[0033] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A fault detection method applied to a motor, characterized in that: The method comprises: Step 1: Acquire the operation data of the motor in real time through the internal sensor set inside the motor, wherein the operation data includes vibration amplitude, vibration frequency, vibration acceleration, rotor temperature, stator temperature, speed, load and current; Step 2: acquiring external environmental data of the motor in real time through an external sensor disposed outside the motor, the external environmental data including ambient temperature, air humidity, dust mass concentration, and dust particle concentration; Step 3: normalize the operation data and input it into the first fully connected neural network model to generate a first vector, and normalize the external environment data and input it into the second fully connected neural network model to generate a second vector; Step 4: concatenate the first vector and the second vector and perform normalization processing to generate a third vector, and input the third vector into a third fully connected neural network model to generate a fourth vector; Step 5: Input the fourth vector into the activation function to generate a detection result, determine the state of the motor according to the detection result, and perform corresponding operations.

2. A fault detection method for a motor as claimed in claim 1, characterized in that: The internal sensors described in step 1 include a vibration sensor, a temperature sensor, a rotation speed sensor, a load sensor and a current sensor.

3. A fault detection method for a motor as claimed in claim 1, characterized in that: The external sensors in step 2 include a temperature sensor, a humidity sensor, and a dust concentration sensor.

4. A fault detection method for a motor as claimed in claim 1, characterized in that: The step 3 specifically includes the following steps: Step 3.1: Arrange the operation data in a predetermined order to generate an operation data vector T, and perform maximum and minimum value normalization on the operation data vector T using formula (1) to generate a standard operation data vector T', where T min The vector representing the theoretical minimum value of the motor operation data is generated by arranging it in the same order as the operation data vector T. max The vector representing the theoretical maximum value of the motor operation data is generated by arranging it in the same order as the operation data vector T: T'=(T - T min ) / (T max - T min ) Equation (1); Step 3.2: Input the standard running data vector T' into the first fully connected neural network model, and calculate and generate the first vector using formula (2), where V represents the first vector, ReLU represents the ReLU activation function, and W 1 and W 2 represents the weight matrix of the first fully connected neural network model, b 1 and b 2 Represents the offset vector of the first fully connected neural network model: V = W 2 ReLU(W 1 T' + b 1 ) + b 2 Equation (2); Step 3.3: Arrange the external environment data in order to generate an external environment data vector E, and perform maximum and minimum value normalization on the external environment data vector E using formula (3) to generate a standard external environment data vector E', where E min The vector representing the theoretical minimum value of the motor external environment data is generated by arranging it in the same order as the external environment data vector E. max The vector representing the theoretical maximum value of the motor's external environment data is generated by arranging it in the same order as the external environment data vector E: E'=(E - E min ) / (E max - E min ) Equation (3); Step 3.4: Input the standard external environment data vector E' into the second fully connected neural network model, and calculate and generate the second vector using formula (4), where O represents the second vector, ReLU represents the ReLU activation function, and W 3 and W 4 represents the weight matrix of the second fully connected neural network model, b 3 and b 4 Represents the offset vector of the second fully connected neural network model: O = W 4 ReLU(W 3 E' + b 3 ) + b 4 Equation (4).

5. A fault detection method for a motor as claimed in claim 1, characterized in that: The step 4 specifically comprises the following steps: Step 4.1: Perform a vector concatenation operation on the first vector and the second vector to generate a concatenated vector, and perform layer normalization on the concatenated vector using formula (5) to generate a third vector, where C i represents the i-th component of the third vector, P i represents the i-th component of the splicing vector, γ represents the scaling factor, which is used to control the scale after normalization, β represents the offset factor, which is used to control the displacement after normalization, μ represents the average value of all components in the third vector, and σ represents the standard deviation of all components in the third vector: C i =γ(P i -μ) / σ+β formula (5); Step 4.2: Input the third vector into the third fully connected neural network model, and calculate and generate the fourth vector using formula (6), where L represents the fourth vector, C represents the third vector, ReLU represents the ReLU activation function, and W 5 and W 6 represents the weight matrix of the third fully connected neural network model, b 5 and b 6 Represents the offset vector of the third fully connected neural network model: L = W 6 ReLU(W 5 C + b 5 ) + b 6 Equation (6).

6. A fault detection method for a motor as claimed in claim 1, characterized in that: The step 5 specifically comprises the following steps: Step 5.1: Input the fourth vector into the softmax activation function to generate a detection result, wherein the fourth vector and the detection result are both vectors of dimension 3, and each component of the detection result is the probability of various states of the motor; Step 5.2: If the first component of the test result is the maximum value among all the components of the test result, it is determined that the motor operation state is normal. If the second component of the test result is the maximum value among all the components of the test result, it is determined that the motor operation state is abnormal and a warning signal is issued. If the third component of the test result is the maximum value among all the components of the test result, it is determined that the motor operation state fails, the motor is controlled to cut off power and stop running, and a fault signal is issued.

7. A fault detection system for a motor, the system comprising: An internal acquisition module, the internal acquisition module is used to obtain the operation data of the motor in real time through an internal sensor arranged inside the motor, the operation data including vibration amplitude, vibration frequency, vibration acceleration, rotor temperature, stator temperature, speed, load and current; An external acquisition module, the external acquisition module is used to obtain external environmental data of the motor in real time through an external sensor arranged outside the motor, the external environmental data including ambient temperature, air humidity, dust mass concentration and dust particle concentration; A dual-tower module, wherein the dual-tower module is used to input the operation data into a first fully connected neural network model after normalization processing to generate a first vector, and input the external environment data into a second fully connected neural network model after normalization processing to generate a second vector; An integration module, wherein the integration module is used to concatenate the first vector and the second vector and perform normalization processing to generate a third vector, and input the third vector into a third fully connected neural network model to generate a fourth vector; A determination module is used to input the fourth vector into the activation function, generate a detection result, determine the state of the motor according to the detection result, and perform a corresponding operation.

Images