A mill load monitoring system and method based on shell vibration signals

Abstract

The application relates to the technical field of ore dressing automation, and discloses a mill load monitoring system and method based on a cylinder vibration signal, a signal collection part of the system is installed on the outer side of a mill cylinder surface, wireless charging and wireless communication are used to realize equipment power supply and data transmission; host computer software is run on an industrial computer to receive vibration signals, save and display related information in real time, meanwhile, the signals are analyzed and the mill load is calculated based on a monitoring method. The system can continuously and reliably operate in an actual production environment, effectively overcomes the influence of field noise and environmental interference, and can realize real-time monitoring of the mill load.

Description

Technical Field

[0001] The present invention relates to the technical field of mineral processing automation, and in particular to a mill load monitoring system and method based on a cylinder vibration signal. Background Art

[0002] The mill is the primary equipment in the grinding process. During operation, the motor drives the mill cylinder to rotate continuously. The constant impact and compression between the steel balls, ore, and mill liner within the cylinder reduces the size of the ore, achieving ore grinding. Mill load, referring to the total loading of ore, water, and steel balls within the mill, is a crucial parameter in mill operation. Excessive mill load (overload) will result in inadequate grinding of the ore, easily causing internal blockages and even production stoppages. Excessive mill load (underload) will cause the ball mill to consume excessive and unnecessary electricity and steel, accelerating equipment damage. Therefore, maintaining the mill under optimal load conditions is crucial for improving energy efficiency and production profitability in a mineral processing plant.

[0003] However, because the material inside the mill rotates continuously within a closed space, it is impossible to directly detect the internal state of the mill and calculate the mill load. Existing mill load monitoring methods generally use indirect measurement and calculation methods, including monitoring the ore input, power, and sound (grinding noise) when the mill is operating stably. Although these variables can provide certain information, the mapping relationship between them and the mill load is indirect and relatively complex, and the long-term stability and accuracy of the monitoring system are difficult to guarantee. In addition, because the grinding process is affected by multiple factors such as ore properties, steel balls, and liners, and interacts with subsequent processing links, it is very difficult to establish a long-term and effective mechanism model to accurately determine the relationship between each variable and the mill load. In summary, in actual production, it is of great significance to use equipment that can collect signals reflecting the internal state of the mill, establish data-driven models based on the collected signals, judge the mill load situation, and then optimize production operations and processes.

[0004] Existing mill load monitoring systems typically collect the sound produced by the mill during operation, known as grinding noise. Using the characteristics of this noise, they develop data-driven or mechanism-based load identification models to assess the mill's internal conditions. While these grinding noise-based load monitoring systems have been implemented in numerous mineral processing plants and have achieved considerable economic benefits, they suffer from the following limitations: 1. Environmental interference: Microphone-based sound acquisition is highly susceptible to industrial noise. The real-world noisy industrial environment includes noise from the operation of other large equipment, the movements of workers, and production operations such as adding steel balls, all of which can be captured by microphones. The presence of these noises often results in significant interference in the grinding noise signal. This is especially true when multiple mills are operating simultaneously, where their individual grinding noises interfere with each other. These grinding noise signals, with similar frequencies, are difficult to distinguish during signal processing. This noise significantly impacts the accuracy and reliability of data-driven models. 2. Signal interpretation: The grinding noise signal can only serve as an indirect basis for assessing mill load. Although there is an empirical rule regarding the amplitude and frequency of the sound when the charge level inside the mill changes: "When the mill load is high, the grinding sound is low and dull, and when the load is low, the grinding sound is loud and clear," this rule cannot directly and clearly characterize the load size inside the mill. Moreover, this type of explanation based on sound characteristics is not always applicable to all production environments. Therefore, in different mineral processing plants, the load identification system based on grinding sound may not achieve consistent high efficiency and accuracy. Therefore, the existing grinding sound-based mill load monitoring system has the problem of not being able to effectively monitor the load changes inside the mill and is not very applicable in the actual grinding process. Summary of the Invention

[0005] The present invention provides a mill load monitoring system and method based on cylinder vibration signals, so as to solve the problems in the prior art of being unable to effectively monitor the load changes inside the mill and having low applicability.

[0006] In order to achieve the above object, the present invention is implemented through the following technical solutions:

[0007] In a first aspect, the present invention provides a mill load monitoring system based on a cylinder vibration signal, comprising: a signal transmitter, a signal receiver, a signal processing terminal, a wireless charging power module, and an active spatial positioning module, wherein the signal transmitter is disposed within a housing, and the housing is disposed on the surface of the mill cylinder and rotates synchronously with the cylinder during mill operation;

[0008] The signal transmitting end is used to collect the vibration signal of the surface of the mill cylinder;

[0009] The signal receiving end is used to receive the vibration signal and send the vibration signal to the signal processing end;

[0010] The wireless charging power supply module is used to supply power to the signal transmitting end, the signal receiving end and the active spatial positioning module;

[0011] The active spatial positioning module is used to calculate the phase of the mill cylinder in space in real time and send the phase to the signal processing end;

[0012] The signal processing end is used to calculate the mill load based on the vibration signal and the phase.

[0013] Optionally, the signal transmitting end includes a vibration sensor, a signal conditioning circuit, an ADC converter, an MCU, a wireless communication module and a first antenna;

[0014] The vibration sensor is arranged on the surface of the mill cylinder and rotates synchronously with the cylinder during the operation of the mill, and is used to convert the vibration of the mill cylinder surface into an analog electrical signal; the signal conditioning circuit is used to pre-process the analog electrical signal and then send it to a channel of the ADC converter; the ADC converter is used to sample the analog electrical signal input to the channel at a set sampling frequency to obtain a digital vibration signal; the MCU is used to drive the ADC converter to perform analog-to-digital conversion, and to communicate bidirectionally with the ADC converter through a high-speed bus to read the sampled digital vibration signal; the wireless communication module is used to perform bidirectional communication with the MCU through a high-speed bus, and to transmit the read information to the signal receiving end through the first antenna through a set frequency band.

[0015] Optionally, the signal transmitting end further includes a DCDC step-down module, and the DCDC step-down module is respectively connected to the signal conditioning circuit, the ADC converter, the MCU, and the wireless communication module to provide power.

[0016] Optionally, the active spatial positioning module includes a photoelectric sensor and a reflector, the photoelectric sensor is disposed in the housing, and the reflector is installed at a position within a set range from the mill cylinder; the photoelectric sensor is used to emit infrared light outward, and when the infrared light is not reflected by any object, the photoelectric sensor outputs a low level; when the infrared light is reflected by the reflector and received by the photoelectric sensor, the photoelectric sensor outputs a high level; wherein the time between two high levels is regarded as a complete circle; the phase is calculated by the ratio of the time difference between the current moment and the previous triggering of the photoelectric sensor and the time interval between two adjacent triggerings;

[0017] The phase calculation process satisfies the following relationship:

[0018] Phase=(Time Current -Trigger1) / (Trigger2-Trigger1)·360;

[0019] In the formula, Trigger2-Trigger1 represents the time interval between two adjacent high levels output by the photoelectric sensor, Time Current Indicates the current moment.

[0020] Optionally, the wireless charging power supply module includes a switching power supply, a wireless charging transmitter host, a wireless charging transmitter coil, a wireless charging receiver host, and a wireless charging receiver coil; the switching power supply, wireless charging transmitter host, and wireless charging transmitter coil are arranged at a position within a set range from the mill cylinder, and the switching power supply is connected to the wireless charging transmitter host for power supply; the wireless charging transmitter host is used to convert the input direct current into high-frequency alternating current, and transmit the high-frequency alternating current to the wireless charging transmitter coil; the wireless charging receiver coil identifies the wireless charging transmitter coil through magnetic field coupling and receives electrical energy, and converts the high-frequency alternating current back into direct current through the wireless charging receiver host; and the output end of the wireless charging receiver host is connected in parallel to the input end of the DCDC step-down module.

[0021] Optionally, the signal receiving end includes a second antenna and a wireless base station, the second antenna is communicatively connected to the first antenna, and the second antenna receives the vibration signal from the first antenna and transmits it to the wireless base station.

[0022] Optionally, the signal processing end includes an industrial computer, which is connected to the wireless base station. The industrial computer is used to read the vibration signal transmitted by the wireless base station, display the waveform of the vibration signal on the upper computer interface in real time, and display related parameters. The industrial computer is also used to identify the mill load using a load identification model. The input of the load identification model is the vibration signal and the phase position, and the output of the identification model is the mill load.

[0023] In a second aspect, the present application further provides a method for monitoring a mill load based on a cylinder vibration signal, which is applied to the mill load monitoring system based on a cylinder vibration signal described in the first aspect, and the method comprises:

[0024] The signal transmitter is used to collect the vibration signal on the surface of the mill cylinder;

[0025] The signal receiving end is used to receive the vibration signal and send the vibration signal to the signal processing end;

[0026] A wireless charging power module is used to power the signal transmitting end, the signal receiving end and the active spatial positioning module;

[0027] The active spatial positioning module is used to calculate the phase of the mill barrel in space in real time and send the phase to the signal processing end;

[0028] calculating the mill load based on the vibration signal and the phase at a signal processing end;

[0029] Optionally, the calculating the mill load based on the vibration signal and the phase at the signal processing end comprises:

[0030] Step 1: dividing the vibration signal according to the phase, truncating the data in a single signal file in the form of a whole circle in time according to the preset phase starting point and ending point;

[0031] Step 2: pre-processing the vibration signal divided by the circle to obtain a signal meeting the basic requirements of the input of the load identification model;

[0032] The pre-processing comprises: first subtracting the amplitude mean value of the vibration signal in the time domain to remove the trend item, then performing maximum absolute value normalization processing on the waveform amplitude to express the vibration signal in the form of relative intensity, and finally performing pre-emphasis processing on the high-frequency band of the vibration signal;

[0033] Step 3: converting the pre-processed vibration signal from the time domain waveform into a time-frequency spectrogram and inputting the time-frequency spectrogram into the load identification model of the mill to obtain the mill load output by the load identification model;

[0034] The time-frequency spectrogram input into the model is a time-frequency spectrogram group formed by superimposing three layers of time-frequency spectrograms, wherein the frequency range of the first layer of time-frequency spectrogram only covers the key frequency band divided based on prior knowledge, the frequency range of the second layer is the remaining frequency band except the key frequency band, and the third layer is a time-frequency representation converted to a logarithmic scale; by adjusting the frame length and frame shift of the time-frequency spectrogram, it is ensured that each layer of time-frequency spectrogram has the same dimension in time and frequency; the three layers of time-frequency spectrograms have consistent sizes and are independent of each other, and are regarded as three different sound channels.

[0035] Beneficial effects:

[0036] The mill load monitoring system based on the cylinder vibration signal provided by the application has the signal acquisition part installed on the outer side of the mill cylinder surface, realizes device power supply and data transmission in a wireless charging and wireless communication mode, runs the upper computer software on the industrial computer to receive the vibration signal, saves and displays relevant information in real time, simultaneously identifies the mill load based on the monitoring method, and monitors the mill load in real time. The system can continuously and reliably operate in the actual production environment, and effectively overcomes the influence of field noise and environmental interference.

[0037] In a further technical solution, the identification model based on the deep learning neural network can accurately extract the internal characteristics of the mill cylinder vibration signal under different load conditions, and realize real-time monitoring of the mill load through analysis of the internal characteristics of the vibration signal. BRIEF DESCRIPTION OF DRAWINGS

[0038] Figure 1This is a module diagram of a mill load monitoring system based on cylinder vibration signals according to a preferred embodiment of the present invention;

[0039] Figure 2 This is a waveform display diagram of the host computer in the preferred embodiment of the present invention;

[0040] Figure 3 A block diagram of a recognition model according to a preferred embodiment of the present invention;

[0041] Figure 4 This is a block diagram of a denoising feature fusion residual submodule based on spatial attention in a preferred embodiment of the present invention;

[0042] Figure 5 4 is a structural diagram of a spatial attention-based denoising feature fusion unit (SAD Unit) according to a preferred embodiment of the present invention;

[0043] Figure 6 The mill load state classification confusion matrix of the preferred embodiment of the present invention;

[0044] Figure 7 A physical on-site diagram of an example of a preferred embodiment of the present invention;

[0045] Figure 8 This is a flow chart of a method for monitoring mill load based on cylinder vibration signals according to a preferred embodiment of the present invention. DETAILED DESCRIPTION

[0046] The following is a clear and complete description of the technical solutions of the present invention. It should be understood that the embodiments described are only a portion of the embodiments of the present invention, not all of them. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without inventive effort are intended to fall within the scope of protection of the present invention.

[0047] Unless otherwise defined, the technical or scientific terms used in the present invention shall have the usual meanings understood by persons of ordinary skill in the field to which the present invention belongs. The words "first", "second" and similar terms used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. Similarly, words such as "one" or "a" do not indicate a quantity limitation, but rather indicate the existence of at least one. Words such as "connected" or "connected" are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the object being described changes, the relative positional relationship also changes accordingly.

[0048] See Figure 1The present application provides a mill load monitoring system based on a cylinder vibration signal, comprising: a signal transmitter, a signal receiver, a signal processing terminal, a wireless charging power module, and an active spatial positioning module. The signal transmitter is disposed in a housing, and the housing is disposed on the surface of the mill cylinder and rotates synchronously with the cylinder during operation of the mill.

[0049] The signal transmitting end is used to collect the vibration signal of the surface of the mill cylinder;

[0050] The signal receiving end is used to receive the vibration signal and send the vibration signal to the signal processing end;

[0051] The wireless charging power supply module is used to supply power to the signal transmitting end, the signal receiving end and the active spatial positioning module;

[0052] The active spatial positioning module is used to calculate the phase of the mill cylinder in space in real time and send the phase to the signal processing end;

[0053] The signal processing end is used to calculate the mill load based on the vibration signal and the phase.

[0054] It is worth pointing out that the mill barrel vibration signal has obvious advantages over the grinding sound signal. First, the barrel vibration signal will not be interfered with by the operation of other equipment on site, and the barrel vibration signal of each mill is only related to the operating status of the mill, thereby ensuring the relevance and accuracy of the signal. Secondly, the vibration of the mill barrel can directly reflect the changes in its internal load, and any change in load will cause a corresponding change in the vibration characteristics. This consistent correlation shows stability between different mills and mineral processing plants, and improves the universality and reliability of the signal. Therefore, in this application, the signal transmitting end is arranged in the shell, and the shell is arranged on the surface of the mill barrel and rotates synchronously with the barrel during the operation of the mill; it can directly reflect the changes in its internal load, thereby improving the universality and reliability of the signal.

[0055] It is further worth pointing out that the mill may continue to operate for several months after startup, and its surface vibrates violently and is in a continuous rotation state for a long time. In addition, the industrial site environment is complex and the conditions are harsh. There is currently a lack of equipment that can stably collect the vibration signals of the mill barrel in the industrial site for a long time. At the same time, the vibration signal required for analysis has certain requirements for the sampling frequency. The LoRa transmission commonly used in industrial sites cannot meet the signal transmission rate at the required sampling frequency. Therefore, in this application, by providing a wireless charging power module for power supply and an antenna for signal transmission, the signal transmission can be quickly realized under self-powered conditions, and continuous monitoring can be achieved.

[0056] In summary, the mill load monitoring system based on cylinder vibration signals has its data acquisition component installed on the outside of the mill cylinder, using wireless charging and wireless communication to power the equipment and transmit data. Host software running on an industrial computer with a monitoring method loaded receives vibration signals, stores and displays relevant information in real time, and uses this monitoring method to identify mill load, enabling real-time monitoring of mill load conditions. This system operates reliably in a continuous production environment and effectively overcomes the effects of on-site noise and environmental interference.

[0057] Optionally, the signal transmitting end includes a vibration sensor, a signal conditioning circuit, an ADC converter, an MCU, a wireless communication module and a first antenna; the vibration sensor is arranged on the surface of the mill cylinder and rotates synchronously with the cylinder during the operation of the mill, and is used to convert the vibration of the mill cylinder surface into an analog electrical signal; the signal conditioning circuit is used to pre-process the analog electrical signal and then send it to a channel of the ADC converter; the ADC converter is used to sample the analog electrical signal input to the channel at a set sampling frequency to obtain a digital vibration signal; the MCU is used to drive the ADC converter to perform analog-to-digital conversion, and to communicate bidirectionally with the ADC converter through a high-speed bus to read the sampled digital vibration signal; the wireless communication module is used to communicate bidirectionally with the MCU through a high-speed bus, and to transmit the read information to the signal receiving end through the first antenna through a set frequency band.

[0058] Furthermore, the signal transmitter also includes a DC-DC step-down module, which is connected to the signal conditioning circuit, ADC converter, MCU, and wireless communication module to provide power. The DC-DC step-down module further reduces the voltage through its own LDO to meet the input voltage requirements of different chip pins.

[0059] Optionally, the active spatial positioning module includes a photoelectric sensor and a reflector, the photoelectric sensor is arranged in the shell, and the reflector is installed at a position within a set range from the grinding mill cylinder; the photoelectric sensor is used to emit infrared light outward, and when the infrared light is not reflected by any object, the photoelectric sensor outputs a low level; when the infrared light is reflected by the reflector and received by the photoelectric sensor, the photoelectric sensor outputs a high level; wherein the time between two high levels is regarded as a complete circle; the phase is calculated by the time difference between the current moment and the previous triggering of the photoelectric sensor, and the ratio of the time interval between two adjacent triggerings.

[0060] The phase calculation process satisfies the following relationship:

[0061] Phase=(Time Current-Trigger1) / (Trigger2-Trigger1)·360;

[0062] In the formula, Trigger2-Trigger1 represents the time interval between two adjacent high levels output by the photoelectric sensor, Time Current Indicates the current moment.

[0063] Optionally, the wireless charging power supply module includes a switching power supply, a wireless charging transmitter host, a wireless charging transmitter coil, a wireless charging receiver host, and a wireless charging receiver coil; the switching power supply, wireless charging transmitter host, and wireless charging transmitter coil are arranged at a position within a set range from the mill cylinder, and the switching power supply is connected to the wireless charging transmitter host for power supply; the wireless charging transmitter host is used to convert the input direct current into high-frequency alternating current, and transmit the high-frequency alternating current to the wireless charging transmitter coil; the wireless charging receiver coil identifies the wireless charging transmitter coil through magnetic field coupling and receives electrical energy, and converts the high-frequency alternating current back into direct current through the wireless charging receiver host; and the output end of the wireless charging receiver host is connected in parallel to the input end of the DCDC step-down module.

[0064] In a feasible embodiment, the shell is installed in the middle of the ball mill cylinder, the wireless charging receiving coil is wound into a circle and installed under the shell top cover, and the wireless charging transmitting coil is installed at a set distance from the top of the shell.

[0065] Optionally, the signal receiving end includes a second antenna and a wireless base station, the second antenna is communicatively connected to the first antenna, and the second antenna receives the vibration signal from the first antenna and transmits it to the wireless base station.

[0066] Optionally, the signal processing end includes an industrial computer, which is connected to the wireless base station. The industrial computer is used to read the vibration signal transmitted by the wireless base station, display the waveform of the vibration signal on the upper computer interface in real time, and display related parameters. The industrial computer is also used to identify the mill load using a load identification model. The input of the load identification model is the vibration signal and the phase position, and the output of the identification model is the mill load.

[0067] Specifically, in one installation example, all components of the signal transmitter, including the wireless charging receiver host of the wireless charging power module, the wireless charging receiving coil, and the photoelectric sensor of the active spatial positioning module, are installed within the collection system housing. The housing is mounted on the surface of the mill barrel and secured by bolts on the barrel surface. During mill operation, it rotates synchronously with the barrel. The wireless base station of the signal receiver, the switching power supply of the wireless charging power module, the wireless charging transmitter host, the wireless charging transmitting coil, and the reflector of the active spatial positioning module are installed near the mill. The wireless charging transmitting coil and the reflector are mounted at an appropriate distance from the mill barrel via a mounting bracket, and the remaining equipment is installed at a suitable location on site. Specifically, the wireless charging transmitting coil is fixed to the edge of the mill barrel, 5-15 cm from the mill wall, and horizontally aligned with the housing mounted on the mill. During mill operation, the housing rotates synchronously with the mill barrel. Whenever it rotates within the effective range of the wireless charging transmitting coil, the transmitting coil and the receiving coil generate magnetic field coupling, and the receiving coil receives the electrical energy transmitted by the transmitting coil.

[0068] Specifically, the working process of the mill load monitoring system for cylinder vibration signals provided by this application is as follows:

[0069] The vibration sensor converts the vibration on the surface of the mill cylinder into an electrical signal (analog quantity), and the generated analog signal is transmitted to the signal conditioning circuit via a cable. The circuit amplifies, filters, and processes the electrical signal. The processed analog vibration signal is transmitted to a channel of the ADC converter via a cable. The ADC converter samples the analog signal input to the channel at a certain frequency to obtain a vibration signal in digital form. The sampling frequency can be adjusted according to the different actual conditions of different sites, and the minimum should not be lower than the lower limit frequency required by the subsequent vibration signal analysis algorithm, and the maximum should not be higher than the upper limit frequency that the wireless module can achieve for stable transmission under field conditions. In this way, the system transmission can be guaranteed to be stable while meeting the data quality required for algorithm analysis.

[0070] Specifically, the normal communication rate (unit: byte / s) is: the sum of the single data length L1 (unit: byte) and the additional overhead L2 (unit: byte) per byte of data brought by the communication protocol, multiplied by the sampling frequency (unit: Hz).

[0071] Speed ​​= (L1 + L2) SampleRate

[0072] The MCU drives the ADC converter to perform analog-to-digital conversion, and performs two-way communication with the ADC converter through a high-speed bus to read the sampled digital vibration signal; the wireless module performs two-way communication with the MCU through a high-speed bus, and transmits it wirelessly to the wireless base station through the 2.4GHz frequency band; the host computer software deployed on the industrial computer reads the vibration signal transmitted to the industrial computer by the wireless base station, such as Figure 2 As shown, the signal waveform is displayed on the host computer interface in real time, and relevant parameters such as communication quality are displayed to help operators understand the current working status of the system more intuitively. Communication quality is defined as the number of bytes actually received per unit time Num RealReceved The number of bytes Num that should be received per unit time Total Ratio:

[0073] Quality=Num RealReceved / Num Total

[0074] At the same time, the host computer software will periodically save the data to the industrial computer, and provide data for subsequent analysis algorithms.

[0075] Specifically, the working process of the mill load monitoring method of the mill load monitoring system based on the cylinder vibration signal is as follows:

[0076] Step 1: Split the vibration signal according to the phase. According to the preset phase start and end points, the data in a single signal file is truncated in the form of a full circle in time.

[0077] Step 2: Preprocess the vibration signal after segmentation by circle to obtain a signal that meets the basic requirements of the load identification model input;

[0078] The preprocessing includes: first, subtracting the amplitude mean of the vibration signal in the time domain to remove the trend term, then performing maximum absolute value normalization processing on the waveform amplitude to express the vibration signal in relative intensity, and finally performing pre-emphasis processing on the high frequency band of the vibration signal;

[0079] Step 3: Convert the pre-processed vibration signal from a time domain waveform into a time-frequency spectrum and input the spectrum into a load identification model of the mill to obtain the mill load output by the load identification model;

[0080] The time-frequency spectrum of the input model is a time-frequency spectrum group formed by superimposing three layers of time-frequency spectrum, wherein the frequency range of the first layer of time-frequency spectrum only covers the key frequency bands divided based on prior knowledge, the frequency range of the second layer is the remaining frequency bands except the key frequency bands, and the third layer is the time-frequency representation converted to a logarithmic scale; by adjusting the frame length and frame shift of the time-frequency spectrum, it is ensured that each layer of the time-frequency spectrum has the same dimensions in time and frequency; the three layers of time-frequency spectrum are of the same size and independent of each other, and are regarded as three different channels. Since the three channels have the same amount of information in the input model, and the key frequency bands only account for a small part of all the frequency bands, this multi-channel design significantly improves the relative importance of the key frequency bands to the model decision, and at the same time enhances the low-frequency resolution by logarithmic scale, so that the input of the model contains more information; the load identification model based on neural network can effectively extract and fuse the features in the above input information, so as to obtain more accurate load identification results. Figure 6 As shown in the figure, the model finally divides the mill load into five different conditions according to the same standards as the industrial site: normal load, full grinding, empty grinding, partially full grinding, and partially empty grinding.

[0081] Specifically, the load identification model of the mill in step 3 includes the following: Figure 3 The network shown in Figure 1 consists of an input convolutional module, three spatial attention-based denoising feature fusion residual modules, and an output fully connected module. Specifically, the input convolutional module includes a convolutional layer, a batch normalization (BN) layer, a ReLU activation function, and a pooling layer. The convolution kernel size of the convolutional layer is set to 5 to give the network a large receptive field at the beginning. The pooling layer performs a max pooling operation.

[0082] Specifically, the spatial attention-based denoising feature fusion residual module (SAD-Res Block) includes two residual submodules and a maximum pooling layer. The residual submodule includes a first convolutional layer, a BN layer, a ReLU activation function, and a second convolutional layer, a BN layer, and a spatial attention-based denoising feature fusion (SAD) unit connected in sequence. The output of the SAD unit is added to the input value directly sent from the submodule input by the residual connection structure, and then passes through a ReLU activation function to obtain the final output of the spatial attention-based denoising feature fusion residual module. Among them, the residual submodule is Figure 4 , SAD unit such as Figure 5 shown.

[0083] Specifically, the SAD unit first performs global maximum pooling and global average pooling on the input features according to the image channel dimension, and obtains a set of 2×H×W pooling features [F MaxPool ,F AvgPool]. This set of pooled features passes through two convolutional layers and Sigmoid to obtain a set of spatial attention α. ​​Separate the α channel dimension into attention A of size 1×H×W MaxPool and A AvgPool , and multiply them with the corresponding pooling features. The two sets of products are averaged to obtain a set of mixed attention A Mix . Mix The channel dimension of is expanded to make it have the same size as the original feature map, and the expanded set of parameters is used as the denoising soft threshold of the original feature map. The thresholding function is shown as follows:

[0084]

[0085] Where x is the input feature, y is the output feature, and τ is the threshold, which is a positive parameter. The input feature is thresholded through the thresholding function, and the output feature map is the feature after noise reduction.

[0086] Specifically, the output fully connected module includes a fully connected layer and a softmax classifier, which is used to convert the extracted features into the final mill load classification result.

[0087] The following is a complete example to illustrate the mill load monitoring system for the cylinder vibration signal:

[0088] See also Figure 1 In this embodiment, the system consists of four components: a signal transmitter, a signal receiver, a wireless charging power module, and an active spatial positioning module. The signal transmitter includes a vibration sensor, signal conditioning circuitry, an ADC converter, an MCU, a wireless communication module, an antenna, and a 5V DC / DC step-down module. The signal receiver includes an antenna, a wireless base station, an industrial computer, and host software deployed on the industrial computer. The wireless charging power module includes a wireless charging transmitter consisting of a 48V switching power supply, a wireless charging transmitter host, and a wireless charging transmitting coil, and a wireless charging receiver host and a wireless charging receiving coil. The active spatial positioning module includes a photoelectric sensor and a reflector. The signal transmitter, wireless charging receiver, and photoelectric sensor are installed in the system housing, which is fixed to the wall of the mill barrel and rotates synchronously with the mill during operation.

[0089] Specifically, for the actual implementation case of a semi-autogenous mill in a mineral processing plant, the semi-autogenous mill has a diameter of 7m, a length of 3.5m, and a speed of about 12rpm during operation. Figure 7As shown in the figure, during system operation, the housing rotates synchronously with the mill barrel, measuring vibrations along the circumference of the mill barrel wall. The active spatial positioning module simultaneously determines the time each time the housing passes the reflector position, thereby dividing the corresponding phase of the vibration signal. Upon system startup, the wireless module establishes a connection with the wireless base station and begins transmitting the signals collected by the vibration sensor. At this point, the host software on the industrial computer receives the signals from the system, displaying the real-time waveform, and saving the data to the computer's disk. Field operation has demonstrated that the system can stably collect the barrel vibration signals of industrial semi-autogenous mills over a long period of time, providing relatively accurate load monitoring.

[0090] As an alternative embodiment, in another example, specifically, a ball mill in a mineral processing plant has a diameter of 5 meters, a length of 8 meters, and an operating speed of approximately 14 rpm. During operation, the system measures the vibration of the mill's cylinder along its circumference and determines its phase using the active spatial positioning module. The remaining configuration is similar to that of Example 1.

[0091] like Figure 8 As shown, the present application also provides a mill load monitoring method based on a cylinder vibration signal, which is applied to the above-mentioned mill load monitoring system based on a cylinder vibration signal, and the method includes:

[0092] The signal transmitter is used to collect the vibration signal on the surface of the mill cylinder;

[0093] The signal receiving end is used to receive the vibration signal and send the vibration signal to the signal processing end;

[0094] A wireless charging power module is used to power the signal transmitting end, the signal receiving end and the active spatial positioning module;

[0095] The active spatial positioning module is used to calculate the phase of the mill barrel in space in real time and send the phase to the signal processing end;

[0096] A signal processing end is used to calculate the mill load based on the vibration signal and the phase.

[0097] Optionally, the signal processing end calculates the mill load based on the vibration signal and the phase, including:

[0098] Step 1: Split the vibration signal according to the phase. According to the preset phase start and end points, the data in a single signal file is truncated in the form of a full circle in time.

[0099] Step 2: Preprocess the vibration signal after segmentation by circle to obtain a signal that meets the basic requirements of the load identification model input;

[0100] The preprocessing includes: first, subtracting the amplitude mean of the vibration signal in the time domain to remove the trend term, then performing maximum absolute value normalization processing on the waveform amplitude to express the vibration signal in relative intensity, and finally performing pre-emphasis processing on the high frequency band of the vibration signal;

[0101] Step 3: Convert the pre-processed vibration signal from a time domain waveform into a time-frequency spectrum and input the spectrum into a load identification model of the mill to obtain the mill load output by the load identification model;

[0102] The time-frequency spectrum of the input model is a time-frequency spectrum group composed of three layers of time-frequency spectrum superimposed on each other, wherein the frequency range of the first layer of time-frequency spectrum only covers the key frequency bands divided based on prior knowledge, the frequency range of the second layer covers the remaining frequency bands except the key frequency bands, and the third layer is a time-frequency representation converted to a logarithmic scale; by adjusting the frame length and frame shift of the time-frequency spectrum, it is ensured that the dimensions of each layer of the time-frequency spectrum are the same in time and frequency; the three layers of time-frequency spectrum are consistent in size and independent of each other, and are regarded as three different channels.

[0103] The mill load monitoring based on the cylinder vibration signal can implement various embodiments of the above-mentioned system and achieve the same beneficial effects, which will not be described in detail here.

[0104] The above describes in detail the preferred embodiments of the present invention. It should be understood that those skilled in the art can make numerous modifications and variations based on the concepts of the present invention without inventive effort. Therefore, any technical solutions that can be derived by those skilled in the art through logical analysis, reasoning, or limited experimentation based on the concepts of the present invention and the prior art should be within the scope of protection defined by the claims.

Claims

1. A mill load monitoring system based on cylinder vibration signal, characterized in that: include: A signal transmitter, a signal receiver, a signal processing terminal, a wireless charging power module, and an active spatial positioning module. The signal transmitter is disposed in a housing that is fixed to the surface of the mill cylinder and rotates synchronously with the cylinder during operation of the mill. The signal transmitting end is used to collect the vibration signal of the surface of the mill cylinder; The signal receiving end is used to receive the vibration signal and send the vibration signal to the signal processing end; The wireless charging power supply module is used to supply power to the signal transmitting end, the signal receiving end and the active spatial positioning module; The active spatial positioning module is used to calculate the phase of the mill cylinder in space in real time and send the phase to the signal processing end; The signal processing end is used to calculate the mill load based on the vibration signal and the phase; The active spatial positioning module includes a photoelectric sensor and a reflector. The photoelectric sensor is disposed within the housing, and the reflector is installed at a position within a set range from the mill cylinder. The photoelectric sensor is configured to emit infrared light. When the infrared light is not reflected by any object, the photoelectric sensor outputs a low level. When the infrared light is reflected by the reflector and received by the photoelectric sensor, the photoelectric sensor outputs a high level. The time between two high levels is considered a complete cycle. The phase is calculated by the ratio of the time difference between the current moment and the last time the photoelectric sensor was triggered to the time interval between the two adjacent triggers. The phase calculation process satisfies the following relationship: ； Where, Indicates the time interval between two adjacent high levels output by the photoelectric sensor. Current Indicates the current moment; The signal processing end calculates the mill load based on the vibration signal and the phase, including: Step 1: Divide the vibration signal according to the phase. According to the preset phase start and end points, the data in a single signal file is truncated in the form of a full circle in time. Step 2: Preprocess the vibration signal after segmentation by circle to obtain a signal that meets the basic requirements of the load identification model input; The preprocessing includes: first, subtracting the amplitude mean of the vibration signal in the time domain to remove the trend term, then performing maximum absolute value normalization processing on the waveform amplitude to express the vibration signal in relative intensity, and finally performing pre-emphasis processing on the high frequency band of the vibration signal; Step 3: Convert the pre-processed vibration signal from a time domain waveform into a time-frequency spectrum and input the spectrum into a load identification model of the mill to obtain the mill load output by the load identification model; The input spectrogram is a spectrogram group composed of three layers of superimposed spectrograms. The frequency range of the first layer of spectrograms only covers the key frequency bands divided based on empirical knowledge, the frequency range of the second layer covers the remaining frequency bands except the key frequency bands, and the third layer is a time-frequency representation converted to a logarithmic scale. By adjusting the frame length and frame shift of the spectrogram, it is ensured that the dimensions of each layer of the spectrogram are the same in time and frequency. The three layers of spectrograms are of the same size and independent of each other, and are regarded as three different channels.

2. The mill load monitoring system based on cylinder vibration signal according to claim 1 is characterized in that: The signal transmitting end includes a vibration sensor, a signal conditioning circuit, an ADC converter, an MCU, a wireless communication module and a first antenna; The vibration sensor is arranged on the surface of the mill cylinder and rotates synchronously with the cylinder during the operation of the mill, and is used to convert the vibration of the mill cylinder surface into an analog electrical signal; the signal conditioning circuit is used to pre-process the analog electrical signal and then send it to a channel of the ADC converter; the ADC converter is used to sample the analog electrical signal input into the channel at a set sampling frequency to obtain a vibration signal; the MCU is used to drive the ADC converter to perform analog-to-digital conversion, and to communicate bidirectionally with the ADC converter through a high-speed bus to read the sampled vibration signal; the wireless communication module is used to perform bidirectional communication with the MCU through a high-speed bus, and to transmit the read information to the signal receiving end through the first antenna through a set frequency band.

3. The mill load monitoring system based on cylinder vibration signal according to claim 2 is characterized in that: The signal transmitting end further includes a DCDC step-down module, which is respectively connected to the signal conditioning circuit, the ADC converter, the MCU, and the wireless communication module to provide power.

4. The mill load monitoring system based on cylinder vibration signal according to claim 3 is characterized in that: The wireless charging power supply module includes a switching power supply, a wireless charging transmitter host, a wireless charging transmitter coil, a wireless charging receiver host, and a wireless charging receiver coil; the switching power supply, the wireless charging transmitter host, and the wireless charging transmitter coil are arranged at a position within a set range from the mill cylinder, and the switching power supply is connected to the wireless charging transmitter host for power supply; the wireless charging transmitter host is used to convert the input direct current into high-frequency alternating current and transmit the high-frequency alternating current to the wireless charging transmitter coil; the wireless charging receiver coil identifies the wireless charging transmitter coil through magnetic field coupling and receives electrical energy, and converts the high-frequency alternating current back into direct current through the wireless charging receiver host; and the output end of the wireless charging receiver host is connected in parallel to the input end of the DCDC step-down module.

5. The mill load monitoring system based on cylinder vibration signal according to claim 2, characterized in that: The signal receiving end includes a second antenna and a wireless base station. The second antenna is communicatively connected to the first antenna. The second antenna receives the vibration signal from the first antenna and transmits it to the wireless base station.

6. The grinding mill load monitoring system based on cylinder vibration signal according to claim 5, characterized in that: The signal processing end includes an industrial computer, which is connected to the wireless base station. The industrial computer is used to read the vibration signal transmitted by the wireless base station, display the waveform of the vibration signal on the host computer interface in real time, and display related parameters. The industrial computer is also used to identify the mill load using a load identification model. The input of the load identification model is the vibration signal and the phase position, and the output of the identification model is the mill load.

7. A method for monitoring a mill load based on a cylinder vibration signal, applied to a mill load monitoring system based on a cylinder vibration signal according to any one of claims 1 to 6, characterized in that: The method comprises: The signal transmitter is used to collect the vibration signal on the surface of the mill cylinder; Using a signal receiving end to receive the vibration signal, and sending the vibration signal to a signal processing end; The active spatial positioning module is used to calculate the phase of the mill barrel in space in real time and send the phase to the signal processing end; A wireless charging power module is used to power the signal transmitting end, the signal receiving end and the active spatial positioning module; A signal processing end is used to calculate the mill load based on the vibration signal and the phase.

Images