Digital factory management system

Abstract

The present invention discloses a digital factory management system, which belongs to the technical field of factory management. Aiming at the needs of automated factories, the present invention collects the vibration signals of various components of equipment processing in real time through an acquisition subsystem. In the automated factory environment, the collected vibration signals are deeply processed to extract edge enhancement signals, internal edge signals and external edge signals, enhance the edge information of the signals, and highlight the internal details and external contours. Then, signal component sequences are constructed for these three signals respectively to reflect the characteristics of each signal. A multi-layer input anomaly detection model is used to process the component sequences of the three signals respectively to obtain abnormal values, which represent the abnormal conditions of the working of the components. Combined with the abnormal values ​​of each component of the equipment, the processing quality score of the automated factory equipment is obtained, thereby improving the processing quality assessment accuracy of the automated factory, significantly improving production efficiency and product quality, and promoting the intelligent development of automated factories.

Description

Technical Field

[0001] The present invention relates to the technical field of factory management, and in particular to a digital factory management system. Background Art

[0002] With the rapid development of industry and intelligent manufacturing, modern automated factories are undergoing digital transformation. In this process, the health status monitoring and processing quality control of equipment in automated factories are becoming increasingly important. Traditional manual inspection methods often have problems such as low efficiency, insufficient accuracy and high cost, which are particularly prominent in automated factories that pursue high efficiency and high precision.

[0003] In industrial production and automated factory environments, the vibration signals of equipment contain a lot of information about their operating status and processing quality. However, these signals are usually complex, contain a lot of noise and interference, and are difficult to analyze directly. In addition, different types of faults and quality problems may manifest themselves in different aspects of the signal, such as edge features, internal structure, etc. This complexity places higher demands on the monitoring and management systems of automated factories. Traditional signal analysis methods often only focus on certain specific features in the time domain or frequency domain, and it is difficult to capture the comprehensive information of the signal. Therefore, there is a problem of low accuracy in processing quality assessment in automated factories. Summary of the invention

[0004] In view of the above-mentioned deficiencies in the prior art, the present invention provides a digital factory management system that solves the problem of low precision in machining quality assessment in the prior art.

[0005] In order to achieve the above-mentioned invention object, the technical solution adopted by the present invention is: a digital factory management system, including: an acquisition subsystem, a signal extraction subsystem, a sequence construction subsystem, an anomaly detection subsystem and an evaluation subsystem;

[0006] The acquisition subsystem is used to collect vibration signals of various components during equipment processing in the factory;

[0007] The signal extraction subsystem is used to extract edge enhancement signals, internal edge signals and external edge signals from the vibration signal;

[0008] The sequence construction subsystem is used to extract instantaneous amplitude features, instantaneous frequency features and instantaneous phase features of the edge enhancement signal, the internal edge signal and the external edge signal, respectively, and to construct signal component sequences of the edge enhancement signal, the internal edge signal and the external edge signal respectively;

[0009] The anomaly detection subsystem is used to obtain an anomaly value based on a multi-layer input anomaly detection model according to a signal component sequence of an edge enhancement signal, an internal edge signal, and an external edge signal of each component;

[0010] The evaluation subsystem is used to obtain a processing quality score based on the abnormal values ​​of each component of the equipment.

[0011] Further, the signal extraction subsystem includes: an erosion operation unit, an expansion operation unit, a first signal extraction unit, a second signal extraction unit and a third signal extraction unit;

[0012] The corrosion operation unit is used to perform corrosion operation on the vibration signal to obtain a corrosion result;

[0013] The expansion operation unit is used to perform expansion operation on the vibration signal to obtain an expansion result;

[0014] The first signal extraction unit is used to subtract the dilation result from the erosion result to obtain an edge enhancement signal;

[0015] The second signal extraction unit is used to subtract the vibration signal from the corrosion result to obtain an internal edge signal;

[0016] The third signal extraction unit is used to subtract the expansion result from the vibration signal to obtain an external edge signal.

[0017] Further, the sequence construction subsystem includes: a decomposition unit, an instantaneous feature calculation unit, a component feature calculation unit and a sequence construction unit;

[0018] The decomposition unit is used to process the original signal using empirical mode decomposition (EMD) to obtain multiple intrinsic mode functions, wherein the original signal is an edge enhancement signal, an internal edge signal or an external edge signal;

[0019] The instantaneous characteristic calculation unit is used to calculate the instantaneous amplitude, instantaneous frequency and instantaneous phase of each intrinsic mode function;

[0020] The component characteristic calculation unit is used to calculate the signal component characteristic value of each intrinsic mode function according to the instantaneous amplitude, instantaneous frequency and instantaneous phase of each intrinsic mode function;

[0021] The sequence construction unit is used to form a signal component sequence from the signal component eigenvalues ​​of each intrinsic mode function.

[0022] Furthermore, the expression of the component characteristic calculation unit is: , where C is the characteristic value of the signal component, AM i is the instantaneous amplitude of the eigenmode function, F i is the i-th instantaneous frequency on the eigenmode function, θ i is the i-th instantaneous phase on the eigenmode function, i is a positive integer, and N is the number of instantaneous amplitudes, instantaneous frequencies, or instantaneous phases.

[0023] Further, the multi-layer input anomaly detection model includes: a first one-dimensional feature extraction unit, a second one-dimensional feature extraction unit, a third one-dimensional feature extraction unit, a feature fusion layer, a hidden layer and an output layer;

[0024] The input end of the first one-dimensional feature extraction unit is used to input the signal component sequence of the edge enhancement signal; the input end of the second one-dimensional feature extraction unit is used to input the signal component sequence of the internal edge signal; the input end of the third one-dimensional feature extraction unit is used to input the signal component sequence of the external edge signal;

[0025] The input end of the feature fusion layer is respectively connected to the output end of the first one-dimensional feature extraction unit, the output end of the second one-dimensional feature extraction unit and the output end of the third one-dimensional feature extraction unit, and its output end is connected to the input end of the hidden layer; the input end of the output layer is connected to the output end of the hidden layer, and its output end serves as the output end of the multi-layer input anomaly detection model.

[0026] Further, the first one-dimensional feature extraction unit, the second one-dimensional feature extraction unit and the third one-dimensional feature extraction unit each include: a fully connected layer and a feature enhancement layer;

[0027] The fully connected layer is used to assign weights and biases to each element in the signal component sequence to obtain a feature sequence;

[0028] The feature enhancement layer is used to perform feature enhancement processing on the feature sequence to obtain an enhanced feature sequence.

[0029] Furthermore, the expression of the feature enhancement layer is: , , where r n To enhance the nth element in the feature sequence, γ n is the nth enhancement coefficient, || is the absolute value operation, x n is the nth element in the signal component sequence, h n is the nth element in the feature sequence, It is the nth element in the standard sequence of signal components.

[0030] Furthermore, the expression of the feature fusion layer is: , where Y is the output sequence of the feature fusion layer, R1 is the enhanced feature sequence output by the first one-dimensional feature extraction unit, R2 is the enhanced feature sequence output by the second one-dimensional feature extraction unit, and R3 is the enhanced feature sequence output by the third one-dimensional feature extraction unit. is element-wise multiplication.

[0031] Furthermore, the expression of the output layer is: , where y is the abnormal value output by the output layer, h m is the output of the mth hidden node in the hidden layer, ω m h m The weight of b m h m , M is the number of hidden nodes, and m is a positive integer.

[0032] The beneficial effects of the present invention are as follows: the present invention is responsible for real-time collection of vibration signals of various components of the equipment processing through the collection subsystem, and the collected vibration signals are deeply processed to extract edge enhancement signals, internal edge signals and external edge signals, enhance the edge information of the signal, highlight the internal details and external contours of the signal, and then construct a signal component sequence for the edge enhancement signal, the internal edge signal and the external edge signal respectively to reflect the characteristics of each signal. A multi-layer input anomaly detection model is used to process the signal component sequences of the three signals respectively to obtain abnormal values ​​representing the abnormal working conditions of the component. The processing quality score is obtained by combining the abnormal values ​​of various components of the equipment. The present invention constructs edge enhancement signals, internal edge signals and external edge signals for vibration signals, so that the constructed signal component sequence fully mines the signal characteristics, captures the comprehensive information of the signal, improves the detection accuracy of abnormal values, and thus improves the processing quality assessment accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 It is a system block diagram of a digital factory management system;

[0034] Figure 2 Schematic diagram of the structure of the multi-layer input anomaly detection model. DETAILED DESCRIPTION

[0035] The specific implementation modes of the present invention are described below so that those skilled in the art can understand the present invention. However, it should be clear that the present invention is not limited to the scope of the specific implementation modes. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the attached claims, these changes are obvious, and all inventions and creations utilizing the concept of the present invention are protected.

[0036] like Figure 1 As shown, a digital factory management system includes: an acquisition subsystem, a signal extraction subsystem, a sequence construction subsystem, an anomaly detection subsystem and an evaluation subsystem;

[0037] The acquisition subsystem is used to collect vibration signals of various components during equipment processing in the factory;

[0038] The signal extraction subsystem is used to extract edge enhancement signals, internal edge signals and external edge signals from the vibration signal;

[0039] The sequence construction subsystem is used to extract instantaneous amplitude features, instantaneous frequency features and instantaneous phase features of the edge enhancement signal, the internal edge signal and the external edge signal, respectively, and to construct signal component sequences of the edge enhancement signal, the internal edge signal and the external edge signal respectively;

[0040] The anomaly detection subsystem is used to obtain an anomaly value based on a multi-layer input anomaly detection model according to a signal component sequence of an edge enhancement signal, an internal edge signal, and an external edge signal of each component;

[0041] The evaluation subsystem is used to obtain a processing quality score based on the abnormal values ​​of each component of the equipment.

[0042] The present invention is aimed at the needs of automated factories. The acquisition subsystem is responsible for real-time acquisition of vibration signals of various components of equipment processing in automated factories, and the acquired vibration signals are deeply processed to extract edge enhancement signals, internal edge signals and external edge signals, enhance the edge information of the signals, highlight the internal details and external contours of the signals, and then construct signal component sequences for the edge enhancement signals, internal edge signals and external edge signals respectively to reflect the characteristics of each signal. The signal component sequences of the three signals are processed respectively using a multi-layer input anomaly detection model to obtain abnormal values, which represent the abnormal conditions of the working of the components. This process can provide more accurate equipment status assessment for automated factories. Combined with the abnormal values ​​of various components of the equipment, a processing quality score is obtained, providing a reliable basis for quality control of automated factories.

[0043] In this embodiment, the equipment includes: CNC machine tools, punching machines, welding equipment, assembly lines, cutting equipment, packaging equipment, etc.

[0044] In this embodiment, the components include: a motor, a bearing, a gear box, a hydraulic system, a pneumatic system, a tool, a transmission system, etc.

[0045] The signal extraction subsystem comprises: an erosion operation unit, an expansion operation unit, a first signal extraction unit, a second signal extraction unit and a third signal extraction unit;

[0046] The corrosion operation unit is used to perform corrosion operation on the vibration signal to obtain a corrosion result;

[0047] The expansion operation unit is used to perform expansion operation on the vibration signal to obtain an expansion result;

[0048] The first signal extraction unit is used to subtract the dilation result from the erosion result to obtain an edge enhancement signal;

[0049] The second signal extraction unit is used to subtract the vibration signal from the corrosion result to obtain an internal edge signal;

[0050] The third signal extraction unit is used to subtract the expansion result from the vibration signal to obtain an external edge signal.

[0051] The present invention can remove noise in the vibration signal and smooth the signal waveform through corrosion and expansion operations, thereby improving the signal-to-noise ratio of the signal. In the corrosion operation, a smoother signal contour is obtained by suppressing isolated point noise, and the expansion operation helps to restore important features and make the main components of the signal more obvious.

[0052] The first signal extraction unit effectively enhances the edge features of the signal by calculating the difference between the expansion result and the corrosion result. This is particularly useful for detecting sudden changes or rapid changes in vibration signals, and helps identify potential faults or abnormal conditions.

[0053] The second signal extraction unit and the third signal extraction unit can respectively extract the internal structure changes and the external contour changes of the signal. In this way, the structure and change characteristics of the signal can be fully captured.

[0054] The sequence construction subsystem includes: a decomposition unit, an instantaneous feature calculation unit, a component feature calculation unit and a sequence construction unit;

[0055] The decomposition unit is used to process the original signal using empirical mode decomposition (EMD) to obtain multiple intrinsic mode functions, wherein the original signal is an edge enhancement signal, an internal edge signal or an external edge signal;

[0056] The instantaneous characteristic calculation unit is used to calculate the instantaneous amplitude, instantaneous frequency and instantaneous phase of each intrinsic mode function;

[0057] The component characteristic calculation unit is used to calculate the signal component characteristic value of each intrinsic mode function according to the instantaneous amplitude, instantaneous frequency and instantaneous phase of each intrinsic mode function;

[0058] The sequence construction unit is used to form a signal component sequence from the signal component eigenvalues ​​of each intrinsic mode function.

[0059] The present invention can effectively decompose complex nonlinear and non-stationary signals into multiple intrinsic mode functions (IMFs) by performing empirical mode decomposition on the original signal through a decomposition unit. The calculation of instantaneous amplitude, instantaneous frequency and instantaneous phase enables the system to fully capture the dynamic change characteristics of the intrinsic mode and reflect the specific state of the signal at different time points. Then, according to the instantaneous amplitude, instantaneous frequency and instantaneous phase of each intrinsic mode function, the signal component characteristic value of each intrinsic mode function is obtained, and the amplitude, frequency and phase characteristics of the intrinsic mode function are reflected through the signal component characteristic value.

[0060] The expression of the component characteristic calculation unit is: , where C is the characteristic value of the signal component, AM i is the instantaneous amplitude of the eigenmode function, F i is the i-th instantaneous frequency on the eigenmode function, θ i is the i-th instantaneous phase of the eigenmode function, i is a positive integer, and N is the number of instantaneous amplitudes, instantaneous frequencies, or instantaneous phases.

[0061] The present invention accumulates the amplitude, frequency and phase according to the weight of each amplitude, frequency and phase in each of them, so as to reflect the signal component characteristics of each intrinsic mode function.

[0062] like Figure 2 As shown, the multi-layer input anomaly detection model includes: a first one-dimensional feature extraction unit, a second one-dimensional feature extraction unit, a third one-dimensional feature extraction unit, a feature fusion layer, a hidden layer and an output layer;

[0063] The input end of the first one-dimensional feature extraction unit is used to input the signal component sequence of the edge enhancement signal; the input end of the second one-dimensional feature extraction unit is used to input the signal component sequence of the internal edge signal; the input end of the third one-dimensional feature extraction unit is used to input the signal component sequence of the external edge signal;

[0064] The input end of the feature fusion layer is respectively connected to the output end of the first one-dimensional feature extraction unit, the output end of the second one-dimensional feature extraction unit and the output end of the third one-dimensional feature extraction unit, and its output end is connected to the input end of the hidden layer; the input end of the output layer is connected to the output end of the hidden layer, and its output end serves as the output end of the multi-layer input anomaly detection model.

[0065] The present invention sets up three one-dimensional feature extraction units, so that the model can extract features from different signal component sequences (edge ​​enhancement signals, internal edge signals and external edge signals). This multi-angle feature extraction can enhance the model's understanding of signal characteristics and provide a more comprehensive information basis.

[0066] The design of the feature fusion layer enables the outputs from different feature extraction units to be concentrated and integrated to form a more representative feature vector.

[0067] In this embodiment, Sigmoid is selected as the activation function of the hidden layer.

[0068] The first one-dimensional feature extraction unit, the second one-dimensional feature extraction unit and the third one-dimensional feature extraction unit all include: a fully connected layer and a feature enhancement layer;

[0069] The fully connected layer is used to assign weights and biases to each element in the signal component sequence to obtain a feature sequence;

[0070] The feature enhancement layer is used to perform feature enhancement processing on the feature sequence to obtain an enhanced feature sequence.

[0071] In this embodiment, the fully connected layer is used to assign weights and biases to each element in the signal component sequence to obtain a feature sequence, the length of which is the same as the length of the signal component sequence.

[0072] The expression of the feature enhancement layer is: , , where r n To enhance the nth element in the feature sequence, γ n is the nth enhancement coefficient, || is the absolute value operation, x n is the nth element in the signal component sequence, h n is the nth element in the feature sequence, It is the nth element in the standard sequence of signal components.

[0073] The present invention calculates an enhancement coefficient according to the difference between a signal component sequence and a signal component standard sequence, and then enhances the elements in the signal component sequence according to the enhancement coefficient, thereby paying more attention to the elements in the signal component sequence that have a large difference with the elements in the signal component standard sequence, thereby increasing the attention of the elements with large deviations.

[0074] In the first one-dimensional feature extraction unit, the standard sequence of signal components is the edge enhancement signal obtained based on the vibration signal of the normal device, and the signal component sequence is obtained based on the edge enhancement signal. In the second one-dimensional feature extraction unit, the standard sequence of signal components is the internal edge signal obtained based on the vibration signal of the normal device, and the signal component sequence is obtained based on the internal edge signal. In the third one-dimensional feature extraction unit, the standard sequence of signal components is the external edge signal obtained based on the vibration signal of the normal device, and the signal component sequence is obtained based on the external edge signal.

[0075] The signal component sequence is the signal feature sequence actually collected during the operation of the device under test, and the signal component standard sequence is the signal feature sequence collected by a normal device under the same conditions.

[0076] The expression of the feature fusion layer is: , where Y is the output sequence of the feature fusion layer, R1 is the enhanced feature sequence output by the first one-dimensional feature extraction unit, R2 is the enhanced feature sequence output by the second one-dimensional feature extraction unit, and R3 is the enhanced feature sequence output by the third one-dimensional feature extraction unit. is element-wise multiplication.

[0077] The present invention fuses the enhanced feature sequences output by three one-dimensional feature extraction units to achieve multi-dimensional feature fusion, which can provide a more comprehensive signal feature representation and enhance the model's ability to understand complex signals.

[0078] The expression of the output layer is: , where y is the abnormal value output by the output layer, h m is the output of the mth hidden node in the hidden layer, ω m h m The weight of b m h m , M is the number of hidden nodes, and m is a positive integer.

[0079] In this embodiment, the weights and biases in the multi-layer input anomaly detection model are trained by the gradient descent method.

[0080] In this embodiment, the evaluation subsystem accumulates the abnormal values ​​of each component of the equipment to obtain a processing quality score.

[0081] The present invention constructs edge enhancement signals, internal edge signals and external edge signals for vibration signals, so that the constructed signal component sequence fully exploits the signal characteristics, captures the comprehensive information of the signal, and improves the detection accuracy of abnormal values, thereby improving the processing quality assessment accuracy in the automated factory. It is particularly suitable for the high-precision requirements of the automated factory and can significantly improve the production efficiency and product quality of the automated factory. By applying the present invention in the automated factory environment, more accurate equipment monitoring and more efficient quality control can be achieved, further promoting the intelligent development of the automated factory.

[0082] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A digital factory management system, characterized in that: include: Acquisition subsystem, signal extraction subsystem, sequence construction subsystem, anomaly detection subsystem and evaluation subsystem; The acquisition subsystem is used to collect vibration signals of various components during equipment processing in the factory; The signal extraction subsystem is used to extract edge enhancement signals, internal edge signals and external edge signals from the vibration signal; The sequence construction subsystem is used to extract instantaneous amplitude features, instantaneous frequency features and instantaneous phase features of the edge enhancement signal, the internal edge signal and the external edge signal, respectively, and to construct signal component sequences of the edge enhancement signal, the internal edge signal and the external edge signal respectively; The anomaly detection subsystem is used to obtain an anomaly value based on a multi-layer input anomaly detection model according to a signal component sequence of an edge enhancement signal, an internal edge signal, and an external edge signal of each component; The evaluation subsystem is used to obtain a processing quality score based on the abnormal values ​​of each component of the equipment; The signal extraction subsystem comprises: an erosion operation unit, an expansion operation unit, a first signal extraction unit, a second signal extraction unit and a third signal extraction unit; The corrosion operation unit is used to perform corrosion operation on the vibration signal to obtain a corrosion result; The expansion operation unit is used to perform expansion operation on the vibration signal to obtain an expansion result; The first signal extraction unit is used to subtract the dilation result from the erosion result to obtain an edge enhancement signal; The second signal extraction unit is used to subtract the vibration signal from the corrosion result to obtain an internal edge signal; The third signal extraction unit is used to subtract the expansion result from the vibration signal to obtain an external edge signal; The sequence construction subsystem includes: a decomposition unit, an instantaneous feature calculation unit, a component feature calculation unit and a sequence construction unit; The decomposition unit is used to process the original signal using empirical mode decomposition (EMD) to obtain multiple intrinsic mode functions, wherein the original signal is an edge enhancement signal, an internal edge signal or an external edge signal; The instantaneous characteristic calculation unit is used to calculate the instantaneous amplitude, instantaneous frequency and instantaneous phase of each intrinsic mode function; The component characteristic calculation unit is used to calculate the signal component characteristic value of each intrinsic mode function according to the instantaneous amplitude, instantaneous frequency and instantaneous phase of each intrinsic mode function; The sequence construction unit is used to form a signal component sequence from the signal component eigenvalues ​​of each intrinsic mode function; The expression of the component characteristic calculation unit is: , where C is the characteristic value of the signal component, AM i is the instantaneous amplitude of the eigenmode function, F i is the i-th instantaneous frequency on the eigenmode function, θ i is the i-th instantaneous phase of the eigenmode function, i is a positive integer, and N is the number of instantaneous amplitudes, instantaneous frequencies, or instantaneous phases.

2. The digital factory management system according to claim 1, characterized in that: The multi-layer input anomaly detection model comprises: a first one-dimensional feature extraction unit, a second one-dimensional feature extraction unit, a third one-dimensional feature extraction unit, a feature fusion layer, a hidden layer and an output layer; The input end of the first one-dimensional feature extraction unit is used to input the signal component sequence of the edge enhancement signal; the input end of the second one-dimensional feature extraction unit is used to input the signal component sequence of the internal edge signal; the input end of the third one-dimensional feature extraction unit is used to input the signal component sequence of the external edge signal; The input end of the feature fusion layer is respectively connected to the output end of the first one-dimensional feature extraction unit, the output end of the second one-dimensional feature extraction unit and the output end of the third one-dimensional feature extraction unit, and its output end is connected to the input end of the hidden layer; the input end of the output layer is connected to the output end of the hidden layer, and its output end serves as the output end of the multi-layer input anomaly detection model.

3. The digital factory management system according to claim 2, characterized in that: The first one-dimensional feature extraction unit, the second one-dimensional feature extraction unit and the third one-dimensional feature extraction unit all include: a fully connected layer and a feature enhancement layer; The fully connected layer is used to assign weights and biases to each element in the signal component sequence to obtain a feature sequence; The feature enhancement layer is used to perform feature enhancement processing on the feature sequence to obtain an enhanced feature sequence.

4. The digital factory management system according to claim 3, characterized in that: The expression of the feature enhancement layer is: , , where r n To enhance the nth element in the feature sequence, γ n is the nth enhancement coefficient, || is the absolute value operation, x n is the nth element in the signal component sequence, h n is the nth element in the feature sequence, It is the nth element in the standard sequence of signal components.

5. The digital factory management system according to claim 2, characterized in that: The expression of the feature fusion layer is: , where Y is the output sequence of the feature fusion layer, R1 is the enhanced feature sequence output by the first one-dimensional feature extraction unit, R2 is the enhanced feature sequence output by the second one-dimensional feature extraction unit, and R3 is the enhanced feature sequence output by the third one-dimensional feature extraction unit. is element-wise multiplication.

6. The digital factory management system according to claim 2, characterized in that: The expression of the output layer is: , where y is the abnormal value output by the output layer, h m is the output of the mth hidden node in the hidden layer, ω m h m The weight of b m h m , M is the number of hidden nodes, and m is a positive integer.

Images