A ball mill analyzer and method of controlling the same

Abstract

The application discloses a kind of mill sound analyzers, receiver, for receiving the audio signal frequency band in the process of ball mill equipment operation;Filter module, amplification module, intelligent analysis module, output module;The processing analysis module includes: energy spectrum calculation module, carries out spectrum analysis to audio signal frequency band, extracts the energy of each frequency band as feature, and constructs feature vector in combination with historical state;Model inference module, using the constructed feature vector is handled to predict the current ball mill capacity.The application receives the audio signal of ball mill equipment in real time, and carries out processing and analysis to it, quickly responds the change of ball mill equipment operating state, realizes accurate prediction to ball mill capacity, optimizes the operating parameter of ball mill equipment by accurately predicting ball mill capacity, improves production efficiency and product quality, the application of intelligent analysis module and model inference module makes the management of ball mill equipment more intelligent and automated, reduces the complexity and cost of manual intervention.

Description

Technical Field

[0001] This invention belongs to the field of ball mill analyzers, and more specifically, relates to a ball mill analyzer and its control method. Background Technology

[0002] For various grinding equipment and ball mills used for ore, due to their working principle, it is difficult to effectively monitor the internal feed rate. This relies entirely on the operator's experience, adjusting the equipment based on the sound and volume emitted by the grinding machine. In actual industrial production, this makes it difficult to effectively improve output quality and operational efficiency. The existing solution involves using a grinding sound analyzer to mimic the operator's ear, monitoring the grinding sound volume, and using a DCS system to control the feed, thereby improving the level of automation.

[0003] Existing grinding sound analyzers cannot dynamically provide feedback on the quality of materials. For example, the grinding time required for ores with different hardness and moisture content varies; at the data acquisition level, the frequency and amplitude of the audio signal differ even with the same feed amount; at the application level, the grinding time required for the same audio feedback will vary, necessitating different control strategies for ores of different qualities.

[0004] This adjustment method results in poor system version consistency; effective ball mill control optimization strategies require designers to have a high level of knowledge in geology, control, and other aspects; for the same production system, it is also difficult to quickly switch parameters and strategies when the ore source or quality changes, resulting in slow response speed, low efficiency, inability to accurately predict ball mill capacity, and low level of intelligence.

[0005] In view of this, the present invention is proposed. Summary of the Invention

[0006] The technical problem to be solved by the present invention is to overcome the shortcomings of the prior art and provide a ball mill analyzer and its control method. By predicting and feeding back the ball mill capacity in real time and dynamically based on the grinding sound signal during the ore ball milling process, and taking into account the different hardness and moisture characteristics of the ore, the production efficiency and system response speed can be improved.

[0007] To solve the above-mentioned technical problems, the basic concept of the technical solution adopted by the present invention is as follows: The present invention provides a grinding sound analyzer, comprising:

[0008] A microphone is used to receive audio signal frequencies during the operation of ball mill equipment.

[0009] The filtering module is used to filter the frequency bands of the acquired audio signal and select the desired audio signal frequency bands.

[0010] The amplification module is used to amplify the frequency bands of the selected audio signal;

[0011] The intelligent analysis module is used to acquire the frequency band of the amplified audio signal, process and analyze the audio signal frequency band, and output predictions.

[0012] The output module is used to convert the output prediction results into actual control signals and then output them.

[0013] The processing and analysis module includes: an energy spectrum calculation module, which performs spectral analysis on the audio signal frequency bands, extracts the energy of each frequency band as features, and constructs a feature vector by combining it with historical states;

[0014] The model inference module uses the constructed feature vectors to predict the current ball mill capacity.

[0015] Furthermore, the energy spectrum calculation module includes: a spectrum analysis unit, used to receive audio signal frequency bands and divide the audio signal into several fixed frequency band intervals;

[0016] The energy spectrum calculation unit is used to calculate the spectral integral value within each frequency band interval;

[0017] The feature extraction unit is used to take the spectral integral value of the energy spectrum calculation module as the feature vector, and combine it with the intermediate state quantity of the previous period. It processes the feature vector using linear transformation and activation function, and constructs the feature vector by combining it with historical states.

[0018] Furthermore, the model inference module includes: a recurrent neural network unit, used to calculate the current cycle state and intermediate state quantities by utilizing the recurrent structure of the recurrent neural network unit and combining the current input and historical input states; the model inference module includes multi-level recurrent neural network units to form a fully connected layer structure;

[0019] The prediction output unit takes the output of the recurrent neural network unit as the input to the fully connected layer structure, and obtains the final prediction value through linear transformation and activation function.

[0020] The present invention also discloses a control method for the ball mill analyzer described above, the control method comprising the following steps:

[0021] The audio signal received by the microphone from the ball mill equipment is acquired, and the audio signal is filtered to remove noise and unnecessary interference.

[0022] The intelligent analysis module processes the frequency bands of the acquired audio signal, extracts the energy of each frequency band as features, and combines historical data to construct features for the grinding progress of ores with different hardness / humidity.

[0023] The model inference module uses a recurrent neural network to extract features from energy spectrum data and calculate predicted values.

[0024] The output module converts the calculated predicted value into an actual control signal and outputs the calculation result as a standard signal.

[0025] Furthermore, the intelligent analysis module also includes the following when processing the acquired audio signal frequency bands:

[0026] The acquired audio signal is converted into a digital signal;

[0027] Perform a Fast Fourier Transform on the converted digital signal to convert the digital signal from the time domain to the frequency domain.

[0028] Furthermore, when extracting energy from each frequency band as features and combining historical states to construct features for the grinding progress of ores with different hardness / humidity, the following steps were also taken:

[0029] The frequency domain signal is divided into several fixed frequency band intervals, and the energy spectrum in each frequency band is calculated.

[0030] The calculated energy spectrum is used as subsequent input. The model inference module uses a recurrent neural network to process the discrete energy spectrum data, extract grinding progress features, and predict the ball mill capacity.

[0031] Furthermore, the model inference module utilizes a recurrent neural network to process the discrete energy spectrum data, extract grinding progress features, and predict the ball mill capacity, specifically including:

[0032] First, Xt is input into the recurrent neural network as the input for subsequent calculations. The recurrent neural network introduces the historical input state Ct-1, processes the feature vector using a predefined weight matrix and activation function, and calculates the current cycle state Ht and intermediate state Ct.

[0033] Furthermore,

[0034] The activation function is Sigmoid; the intermediate state Ct = Sig(W1·Xt) + (W2·Ct-1);

[0035] The Ht=(Sig(W0﹒Xt)☉(Sig(Sig(W1﹒Xt)+(W2﹒Ct-1)));

[0036] Where: Ht is the state of the current period; Sig is the activation function; W0, W1, and W2 are predefined weight matrices; Xt is the input quantity; and Ct-1 is the historical input state.

[0037] Furthermore, the recurrent neural network is composed of multiple recurrent neural networks connected in series.

[0038] Furthermore, the output module converts the calculated predicted value into an actual control signal and outputs the calculation result as a standard signal, specifically including:

[0039] The output module outputs Ht based on the current cycle state of the recurrent neural network, calculates the final output value Out through a fully connected layer, with the value between [0,1], and converts it into a 4-20mA current signal for final output;

[0040] The Out = Sig(Sum(Wk));

[0041] Where Sig is the activation function and Wk is the final weighted vector before the output of the fully connected layer.

[0042] By adopting the above technical solution, the present invention has the following beneficial effects compared with the prior art:

[0043] (1) The grinding sound analyzer of the present invention receives the audio signal of the ball mill in real time, processes and analyzes it, and can quickly respond to changes in the operating status of the ball mill, realize accurate prediction of the ball mill capacity. By accurately predicting the ball mill capacity, the operating parameters of the ball mill, such as the feeding speed and grinding time, can be optimized, thereby improving production efficiency and product quality. The application of the intelligent analysis module and the model reasoning module makes the management of the ball mill more intelligent and automated, reducing the complexity and cost of manual intervention.

[0044] (2) By dividing the detection frequency band into multiple discrete segments and calculating the energy within each segment, the system can capture the spectral characteristics of the input signal in greater detail. This refinement helps identify subtle differences in the ore produced during grinding, improving the system's resolution. Simultaneously, the frequency band processing reduces noise interference to the useful signal across the entire band, enabling more effective extraction of features related to the ore's grinding state. It also allows for more flexible adaptation to different types of ore, thereby improving the system's robustness and versatility.

[0045] (3) By using a recurrent neural network to process time series data, this invention can capture the time dependence and dynamic changes in the data. In ore grinding monitoring, the introduction of historical computation as input can help the RNN network better understand the dynamic changes in the grinding process, thereby improving the accuracy of prediction.

[0046] (4) By connecting multiple recurrent neural networks in series, the system can build a deeper network structure, thereby having stronger feature extraction and pattern recognition capabilities, which helps to mine deeper information in the data and improve the prediction accuracy of the system. Using a fully connected network as the output layer, the features extracted by the recurrent neural network can be flexibly mapped to the required output range. The Sigmoid function, as the activation function, can limit the output value to (0,1), which is convenient for scaling and standardization, and improves the applicability to different scenarios.

[0047] The specific embodiments of the present invention will now be described in further detail with reference to the accompanying drawings. Attached Figure Description

[0048] The accompanying drawings, as part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments and descriptions of the invention are used to explain the invention, but do not constitute an undue limitation of the invention. Obviously, the drawings described below are merely some embodiments, and those skilled in the art can obtain other drawings based on these drawings without creative effort. In the drawings:

[0049] Figure 1 This is a structural block diagram of a grinding sound analyzer according to the present invention;

[0050] Figure 2 This is an overall structural diagram of a grinding sound analyzer according to the present invention;

[0051] Figure 3 These are multiple fixed-frequency band energy spectrum bar charts in this invention;

[0052] Figure 4 This is a design diagram of a single recurrent neural network structure in this invention;

[0053] Figure 5 This is a design diagram of the multi-level recurrent neural network structure in this invention;

[0054] Figure 6 This is a flowchart of a control method for a grinding sound analyzer according to the present invention.

[0055] It should be noted that these accompanying drawings and textual descriptions are not intended to limit the scope of the invention in any way, but rather to illustrate the concept of the invention to those skilled in the art by referring to specific embodiments. Detailed Implementation

[0056] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, embodiments of the present disclosure will be described below with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the present disclosure. In the following detailed description, many specific details are set forth to provide a comprehensive understanding of the embodiments of the present disclosure for ease of explanation. However, it will be apparent that one or more embodiments may be practiced without these specific details. Furthermore, descriptions of well-known structures and technologies are omitted in the following description to avoid unnecessarily obscuring the concepts of the present disclosure.

[0057] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit this disclosure. The terms “comprising,” “including,” etc., as used herein indicate the presence of features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0058] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein are to be interpreted in a manner consistent with the context of this specification, and not in an idealized or overly rigid way.

[0059] When using expressions such as "at least one of A, B, and C," the expression should generally be interpreted in accordance with the meaning commonly understood by a person skilled in the art (e.g., "a system having at least one of A, B, and C" should include, but is not limited to, systems having A alone, having B alone, having C alone, having A and B, having A and C, having B and C, and / or having A, B, and C, etc.). Similarly, when using expressions such as "at least one of A, B, or C," the expression should generally be interpreted in accordance with the meaning commonly understood by a person skilled in the art (e.g., "a system having at least one of A, B, or C" should include, but is not limited to, systems having A alone, having B alone, having C alone, having A and B, having A and C, having B and C, and / or having A, B, and C, etc.).

[0060] The accompanying drawings illustrate several block diagrams and / or flowcharts. It should be understood that some blocks, or combinations thereof, in the block diagrams and / or flowcharts can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that, when executed by the processor, these instructions can create means for implementing the functions / operations described in these block diagrams and / or flowcharts. The technology of this disclosure can be implemented in hardware and / or software (including firmware, microcode, etc.). Alternatively, the technology of this disclosure can take the form of a computer program product stored on a computer-readable storage medium, which is available for use by or in conjunction with an instruction execution system.

[0061] A grinding sound analyzer is a typical audio acquisition and analysis system. It collects the sound emitted by grinding equipment on-site, filters and extracts the effective frequency components, and establishes a mapping relationship between audio intensity and the amount and particle size of material in the grinding equipment.

[0062] like Figures 1 to 2 As shown, the grinding sound analyzer of the present invention includes: a microphone, a filtering module, an amplification module, an intelligent analysis module, and an output module.

[0063] The receiver is used to receive audio signal frequency bands during the operation of the ball mill equipment.

[0064] The filtering module is used to filter the acquired audio signal frequency bands and select the desired audio signal frequency bands.

[0065] The amplification module is used to amplify the selected audio signal frequency bands.

[0066] The intelligent analysis module is used to acquire the frequency band of the amplified audio signal, process and analyze the audio signal frequency band, and output predictions.

[0067] The output module is used to convert the output prediction result into an actual control signal and output it.

[0068] The processing and analysis module includes an energy spectrum calculation module, which performs spectral analysis on the audio signal frequency bands, extracts the energy of each frequency band as features, and constructs feature vectors by combining historical states.

[0069] The model inference module uses the constructed feature vectors to predict the current ball mill capacity.

[0070] This invention first uses a microphone installed near the ball mill to receive audio signals during operation. A filtering module then filters the received audio signals, removing noise and unwanted frequency bands to select those closely related to the ball mill's operating status. An amplification module then amplifies the selected audio signals to improve the signal-to-noise ratio and clarity, providing a high-quality signal source for subsequent analysis and prediction. Next, the energy spectrum calculation module in the intelligent analysis module performs spectral analysis on the amplified audio signals, extracting the energy of each frequency band as features. These extracted features are combined with historical state data to construct a feature vector that comprehensively reflects the current state of the ball mill. Using this feature vector, a model inference module processes and predicts the current ball mill capacity. Finally, the output module converts the prediction results from the model inference into actual control signals, such as control signals to adjust parameters like feed rate and grinding time. These control signals act on the ball mill through corresponding actuators, enabling real-time adjustment and optimization of the equipment's operating status.

[0071] The present invention, through the above-mentioned settings, can quickly respond to changes in the operating status of ball mill equipment and achieve accurate prediction of ball mill capacity. By accurately predicting the ball mill capacity, the operating parameters of the ball mill equipment, such as feed rate and grinding time, can be optimized, thereby improving production efficiency and product quality. The application of intelligent analysis module and model reasoning module makes the management of ball mill equipment more intelligent and automated, reducing the complexity and cost of manual intervention.

[0072] Furthermore, the energy spectrum calculation module includes a spectrum analysis unit, an energy spectrum calculation unit, and a feature extraction unit.

[0073] The spectrum analysis unit is used to receive audio signal frequency bands and divide the audio signal into several fixed frequency band intervals.

[0074] The energy spectrum calculation unit is used to calculate the spectral integral value within each frequency band interval.

[0075] The feature extraction unit is used to take the spectral integral value of the energy spectrum calculation module as the feature vector, and combine it with the intermediate state quantity of the previous period. It processes the feature vector using linear transformation and activation function, and constructs the feature vector by combining it with historical states.

[0076] By setting up the spectrum analysis unit as described above, the audio signal is divided into several fixed frequency band intervals, making the energy spectrum calculation more accurate and detailed, and more accurately reflecting the operating status of the ball mill equipment. The energy spectrum calculation unit realizes the quantitative analysis of the audio signal energy by calculating the spectrum integral value within each frequency band interval. The feature extraction unit combines the intermediate state quantities and historical state information of the previous period, and further processes the feature vector through linear transformation and activation function to improve the expressive power and prediction accuracy of the feature vector.

[0077] Specifically, after receiving the audio signal frequency band, the spectrum analysis unit uses Fourier transform (or other suitable spectrum analysis methods) to convert the audio signal from the time domain to the frequency domain, dividing the frequency domain signal into several fixed frequency band intervals. These intervals can be set according to the characteristics of the ball mill equipment and prediction requirements. Next, the energy spectrum calculation unit integrates the spectrum signal within each frequency band interval to obtain the spectrum integral value (i.e., energy value) within that interval and records the energy value for each frequency band interval, forming preliminary energy spectrum data. Finally, the feature extraction unit uses the spectrum integral value obtained by the energy spectrum calculation unit as part of the input features, while simultaneously acquiring the intermediate state quantities from the previous period. The spectrum integral value and the intermediate state quantities from the previous period are combined to form a preliminary feature vector. Then, a linear transformation (such as weight matrix multiplication) is applied to process the preliminary feature vector. An activation function (such as ReLU, Sigmoid, etc.) is used to perform a nonlinear transformation on the linearly transformed feature vector. Combined with historical state data (such as historical ball mill capacity, equipment operating time, etc.), the nonlinearly transformed feature vector is further processed to construct the final feature vector used for model prediction.

[0078] Furthermore, the model inference module includes a recurrent neural network unit and a prediction output unit.

[0079] The recurrent neural network unit is used to calculate the current cycle state and intermediate state quantities by utilizing the recurrent structure of the recurrent neural network unit and combining the current input and historical input states; the model inference module includes multi-level recurrent neural network units to form a fully connected layer structure.

[0080] The prediction output unit takes the output of the recurrent neural network unit as the input to the fully connected layer structure, and obtains the final prediction value through linear transformation and activation function.

[0081] Recurrent Neural Network (RNN) units, through their built-in recurrent structure, are capable of processing data with time-series characteristics. At each time step, the RNN unit receives the current input features (feature vectors from the energy spectrum calculation module) and the output state of the previous time step, and calculates the output state of the current time step and the new internal state accordingly.

[0082] The RNN units in the model inference module of this application are organized into a multi-level structure. Each level receives the output of the previous level as input and produces its own output, which can capture more complex temporal dependencies and improve the model's predictive ability. Information is transmitted through a fully connected layer structure formed by multiple RNNs connected in series.

[0083] The prediction output unit receives the final output of the RNN unit as input, performs a linear transformation on the input (i.e., multiplies it by a weight matrix and adds a bias term) to adjust the contribution of different features to the prediction result.

[0084] An activation function (such as Sigmoid, Softmax, etc., selected according to the specific needs of the prediction task) is applied to perform a nonlinear transformation on the result after the linear transformation to obtain the final predicted value.

[0085] Specifically, such as Figure 3 As shown, the input signal is first divided into several fixed frequency band intervals through spectrum analysis, and the energy spectrum of the input signal within the frequency band intervals is calculated. The entire frequency band range is up to 5KHz. The spectrum integral calculation value is x0 in the range of 0 to 100Hz, x1 in the range of 100 to 200Hz, ..., and xn in the range of 4.9KHz to 5KHz. x1 to xn are used as input quantities for subsequent calculations. The current ball mill capacity is predicted using a recurrent neural network (RNN) structure. By introducing the historical input state Ct-1, the grinding progress of ores with different hardness / moisture content is extracted.

[0086] Where: Xt=[x0t,x1t…xnt], Ct-1 is the historical input state;

[0087] Ht=(Sig(W0﹒Xt)☉(Sig(Sig(W1﹒Xt)+(W2﹒Ct-1)))

[0088] Where Sig is the activation function Sigmoid;

[0089] The aforementioned intelligent analysis module uses a field-programmable gate array (FPGA) as a processor to provide parallel computing for the recurrent neural network.

[0090] Figure 6 As shown, the present invention also discloses a control method applied to the ball mill analyzer, the control method comprising the following steps:

[0091] S1. Obtain the audio signal from the ball milling equipment received by the microphone, and filter the audio signal to remove noise and unnecessary interference.

[0092] S2, the intelligent analysis module processes the frequency bands of the acquired audio signal, extracts the energy of each frequency band as features, and combines historical states to construct features for the grinding progress of ores with different hardness / humidity.

[0093] S3, the model inference module uses a recurrent neural network to extract features from the energy spectrum data and calculate predicted values.

[0094] S4. The output module converts the calculated predicted value into an actual control signal and outputs the calculation result as a standard signal.

[0095] This application, by directly analyzing the audio signals inside the ball mill, can reflect the state of ore grinding in real time, including changes in key parameters such as hardness and moisture, thereby improving the control precision of the grinding process. Combined with intelligent analysis and model reasoning modules, it can automatically learn and adapt to changes in different ore characteristics, realize intelligent prediction and control of grinding progress, improve production efficiency and product quality, and accurately control the grinding process. This can avoid energy waste and equipment wear caused by over-grinding or under-grinding, and reduce production costs.

[0096] Specifically, the process begins with audio signal acquisition and preprocessing. The acquired audio signal is filtered to remove unnecessary interference such as environmental noise and mechanical vibration, ensuring signal quality. Next, an intelligent analysis module performs spectral analysis on the processed audio signal, dividing it into multiple frequency bands and extracting the energy of each band as feature values. These feature values ​​reflect changes in the physical properties of the ore during grinding, such as hardness and moisture content. Then, combined with historical data, a feature library tailored to different ore characteristics is constructed, providing a foundation for subsequent model inference. A recurrent neural network (RNN) is used to further process the extracted feature values, and the trained model calculates predicted values. Based on the predicted values, the current grinding progress is assessed, and future grinding trends are predicted. The output module converts the predicted values ​​into actual control signals, such as adjusting parameters like the ball mill's rotation speed and grinding time. These control signals are output to the ball mill control system in standard signal form, achieving automated control of the grinding process. Finally, feedback on the control effect is collected, and the model is continuously optimized and adjusted to improve prediction accuracy and control effectiveness.

[0097] Furthermore, the intelligent analysis module also includes the following when processing the acquired audio signal frequency bands:

[0098] The acquired audio signal is converted into a digital signal;

[0099] Perform a Fast Fourier Transform on the converted digital signal to convert the digital signal from the time domain to the frequency domain.

[0100] like Figure 3 As shown, converting a signal from the time domain to the frequency domain using the Fast Fourier Transform (FFT) allows for a more intuitive view of the signal's frequency attributes, such as high-frequency and low-frequency components and their bandwidth. The spectrum clearly displays the signal's distribution in the frequency domain, aiding in the identification and analysis of key signal features. In the frequency domain, FFT can separate noise components from the signal and remove them through filtering, thereby improving the signal-to-noise ratio. This is particularly important in monitoring the ore grinding process, as the sound signals generated during grinding of different ores may exhibit significant differences in the frequency domain.

[0101] Based on FFT, the system can also predict the spectral characteristics of future signals by analyzing historical spectral data, thereby taking corresponding control measures or optimization strategies in advance.

[0102] Specifically, the acquired sound signal undergoes filtering to remove noise and interference, improving signal quality. Subsequently, the signal is amplified for further digital processing; an ADC (Analog-to-Digital Converter) converts the analog sound signal into a digital signal for processing. FFT (Fast Fourier Transform) analysis transforms the digital signal from the time domain to the frequency domain, facilitating subsequent energy spectrum calculations.

[0103] Furthermore, when extracting energy from each frequency band as features and combining historical states to construct features for the grinding progress of ores with different hardness / humidity, the following steps were also taken:

[0104] The frequency domain signal is divided into several fixed frequency band intervals, and the energy spectrum in each frequency band is calculated.

[0105] The calculated energy spectrum is used as subsequent input. The model inference module uses a recurrent neural network to process the discrete energy spectrum data, extract grinding progress features, and predict the ball mill capacity.

[0106] By dividing the frequency domain signal into several fixed frequency bands and calculating the energy spectrum, features related to grinding progress in the audio signal can be captured more accurately, which helps improve the accuracy of the prediction model. Combining historical state data to construct a feature set enables the model to capture dynamic changes during the grinding process, enhancing its adaptability to different ore types and varying grinding conditions, and improving its robustness. Using RNNs to process discrete energy spectrum data in real time allows for real-time monitoring and prediction of grinding progress, thereby optimizing the grinding process.

[0107] Furthermore, the model inference module utilizes a recurrent neural network to process the discrete energy spectrum data, extract grinding progress features, and predict the ball mill capacity, specifically including:

[0108] First, Xt is input into the recurrent neural network as the input for subsequent calculations. The recurrent neural network introduces the historical input state Ct-1, processes the feature vector using a predefined weight matrix and activation function, and calculates the current cycle state Ht and intermediate state Ct.

[0109] like Figures 3 to 4 As shown, specifically, in the frequency domain, the signal is divided into several fixed frequency band intervals, and the energy spectrum within each frequency band, i.e., x0, x1, ..., xn, is calculated. The energy spectrum data reflects the energy distribution of the sound signal in different frequency bands and serves as the input for subsequent model inference. The model inference module mainly utilizes an RNN to process the discrete energy spectrum data. The RNN captures the time dependence in the sequence data by introducing historical input states Ct-1, extracting features related to the grinding progress. The parallel processing capability of the FPGA (Field-Programmable Gate Array) is used to accelerate the RNN's computation process and improve the system response time.

[0110] Furthermore,

[0111] The activation function is Sigmoid; the intermediate state Ct = Sig(W1·Xt) + (W2·Ct-1);

[0112] The Ht=(Sig(W0﹒Xt)☉(Sig(Sig(W1﹒Xt)+(W2﹒Ct-1)));

[0113] Where: Ht is the state of the current period; Sig is the activation function; W0, W1, and W2 are predefined weight matrices; Xt is the input quantity; and Ct-1 is the historical input state.

[0114] As shown in the figure, for example, taking a three-dimensional vector:

[0115] Xt = [0.5470 0.2963 0.7447]

[0116] W0=W1=[0.1890 0.6868 0.1835]T

[0117] W2 = [0.0811 0.9294 0.7757]T

[0118] Ct-1 = [0.3685 0.6256 0.7802]

[0119] W1﹒ Xt=W0﹒ Xt=[0.1034 0.0560 0.1407 0.3757 0.2035 0.5114 0.1004 0.0544 0.1367]

[0122] W2﹒ Ct-1=[0.0299 0.0508 0.0633 0.3425 0.5814 0.7251 0.2858 0.4853 0.6052]

[0125] Sig(W0﹒Xt)=Sig(W0﹒Xt)= [0.5258 0.5140 0.5351 0.5928 0.5507 0.6251 0.5251 0.5136 0.5341]

[0129] (Sig(Sig(W1﹒Xt)+(W2﹒Ct-1)))=[ 0.6355 0.6376 0.6453 0.7181 0.7562 0.7942 0.6923 0.7308 0.7576]

[0133] Ht = [1.1234 1.1195 1.2464]

[0134] Ct = Sig(W1·Xt) + (W2·Ct-1) can be used for subsequent data calculations.

[0135] Ct = [0.5557 0.5647 0.5984] 0.9353 1.1321 1.3503 0.8109 0.9989 1.1393]

[0138] Define the structure of an RNN, including an input layer, hidden layers (there may be multiple layers), and an output layer.

[0139] Forward propagation is performed using predefined weight matrices (such as W0, W1, W2) and activation functions (such as Sigmoid) to obtain the hidden layer state Ht and the final output Out.

[0140] By leveraging the parallel processing capabilities of FPGAs, the computation process of RNNs can be accelerated, thereby improving system response time.

[0141] Furthermore, the recurrent neural network is composed of multiple recurrent neural networks connected in series.

[0142] like Figure 5 As shown, the system can be composed of a multi-level RNN structure. The output is fully connected using Ht, and Out is the final output with a value between [0,1]. If a 4-20mA output is used, the final output current is 4+16*Out(mA).

[0143] For example: Taking the parameters of the fully linked layer:

[0144] Wq = [0.04417 0.01967 0.04561] 0.00133 0.00934 0.01017 0.08972 0.03074 0.09954]

[0147] Then Wk = [0.1629 0.0709 0.1867]

[0148] Final output data: Out = Sig(Sum(Wk)) = 0.6036

[0149] Output current: 13.6mA.

[0150] Where: Wq is the mapping weight used to map from the hidden state (or similar state) of the RNN to the input of the fully connected layer.

[0151] Furthermore, the output module converts the calculated predicted value into an actual control signal and outputs the calculation result as a standard signal, specifically including:

[0152] The output module outputs Ht based on the current cycle state of the recurrent neural network, calculates the final output value Out through a fully connected layer, with the value between [0,1], and converts it into a 4-20mA current signal for final output;

[0153] The Out = Sig(Sum(Wk));

[0154] Where Sig is the activation function and Wk is the final weighted vector before the output of the fully connected layer.

[0155] In the model inference module of this application, the RNN calculates the current hidden layer state Ht and the final output Out based on the input energy spectrum data and historical state information. Out is a value between [0,1], representing the predicted value of the ball mill capacity.

[0156] The output module calculates the final prediction result through a fully connected layer based on the output value of the RNN. This prediction result is then converted into a 4–20 mA current signal for interfacing with the industrial control system.

[0157] The standard signal, based on the Modbus protocol, is used to send current signals to the control system or display device for remote monitoring and adjustment. The final output is calculated using a fully connected network and employs the Sigmoid function as the activation function. The system output can be directly mapped to 4-20mA by the activation function (or the ratio can be output via Modbus).

[0158] This application's entire device, through close collaboration and integration among its modules, achieves a complete process from sound signal acquisition, processing, model inference to output control. High-speed data processing and transmission capabilities enable real-time monitoring and prediction of the ball mill's grinding progress. Simultaneously, precise calculations and conversions ensure the accuracy and reliability of the prediction results.

[0159] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-described technical content to create equivalent embodiments without departing from the scope of the present invention. The implementation schemes in the above embodiments can also be further combined or replaced. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A ball milling analyzer, characterized in that, include: A microphone is used to receive audio signal frequencies during the operation of ball mill equipment. The filtering module is used to filter the frequency bands of the acquired audio signal and select the desired audio signal frequency bands. The amplification module is used to amplify the frequency bands of the selected audio signal; The intelligent analysis module is used to acquire the frequency band of the amplified audio signal, process and analyze the audio signal frequency band, and output predictions. The output module is used to convert the output prediction results into actual control signals and then output them. The intelligent analysis module includes: an energy spectrum calculation module, which performs spectrum analysis on the audio signal frequency bands, extracts the energy of each frequency band as features, and constructs a feature vector by combining historical states; The model inference module uses the constructed feature vectors to predict the current ball mill capacity. The energy spectrum calculation module includes: a spectrum analysis unit, used to receive audio signal frequency bands and divide the audio signal into several fixed frequency band intervals; The energy spectrum calculation unit is used to calculate the spectral integral value within each frequency band interval; The feature extraction unit is used to take the spectral integral value of the energy spectrum calculation module as the feature vector, and combine it with the intermediate state quantity of the previous period. It processes the feature vector using linear transformation and activation function, and constructs the feature vector by combining it with historical states.

2. The ball milling analyzer according to claim 1, characterized in that, The model inference module includes: a recurrent neural network unit, used to calculate the state and intermediate state of the current cycle by utilizing the recurrent structure of the recurrent neural network unit and combining the current input and the intermediate state of the previous cycle; the model inference module includes multi-level recurrent neural network units to form a fully connected layer structure. The prediction output unit takes the output of the recurrent neural network unit as the input to the fully connected layer structure, and obtains the final prediction value through linear transformation and activation function.

3. A control method for a ball mill analyzer as described in any one of claims 1-2, characterized in that, The control method includes the following steps: The audio signal received by the microphone from the ball milling equipment is acquired, and the audio signal is filtered to remove noise and unnecessary interference. The intelligent analysis module processes the frequency bands of the acquired audio signal, extracts the energy of each frequency band as features, and combines historical data to construct features for the grinding progress of ores with different hardness / humidity. The model inference module uses a recurrent neural network to extract features from energy spectrum data and calculate predicted values. The output module converts the calculated predicted value into an actual control signal and outputs the calculation result as a standard signal.

4. The control method for a ball mill analyzer according to claim 3, characterized in that, The intelligent analysis module also includes the following when processing the frequency bands of the acquired audio signals: The acquired audio signal is converted into a digital signal; Perform a Fast Fourier Transform on the converted digital signal to convert the digital signal from the time domain to the frequency domain.

5. The control method for a ball mill analyzer according to claim 4, characterized in that, The process of extracting energy from each frequency band as features and combining historical data to construct features for the grinding progress of ores with different hardness / humidity also includes: The frequency domain signal is divided into several fixed frequency band intervals, and the energy spectrum in each frequency band is calculated. The calculated energy spectrum is used as subsequent input. The model inference module uses a recurrent neural network to process the discrete energy spectrum data, extract grinding progress features, and predict the ball mill capacity.

6. The control method for a ball mill analyzer according to claim 5, characterized in that, The model inference module uses a recurrent neural network to process discrete energy spectrum data, extract grinding progress features, and predict ball mill capacity, specifically including: First, Xt is input into the recurrent neural network as the input for subsequent calculations. The recurrent neural network introduces the historical input state Ct-1, processes the feature vector using a predefined weight matrix and activation function, and calculates the current cycle state Ht and intermediate state Ct.

7. The control method for a ball mill analyzer according to claim 6, characterized in that, The activation function is Sigmoid; the intermediate state Ct = Sig(W1·Xt) + (W2·Ct-1); The given Ht = Sig(W0﹒Xt)☉Sig(Sig(W1﹒Xt)+(W2﹒Ct-1)); Where: Ht is the state of the current period; Sig is the activation function; W0, W1, and W2 are predefined weight matrices; Xt is the input quantity; and Ct-1 is the historical input state.

8. The control method for a ball mill analyzer according to claim 3, characterized in that, The recurrent neural network is composed of multiple recurrent neural networks connected in series.

9. The control method for a ball mill analyzer according to claim 6, characterized in that, The output module converts the calculated predicted values ​​into actual control signals and outputs the calculation results as standard signals, specifically including: The output module outputs Ht based on the current cycle state of the recurrent neural network, calculates the final output value Out through a fully connected layer, with the value between [0,1], and converts it into a 4~20mA current signal for final output; The Out = Sig(Sum(Wk)); Where Sig is the activation function and Wk is the final weighted vector before the output of the fully connected layer.

Images