Calcium carbide solid waste grinding process intelligent control method and system

Abstract

The application provides a calcium carbide solid waste grinding process intelligent control method and system, the method comprises the following steps: obtaining calcium carbide solid waste raw material characteristic data; based on the raw material characteristic data, the initial proportioning scheme and the initial grinding process parameters of solid waste mixed grinding are calculated through a preset proportioning model; in the grinding process according to the initial proportioning scheme and the initial grinding process parameters, the grinding condition data is collected in real time, the grinding condition data at least includes the running current of the ball mill, the inlet and outlet pressure difference, the grinding sound signal and the discharge particle size; the grinding condition data is input into the pre-trained intelligent control model, the intelligent control model outputs the state evaluation result and the process parameter adjustment instruction of the grinding process, and the running parameters of the ball mill are dynamically adjusted according to the process parameter adjustment instruction. The application can more accurately determine the initial proportioning and process parameters, and try to avoid the drastic fluctuation of the working condition after starting due to the experience setting.

Description

Technical Field

[0001] This application relates to the field of intelligent control technology, and in particular to an intelligent control method and system for the grinding process of calcium carbide solid waste. Background Technology

[0002] Calcium carbide solid waste (such as calcium carbide slag and calcium carbide dust) is a major industrial solid waste generated during the calcium carbide process for producing polyvinyl chloride (PVC) and acetylene. Its resource utilization typically requires grinding to meet the particle size requirements of subsequent applications in building materials and chemicals. However, the raw materials for calcium carbide solid waste are complex, with significant fluctuations in chemical composition (such as the content of CaO, SiO2, Al2O3, and Fe2O3), particle size distribution, and moisture content between different batches. This instability in raw material characteristics directly leads to complex characteristics in the grinding process, including strong nonlinearity, large hysteresis, and multivariate coupling.

[0003] Currently, the process control of calcium carbide solid waste grinding mainly adopts the following technical methods: Operators manually set process parameters such as the initial speed, filling rate, and grinding time of the ball mill based on raw material characteristic data obtained from periodic sampling and testing, combined with empirical formulas or simple calculation tables. During the grinding process, they intermittently adjust the operating parameters manually based on inspection records, offline screening of the output particle size, and the operator's auditory judgment of changes in grinding sound. Some enterprises have introduced distributed control systems (DCS), which realize real-time monitoring and over-limit alarms for key parameters such as ball mill operating current and inlet / outlet pressure difference. Operators can remotely issue adjustment commands from the control room based on the monitoring data.

[0004] However, the initial process parameters in the above-mentioned technologies rely solely on static empirical formulas and fail to fully integrate multi-dimensional characteristic data of raw materials. This results in poor adaptability of the initial proportioning scheme and process parameters to different batches of raw materials, drastic fluctuations in grinding conditions after startup, and a long adjustment time required to reach steady state. Summary of the Invention

[0005] In order to more accurately determine the initial ratio and process parameters based on the chemical composition, particle size and other characteristics of different batches of raw materials, and to minimize the drastic fluctuations in operating conditions after startup caused by experience-based settings, this application provides an intelligent control method and system for the grinding process of calcium carbide solid waste.

[0006] Firstly, this application provides an intelligent control method for the grinding process of calcium carbide solid waste, which adopts the following technical solution: A method for intelligent control of calcium carbide solid waste grinding process includes the following steps: Obtain characteristic data of calcium carbide solid waste raw materials, including chemical composition, particle size distribution and moisture content of calcium carbide solid waste; Based on the raw material characteristic data, the initial proportioning scheme and initial grinding process parameters for solid waste mixing and grinding are calculated through a preset proportioning model. The initial grinding process parameters include ball mill speed, filling rate and grinding time. During the grinding process according to the initial proportioning scheme and initial grinding process parameters, grinding condition data is collected in real time. The grinding condition data includes at least the ball mill's operating current, inlet and outlet pressure difference, grinding sound signal, and discharge particle size. The grinding condition data is input into a pre-trained intelligent control model, which outputs the state evaluation results of the grinding process and process parameter adjustment instructions. Based on the process parameter adjustment instructions, the operating parameters of the ball mill are dynamically adjusted.

[0007] This application obtains raw material characteristic data such as chemical composition, particle size distribution, and moisture content of calcium carbide solid waste, and pre-calculates the initial proportioning scheme and key process parameters such as ball mill speed, filling rate, and grinding time based on a proportioning model. This provides a scientific starting point for the grinding process, minimizing the blindness and instability caused by relying on manual experience to set parameters. Furthermore, this application collects operating condition data such as operating current, inlet and outlet pressure difference, grinding sound signal, and discharge particle size in real time during the grinding process. This data is then input into a pre-trained intelligent control model, enabling the system to automatically evaluate the current grinding status and dynamically adjust operating parameters. This achieves closed-loop adaptive control of the grinding process, effectively improving the stability of the finished product particle size and grinding efficiency, while reducing energy consumption and equipment wear.

[0008] Optionally, before inputting the grinding condition data into the pre-trained intelligent control model, the method further includes: comparing the operating current, inlet and outlet pressure difference, grinding sound signal and discharge particle size with their respective preset reasonable threshold ranges; when any data exceeds the corresponding threshold range, the data is discarded and marked as invalid data, and the data from the previous valid moment is used to replace it, and a data abnormality alarm signal is generated.

[0009] Before inputting operating data into the intelligent control model, this application adds a data validity verification step. The operating current, inlet and outlet pressure difference, grinding sound signal, and discharge particle size are compared with their respective reasonable threshold ranges. When any data exceeds the threshold, it is actively rejected and marked as invalid data. At the same time, the data from the previous valid moment is used to replace it and generate an alarm signal. This effectively prevents erroneous data caused by sensor failure, signal interference, or instantaneous abnormal fluctuations from entering the control model. It also minimizes the possibility of the model outputting unreasonable adjustment commands due to dirty data, improves the anti-interference capability and decision reliability of the intelligent control system, and ensures the safety and continuity of the grinding process control.

[0010] Optionally, the method further includes: After dynamically adjusting the operating parameters of the ball mill according to the process parameter adjustment instructions, the changing trends of the discharge particle size and the grinding sound signal within a preset response time are continuously monitored. If the trend of change is consistent with the expected adjustment direction of the process parameter adjustment instruction, then the adjustment is confirmed to be effective and the adjustment result is recorded. If the change trend is inconsistent with the expected direction or the output particle size does not undergo an identifiable change within the preset response time, the adjustment will be revoked and the operating parameters will be restored to those before the adjustment, and an alarm signal will be issued.

[0011] This application continuously monitors the changing trends of discharge particle size and grinding sound signal within a preset response time after executing the operating parameter adjustment command according to the process parameter adjustment command. The effectiveness of the adjustment measures is verified by judging whether the changing trend is consistent with the expected adjustment direction. When the changing trend meets the expectation, the adjustment is confirmed as effective and the result is recorded, providing data support for subsequent optimization of control strategies. When the trend is inconsistent or the particle size does not change significantly, the adjustment is promptly revoked and the system is restored to the state before adjustment, while an alarm signal is issued. This effectively avoids the risk of process runaway caused by model misjudgment or abnormal response of the actuator, enhances the robustness and self-correction capability of the control system, and effectively prevents ineffective or even harmful adjustment actions from adversely affecting the grinding quality.

[0012] Optionally, during the grinding process according to the initial proportioning scheme and initial grinding process parameters, when the characteristic data of any batch of raw materials in the initial proportioning scheme are updated, the proportioning model is used to recalculate the corrected proportioning scheme and corrected grinding process parameters of the current remaining materials; the intelligent control model uses the corrected proportioning scheme and corrected grinding process parameters as the new benchmark values ​​and continues to output process parameter adjustment instructions.

[0013] During the grinding process, when the characteristic data of any batch of raw materials in the initial proportioning scheme is updated, this application automatically triggers the proportioning model to recalculate the corrected proportioning scheme and corrected grinding process parameters of the current remaining materials, and uses the corrected parameters as the new benchmark value of the intelligent control model to continue to output adjustment instructions. This can respond in a timely manner to the differences in raw material characteristics caused by changes in raw material source, stockpile, or batch fluctuations, and try to avoid the subsequent grinding effect from deviating from the target due to the use of the initial proportioning scheme. This achieves flexible adaptation to multiple batches of heterogeneous calcium carbide solid waste raw materials, and improves the product quality consistency and process stability throughout the grinding cycle.

[0014] Optionally, the acquisition of calcium carbide solid waste raw material characteristic data is carried out intermittently according to a preset sampling period; between two adjacent sampling periods, the intelligent control model calculates the changing trend of raw material characteristics in the current grinding process in real time based on the grinding condition data, and outputs a feedforward adjustment command based on the calculated changing trend. The feedforward adjustment command is executed before the raw material characteristic data of the next sampling period is acquired.

[0015] This application acquires raw material characteristic data intermittently according to a preset sampling period. Between two adjacent sampling periods, it uses an intelligent control model to calculate the changing trend of raw material characteristics during the current grinding process based on real-time grinding condition data. Based on the calculation result, it outputs a feedforward adjustment command, which is executed before the next batch of measured data arrives. This effectively compensates for the information lag caused by intermittent sampling, enabling the system to proactively predict the impact of raw material fluctuations on the grinding process during periods of missing measured data and make compensation adjustments in advance. This improves the control system's ability to respond quickly to gradual or sudden changes in raw materials, further enhancing the smoothness of the grinding process and the stability of the finished product quality.

[0016] Optionally, the method further includes: The operating parameters of the ball mill are divided into safety-sensitive parameters and non-safety-sensitive parameters; the safety-sensitive parameters include at least the filling rate, and the non-safety-sensitive parameters include at least the ball mill speed and grinding time; the intelligent control model has automatic adjustment authority for the non-safety-sensitive parameters, and only outputs adjustment suggestions for the safety-sensitive parameters, and waits for operator confirmation before execution.

[0017] This application categorizes the operating parameters of a ball mill into safety-sensitive parameters such as filling rate and non-safety-sensitive parameters such as rotational speed and grinding time. It grants the intelligent control model automatic adjustment authority for non-safety-sensitive parameters, while only outputting adjustment suggestions for safety-sensitive parameters and awaiting operator confirmation before execution. This approach fully leverages the automated adjustment capabilities of the intelligent model to improve control efficiency, while retaining final human decision-making authority for critical parameters involving equipment safety, overload risk, or mechanical limits. This effectively avoids the risk of over-adjustment due to model prediction deviations or changes in boundary conditions, achieving a reasonable balance between automation and operational safety, and better meeting the actual reliability requirements of industrial sites.

[0018] Optionally, when the status assessment result indicates that the current working condition is in a preset emergency state, the intelligent control model temporarily obtains the automatic adjustment authority of the safety sensitive parameters.

[0019] When the status assessment results indicate that the current operating condition is in a preset emergency state, this application temporarily grants the intelligent control model the authority to automatically adjust safety-sensitive parameters. Under sudden operating conditions, the system can quickly intervene and adjust key parameters such as filling rate without waiting for manual confirmation from the operator. This enables the system to take protective measures immediately in emergency situations such as equipment overload, mill blockage, and sudden increase in differential pressure, effectively preventing the situation from deteriorating, improving the active safety protection capability and emergency response timeliness of the grinding system, and ensuring the safety of equipment and personnel.

[0020] Optionally, the process parameter adjustment instructions output by the intelligent control model include coarse adjustment instructions and fine adjustment instructions; When the state evaluation result is that the particle size deviation of the output exceeds the first preset deviation threshold, the coarse adjustment command is output. When the state evaluation result is that the output particle size deviation is between the first preset deviation threshold and the second preset deviation threshold, the fine-tuning command is output, and the second preset deviation threshold is less than the first preset deviation threshold. When dynamically adjusting the operating parameters of the ball mill, the execution cycles of the coarse adjustment command and the fine adjustment command are different.

[0021] This application outputs two adjustment commands with different amplitudes—coarse adjustment and fine adjustment—through an intelligent control model. The type of command used is determined based on the degree of particle size deviation in the output material: coarse adjustment is executed when the deviation is large, and fine adjustment is executed when the deviation is small. At the same time, different execution cycles are set for the two types of commands, enabling the control system to quickly and significantly correct process parameters to shorten the recovery time when the deviation is significant, while performing fine and small-step adjustments when the deviation is small to avoid overshoot and oscillation. This balances the rapid response and steady-state accuracy of the control system, effectively improving the control quality of the grinding process under different deviation states and reducing frequent fluctuations or over-adjustment problems caused by improper adjustment amplitude.

[0022] Secondly, this application provides an intelligent control system for the grinding process of calcium carbide solid waste, which adopts the following technical solution: An intelligent control system for the grinding process of calcium carbide solid waste includes: A processor, and a memory communicatively connected to the processor; The memory is provided with a computer-readable storage medium, and a computer program is stored on the computer-readable storage medium. When the processor processes a computer program stored on the computer-readable storage medium, it implements the method as described in the first aspect.

[0023] In summary, this application includes at least one of the following beneficial technical effects: 1. This application obtains raw material characteristic data such as chemical composition, particle size distribution and moisture content of calcium carbide solid waste, and pre-calculates the initial proportioning scheme and key process parameters such as ball mill speed, filling rate and grinding time based on the proportioning model. This provides a scientific starting benchmark for the grinding process and can minimize the blindness and instability caused by relying on manual experience to set parameters.

[0024] 2. This application collects operating condition data such as operating current, inlet and outlet pressure difference, grinding sound signal and discharge particle size in real time during the grinding process, and inputs these data into a pre-trained intelligent control model, enabling the system to automatically evaluate the current grinding status and dynamically adjust the operating parameters. This achieves closed-loop adaptive control of the grinding process, effectively improving the stability of the finished product particle size and grinding efficiency, while reducing energy consumption and equipment wear.

[0025] 3. Before inputting the operating data into the intelligent control model, this application adds a data validity verification step. The operating current, inlet and outlet pressure difference, grinding sound signal, and discharge particle size are compared with their respective reasonable threshold ranges. When any data exceeds the threshold, it is actively rejected and marked as invalid data. At the same time, the data from the previous valid moment is used to replace it and generate an alarm signal. This effectively prevents erroneous data caused by sensor failure, signal interference, or instantaneous abnormal fluctuations from entering the control model. It also minimizes the possibility of the model outputting unreasonable adjustment commands due to dirty data, improves the anti-interference capability and decision reliability of the intelligent control system, and ensures the safety and continuity of the grinding process control. Attached Figure Description

[0026] Figure 1 This is a flowchart of Embodiment 1 of this application. Detailed Implementation

[0027] The following combination Figure 1 This application will be described in further detail.

[0028] Example 1: This example discloses an intelligent control method for the grinding process of calcium carbide solid waste, referring to... Figure 1 The method includes: S1 data acquisition, S2 parameter calculation, S3 real-time acquisition of grinding condition data, S4 data verification, S5 output, and S6 verification and rollback. It uses raw material characteristic data of calcium carbide solid waste as input, determines the initial proportioning scheme and grinding process parameters using a proportioning model, and collects condition data in real time during the grinding process. Then, an intelligent control model evaluates the grinding status and outputs adjustment commands to achieve dynamic closed-loop control of the grinding process. The execution process of each step in this embodiment is as follows: S1 Data Acquisition: In this embodiment, the calcium carbide solid waste raw materials mainly originate from industrial waste residue generated during the calcium carbide-based acetylene production process, including but not limited to calcium carbide slag, silica corrective materials (such as fly ash, silica powder, etc.), and alumina corrective materials (such as bauxite, coal gangue, etc.). Raw material characteristic data are acquired through online detection.

[0029] In this embodiment, a near-infrared spectroscopy analyzer and an X-ray fluorescence spectroscopy analyzer are installed on the raw material conveyor belt or feeding pipeline to acquire the characteristic data of each batch of raw materials in real time using an online joint measurement method. The raw material characteristic data includes: the chemical composition, particle size distribution, and moisture content of calcium carbide solid waste.

[0030] Chemical composition: mainly includes the mass percentage of components such as calcium oxide (CaO), silicon dioxide (SiO2), aluminum oxide (Al2O3), iron oxide (Fe2O3), and magnesium oxide (MgO).

[0031] Particle size distribution: The particle size distribution of the raw material was detected online using a laser particle size analyzer. In this embodiment, it was characterized by the residue on different mesh sizes or the volume average particle size.

[0032] Moisture content: The moisture content of the raw materials was obtained using an online moisture analyzer.

[0033] In this embodiment, the acquisition of calcium carbide solid waste raw material characteristic data is carried out intermittently according to a preset sampling cycle. For example, a complete raw material characteristic detection and sampling record is completed every 10 minutes, and the detection data is uploaded to the data server of the central control system. During two adjacent sampling cycles, if the raw material source is relatively stable, the latest valid detection data is used as the input for the proportioning model.

[0034] Between two adjacent sampling periods, the characteristics of raw materials may also change gradually due to fluctuations in upstream processes. If control is based solely on periodic sampling data, there is a risk of detection lag.

[0035] This embodiment introduces a feedforward adjustment strategy to reduce problems after detection. Specifically, between two adjacent sampling cycles, the intelligent control model calculates the changing trend of raw material characteristics in the current grinding process in real time based on the grinding condition data, and outputs a feedforward adjustment command based on the calculated changing trend. The feedforward adjustment command is executed before the raw material characteristic data of the next sampling cycle is acquired.

[0036] The intelligent control model infers the possible direction of change in raw material characteristics by monitoring grinding condition data (such as the slow drift of operating current, the gradual change of grinding sound signal, and the subtle change trend of output particle size). In this embodiment, a qualitative / semi-quantitative mapping rule library between operating condition parameters and raw material characteristics is established in advance, and the change trend is identified by the rules, as shown in Table 1.

[0037] Table 1

[0038] For example, if the operating current shows a slow upward trend and the output particle size tends to coarsen over three consecutive minutes, the intelligent control model infers that the moisture content in the raw material may be increasing. Before the next batch of raw material test data is available, the intelligent control model outputs a feedforward adjustment command to slightly increase the ball mill speed. This feedforward adjustment is smaller than the feedback adjustment made based on the actual test data, in order to offset the expected disturbance in advance and achieve the effect of predictive control.

[0039] Once the actual raw material characteristic data for the next sampling period is obtained, the intelligent control model will compare the actual data with the previously calculated trend, conduct a self-evaluation of the accuracy of the feedforward adjustment, and use the evaluation results to correct subsequent feedforward calculations.

[0040] After obtaining the raw material characteristic data by calculating the S2 parameters, the data is input into the proportioning model preset in the central control system or edge computing terminal. The proportioning model is used to determine the initial proportioning scheme and initial grinding process parameters of solid waste mixing and grinding based on the actual quality of the current batch of raw materials.

[0041] In this embodiment, the matching model can be an optimization model based on linear programming or a matching decision model based on fuzzy logic.

[0042] This embodiment uses a fuzzy logic matching model as an example to further illustrate this step.

[0043] The chemical composition, particle size distribution, and moisture content of the calcium carbide solid waste are fuzzified into corresponding fuzzy sets. In this embodiment, the CaO content is divided into three fuzzy subsets: low, moderate, and high; the particle size distribution is divided into three fuzzy subsets: fine, medium, and coarse; and the moisture content is divided into three fuzzy subsets: low, medium, and high.

[0044] Fuzzy inference is performed based on a pre-defined fuzzy rule base, which is pre-established based on the mutual grinding-aiding effect between the components in the calcium carbide solid waste and the clinker yield correction coefficient. For example, a fuzzy rule can be expressed as: if the CaO content in the current calcium carbide slag is low and the SiO2 content in the silica corrector is moderate, then the proportion of calcium carbide slag should be appropriately increased and the proportion of silica corrector should be decreased, while the ball mill speed should be set to a higher level.

[0045] The result of fuzzy reasoning consists of two parts: Initial proportioning scheme: Determine the mass percentage of each of the carbide slag, silica corrective material, and alumina corrective material.

[0046] In this embodiment, the proportioning model is constrained by the mass percentages of calcium oxide, silicon dioxide, aluminum oxide and iron oxide in the ground product meeting the cement raw meal ratio requirements (e.g., KH=0.88~0.92, SM=2.4~2.8, IM=1.2~1.6), and the proportioning is optimized with the goal of minimizing grinding energy consumption. The sum of the mass percentages of each component is 100%.

[0047] Initial grinding process parameters include ball mill speed (unit: r / min), filling rate (percentage of grinding media in the effective volume of the cylinder) and grinding time (unit: min or h).

[0048] Taking a specific batch of raw materials as an example, assume that the initial proportioning scheme output after calculation by the proportioning model is: 78% calcium carbide slag, 15% silica correcting material, and 7% alumina correcting material; The initial grinding process parameters are: ball mill speed is 72% of the critical speed of the cylinder, filling rate is 30%, and grinding time is 45 minutes.

[0049] After the proportioning model is calculated, the initial proportioning scheme is sent to the batching control system to execute automatic batching, and the initial grinding process parameters are sent to the ball mill control system to start the grinding operation.

[0050] S3 collects grinding condition data in real time. During the grinding process according to the initial proportioning scheme and initial grinding process parameters, it collects grinding condition data in real time through multiple sensors arranged on the ball mill and its auxiliary equipment.

[0051] The grinding condition data includes the following four categories: Operating current: Collected by a current transformer installed on the main motor power supply line of the ball mill, and the unit is amperes (A). The operating current reflects the load condition of the ball mill. An increase in current usually means that the amount of material filling the mill has increased or the grinding resistance has increased, while a decrease in current may indicate that the mill is empty or that the grinding media is worn out.

[0052] Inlet and outlet pressure difference: Data is collected by pressure transmitters installed at the feed and discharge ends of the ball mill, and the unit is kilopascal (kPa). The inlet and outlet pressure difference reflects the ventilation resistance inside the mill, and changes in the pressure difference can indirectly indicate the material level and ventilation status inside the mill.

[0053] Grinding sound signal: Collected by a grinding sound sensor (such as an electric ear) installed near the ball mill cylinder (e.g., outside the cylinder or near the grinding head), the unit is decibel (dB) or an amplified voltage / current signal. The strength of the grinding sound signal reflects the impact state between the grinding media and the material: a crisp and loud grinding sound indicates normal grinding, a low grinding sound indicates too much material or too thick a material layer, and a sharp and harsh grinding sound may indicate empty grinding or insufficient grinding media.

[0054] Discharge particle size: Collected by an online laser particle size analyzer installed on the discharge pipe of the ball mill, and expressed as 80μm sieve residue or specific surface area (m² / kg).

[0055] In other embodiments, the active power (kW) and bearing temperature (°C) of the main motor of the ball mill can also be collected to help determine the health status of the equipment. Vibration spectrum signals can also be collected by vibration sensors installed on the surface of the ball mill cylinder; these vibration spectrum signals, together with the grinding sound signals, are used to determine the impact state of the grinding media and the degree of liner wear.

[0056] All operating data are transmitted in real time to the data acquisition and monitoring control (SCADA) platform of the central control system via fieldbus (such as Profibus, Modbus, etc.) or industrial Ethernet. The sampling frequency is usually set to 1 to 10 times per second to meet the processing requirements of the subsequent intelligent control model.

[0057] In other embodiments, during continuous grinding production according to the initial formulation and initial grinding process parameters, the raw materials in each batch of the initial formulation are gradually consumed as grinding progresses. When a new batch of raw materials enters the production line, the material characteristics data of the new batch may be updated relative to the initial batch due to reasons such as changes in the source of the ore, uneven mixing in the stockpile, or changes in the supplier.

[0058] In this embodiment, when the characteristic data of any batch of raw materials in the initial proportioning scheme is updated (i.e., the chemical composition, particle size distribution or moisture content of the new batch is significantly different from the previous batch), the system automatically triggers the proportioning model to recalculate the corrected proportioning scheme and corrected grinding process parameters of the current remaining materials.

[0059] Obtain online testing data for the new batch, and perform a weighted average of the new batch data and the current remaining inventory to obtain the comprehensive raw material characteristics of the current remaining mixture; The comprehensive raw material characteristics were re-input into the proportioning model, and the calculation was re-optimized with the goals of achieving the target grinding product yield and minimizing energy consumption. The proportioning model outputs a corrected proportioning scheme (i.e., the new proportions of the remaining materials) and corrected grinding process parameters (i.e., the rotation speed, filling rate, and grinding time under the new conditions).

[0060] After recalculation, the intelligent control model uses the corrected proportioning scheme and corrected grinding process parameters as new benchmark values ​​and continues to output process parameter adjustment commands in subsequent grinding processes.

[0061] S4 Data Verification: In this embodiment, the operating current, inlet and outlet pressure difference, grinding sound signal, and discharge particle size are compared with their respective preset reasonable threshold ranges. The reasonable threshold ranges for each parameter can be determined based on the ball mill's equipment specifications, design parameters, and historical normal operating data.

[0062] For example: the reasonable threshold range of operating current: [rated current × 60%, rated current × 110%]; the reasonable threshold range of inlet and outlet pressure difference: [0.5kPa, 3.5kPa]; the reasonable threshold range of grinding sound signal: [60dB, 110dB]; the reasonable threshold range of discharge particle size: [5% (80μm sieve residue), 25% (80μm sieve residue)].

[0063] When any data exceeds the corresponding threshold range, the system determines that the data is invalid, removes it, and marks it as an invalid data record. Simultaneously, it replaces the data with the data collected at the previous valid time point and generates a data anomaly alarm signal. The alarm signal is sent to the operator interface in the central control room in the form of an audible and visual alarm or a pop-up window, reminding the operator to check the working status of the corresponding sensor.

[0064] This data validity verification step can prevent abnormal data caused by sensor failure, signal interference, or communication interruption from entering the intelligent control model, thus avoiding the model from outputting incorrect adjustment commands.

[0065] S5 output: The intelligent control model can be a neural network model trained based on reinforcement learning algorithms such as Deep Q Network (DQN) or Proximal Policy Optimization (PPO), or a hybrid model combining a time-series prediction model based on Long Short-Term Memory Network (LSTM) with a rule base.

[0066] The training data for the intelligent control model comes from historical grinding production data, including multiple sets of historical raw material characteristic data, historical operating condition data, corresponding historical process parameter adjustment records, and final product quality data.

[0067] After receiving the grinding condition data at the current moment, the intelligent control model executes two output tasks: the state evaluation result of the grinding process and the process parameter adjustment instruction.

[0068] The status assessment results include at least one of the following or any combination thereof: (1) Health score of the current grinding state. The score can be a continuous value between 0 and 100. The higher the score, the better the grinding state.

[0069] (2) Abnormal operating condition type identification, wherein the abnormal operating condition type includes at least overheating, empty grinding and blocked grinding. For example, when the operating current is continuously high and the inlet and outlet pressure difference increases significantly and the grinding sound is low and muffled, the model identifies it as "overheating"; when the operating current is continuously low and the grinding sound is sharp, the model identifies it as "empty grinding"; when the inlet and outlet pressure difference increases sharply and the discharge particle size coarsens, the model identifies it as "blocked grinding".

[0070] (3) Mill filling status level and grinding efficiency level. The mill filling status level is determined based on the correlation trend between the operating current and the inlet / outlet pressure difference. For example, if the operating current and the inlet / outlet pressure difference increase simultaneously, the filling status is considered to be oversaturated; if both decrease simultaneously, the filling status is considered insufficient. The grinding efficiency level is determined based on the mapping relationship between the grinding sound signal and the output particle size. For example, when the grinding sound signal is in the normal range and the output particle size continues to refine, the grinding efficiency is considered high; when the grinding sound signal is normal but the output particle size does not change significantly, the grinding efficiency is considered to be saturated.

[0071] (4) Equipment health status level, which is determined by a combination of the rate of change of active power and the rate of increase of bearing temperature, and is used to assess whether there is abnormal wear or overheating risk in the equipment.

[0072] The intelligent control model generates corresponding process parameter adjustment instructions based on the state assessment results. These instructions are used to dynamically adjust the operating parameters of the ball mill, including at least the adjustment amounts for the ball mill speed, filling rate, and grinding time.

[0073] In this embodiment, the process parameter adjustment commands are divided into two types: coarse adjustment commands and fine adjustment commands, to adapt to different degrees of deviation in operating conditions. When the state assessment result indicates that the discharge particle size deviation exceeds the first preset deviation threshold, meaning the deviation between the current discharge particle size and the target particle size is large (e.g., the 80μm sieve residue exceeds the target value by more than ±5 percentage points), the intelligent control model outputs the coarse adjustment command. The coarse adjustment command corresponds to a larger adjustment range and a faster adjustment rate, aiming to quickly bring the discharge particle size back to near the target range.

[0074] When the state assessment result indicates that the discharge particle size deviation is between the first preset deviation threshold and the second preset deviation threshold, and the second preset deviation threshold is less than the first preset deviation threshold (for example, the 80μm sieve residue exceeds the target value by ±2~5 percentage points), the intelligent control model outputs the fine-tuning instruction. The fine-tuning instruction corresponds to a smaller adjustment range and a slower adjustment rate, used for fine adjustment to avoid fluctuations caused by over-adjustment.

[0075] When dynamically adjusting the operating parameters of the ball mill, the execution cycles of the coarse adjustment command and the fine adjustment command are different. For example, the coarse adjustment command has a shorter execution cycle (e.g., once every 30 seconds) to meet the need for rapid response; the fine adjustment command has a longer execution cycle (e.g., once every 5 minutes) to allow the system to stabilize sufficiently between adjustments.

[0076] Furthermore, as a more preferred embodiment, the process parameter adjustment command also carries an adjustment range level identifier, which is determined based on the difference between the current output particle size and the target particle size. When dynamically adjusting the ball mill's operating parameters, the process is executed in the following order: first, the ball mill speed; then, the filling rate; and finally, the grinding time. Before executing a subsequent adjustment, the change in the output particle size after the previous adjustment is evaluated. If the change has brought the output particle size within the target allowable deviation range, the adjustment command for the subsequent item is terminated. This sequential execution strategy avoids coupling interference caused by adjusting multiple parameters simultaneously, facilitating accurate attribution of the actual effect of each adjustment.

[0077] This embodiment uses reinforcement learning (DQN algorithm) as an example to further illustrate this step.

[0078] The state space is:

[0079] In this embodiment, the state space includes 12-dimensional raw material characteristics, 10-dimensional operating condition characteristics, 4-dimensional current parameters, and 2-dimensional quality indicators.

[0080] The motion space is:

[0081] In this embodiment, each action dimension is discretized into 5 levels (e.g., -2 levels, -1 levels, 0, +1 levels, +2 levels), resulting in a combined action space of approximately 625 discrete actions.

[0082] The reward function is:

[0083]

[0084] in, For the reward function; The reward is based on energy consumption per unit of output; the lower the energy consumption, the higher the score. The reward is for fluctuations in process parameters; the smaller the fluctuation, the higher the reward. For product quality rewards, positive rewards are given for meeting the standards, and the greater the deviation from the target value, the greater the penalty. In this embodiment, product quality rewards... The calculation model is shown in Table 2.

[0085] Table 2

[0086] In other embodiments, other values ​​may be set as needed.

[0087] The model parameters of the DQN are as follows: the input layer is used to receive a 28-dimensional state vector, including raw material characteristics (12-dimensional), operating condition data (10-dimensional), current process parameters (4-dimensional), and recent quality indicators (2-dimensional).

[0088] There are three hidden layers. The first hidden layer contains 256 neurons and uses the ReLU activation function to extract preliminary features from the original input. The second hidden layer contains 256 neurons and uses the ReLU activation function to perform deeper feature combination based on the first layer. The third hidden layer contains 128 neurons and uses the ReLU activation function.

[0089] The output layer outputs 625 values, each corresponding to an evaluation score for a combination of actions. These 625 actions come from four control parameters (feed rate, grinding pressure, classifier speed, and ventilation volume), each with five levels.

[0090] Ultimately, the action with the highest Q score is selected for execution.

[0091] This embodiment also divides the operating parameters of the ball mill into safety-sensitive parameters and non-safety-sensitive parameters: Safety-sensitive parameters include at least the filling rate. The filling rate determines the amount of grinding media loaded, which directly affects the starting torque, operating load, and cylinder stress of the ball mill. If the filling rate exceeds the upper limit of the equipment design, it may cause the motor to overload and burn out or the cylinder to deform. Therefore, it is classified as a safety-sensitive parameter.

[0092] Non-safety-sensitive parameters include at least the ball mill speed and grinding time. When the speed is adjusted within a certain range (such as 60% to 80% of the critical speed), the impact on equipment safety is relatively small; changes in grinding time do not affect the instantaneous load of the equipment, and the safety risk is low.

[0093] Under normal operating conditions, the intelligent control model has the authority to automatically adjust the non-safety sensitive parameters (speed and grinding time). That is, the speed and grinding time adjustment commands output by the model can be directly sent to the actuators (frequency converter, timer) for automatic execution without manual confirmation.

[0094] Regarding the safety-sensitive parameter (fill rate), the intelligent control model only outputs adjustment suggestions and displays them to the operator through a human-machine interface (HMI). It also includes the suggested adjustment amount, the reason for the adjustment, and an explanation of the expected effect. The system will only perform the fill rate adjustment operation after the operator clicks to confirm on the interface.

[0095] However, when the status assessment result indicates that the current operating condition is in a preset emergency state, the intelligent control model temporarily gains the authority to automatically adjust the safety-sensitive parameters. An emergency state, for example, might be identified as "overheating" with the operating current approaching the rated current limit and a sharp increase in the inlet / outlet pressure difference potentially causing mill blockage and shutdown; or as "dry grinding" with abnormally sharp grinding noise and a rapid rise in bearing temperature potentially causing equipment damage. In such emergency states, time is limited to waiting for operator confirmation at each level; therefore, the model temporarily gains the authority to automatically adjust the fill rate, immediately executing operations to reduce or increase the fill rate to bring the operating condition back to a safe range, while simultaneously notifying the operator of the emergency adjustment event as a high-priority alarm.

[0096] Once the emergency is lifted, the system will automatically switch back to the normal access control mode, meaning the fill rate will be restored to require operator confirmation before execution.

[0097] S6 Verification and Rollback: After a certain process parameter adjustment, continuously monitor the changing trends of the output particle size and the grinding sound signal within a preset response time (e.g., 2-5 minutes). The determination method is as follows: If the trend of change is consistent with the expected adjustment direction of the process parameter adjustment command, the adjustment is confirmed to be effective and the result is recorded. For example, when the intelligent control model outputs a command to increase the ball mill speed due to excessively coarse discharge particle size, if the discharge particle size is detected to gradually become finer and the grinding sound signal tends to normal within a preset response time, the adjustment is determined to be effective. The corresponding operating data, adjustment command, and response result are recorded in the system log for subsequent model optimization or process traceability.

[0098] If the trend of change is inconsistent with the expected direction, or if the output particle size does not undergo a identifiable change within the preset response time (i.e., the change is less than the resolution or noise level of the detection instrument), then the adjustment is deemed invalid, the adjustment is revoked, and the operating parameters are restored to their previous state. Simultaneously, an alarm signal is issued to alert the operator to intervene and check. The alarm signal may include audible and visual alarms, pop-up windows on the central control screen, or SMS notifications.

[0099] Example 2: This example discloses an intelligent control system for the grinding process of calcium carbide solid waste. The system includes: A processor, and a memory communicatively connected to the processor; The memory is provided with a computer-readable storage medium, and a computer program is stored on the computer-readable storage medium. When the processor processes a computer program stored on the computer-readable storage medium, it implements the method.

[0100] The above are all preferred embodiments of this application, and are not intended to limit the scope of protection of this application. Therefore, all equivalent changes made in accordance with the structure, shape and principle of this application should be covered within the scope of protection of this application.

Claims

1. A method for intelligent control of calcium carbide solid waste grinding process, characterized in that, include: Obtain characteristic data of calcium carbide solid waste raw materials, including chemical composition, particle size distribution and moisture content of calcium carbide solid waste; Based on the raw material characteristic data, the initial proportioning scheme and initial grinding process parameters for solid waste mixing and grinding are calculated through a preset proportioning model. The initial grinding process parameters include ball mill speed, filling rate and grinding time. During the grinding process according to the initial proportioning scheme and initial grinding process parameters, grinding condition data is collected in real time. The grinding condition data includes at least the ball mill's operating current, inlet and outlet pressure difference, grinding sound signal, and discharge particle size. The grinding condition data is input into a pre-trained intelligent control model, which outputs the state evaluation results of the grinding process and process parameter adjustment instructions. Based on the process parameter adjustment instructions, the operating parameters of the ball mill are dynamically adjusted.

2. The intelligent control method for the calcium carbide solid waste grinding process according to claim 1, characterized in that, Before inputting the grinding condition data into the pre-trained intelligent control model, the method further includes: comparing the operating current, inlet and outlet pressure difference, grinding sound signal and discharge particle size with their respective preset reasonable threshold ranges. When any data exceeds the corresponding threshold range, the data is discarded and marked as invalid data, and the data from the previous valid moment is used to replace it, and a data abnormality alarm signal is generated.

3. The intelligent control method for the calcium carbide solid waste grinding process according to claim 1, characterized in that, The method further includes: After dynamically adjusting the operating parameters of the ball mill according to the process parameter adjustment instructions, the changing trends of the discharge particle size and the grinding sound signal within a preset response time are continuously monitored. If the trend of change is consistent with the expected adjustment direction of the process parameter adjustment instruction, then the adjustment is confirmed to be effective and the adjustment result is recorded. If the change trend is inconsistent with the expected direction or the output particle size does not undergo an identifiable change within the preset response time, the adjustment will be revoked and the operating parameters will be restored to those before the adjustment, and an alarm signal will be issued.

4. The intelligent control method for the calcium carbide solid waste grinding process according to claim 1, characterized in that, During the grinding process according to the initial proportioning scheme and initial grinding process parameters, when the characteristic data of any batch of raw materials in the initial proportioning scheme are updated, the proportioning model is used to recalculate the corrected proportioning scheme and corrected grinding process parameters of the current remaining materials; the intelligent control model uses the corrected proportioning scheme and corrected grinding process parameters as the new benchmark values ​​and continues to output process parameter adjustment instructions.

5. The intelligent control method for the calcium carbide solid waste grinding process according to claim 1, characterized in that, The acquisition of calcium carbide solid waste raw material characteristic data is carried out intermittently according to a preset sampling period. Between two adjacent sampling periods, the intelligent control model calculates the changing trend of raw material characteristics in the current grinding process in real time based on the grinding condition data, and outputs a feedforward adjustment command based on the calculated changing trend. The feedforward adjustment command is executed before the raw material characteristic data of the next sampling period is acquired.

6. The intelligent control method for the calcium carbide solid waste grinding process according to claim 1, characterized in that, The method further includes: The operating parameters of the ball mill are divided into safety-sensitive parameters and non-safety-sensitive parameters; the safety-sensitive parameters include at least the filling rate, and the non-safety-sensitive parameters include at least the ball mill speed and grinding time; the intelligent control model has automatic adjustment authority for the non-safety-sensitive parameters, and only outputs adjustment suggestions for the safety-sensitive parameters, and waits for operator confirmation before execution.

7. The intelligent control method for the calcium carbide solid waste grinding process according to claim 6, characterized in that, When the status assessment result indicates that the current operating condition is in a preset emergency state, the intelligent control model temporarily obtains the automatic adjustment authority of the safety sensitive parameters.

8. The intelligent control method for the calcium carbide solid waste grinding process according to claim 1, characterized in that, The process parameter adjustment instructions output by the intelligent control model include coarse adjustment instructions and fine adjustment instructions; When the state evaluation result is that the particle size deviation of the output exceeds the first preset deviation threshold, the coarse adjustment command is output. When the state evaluation result is that the output particle size deviation is between the first preset deviation threshold and the second preset deviation threshold, the fine-tuning command is output, and the second preset deviation threshold is less than the first preset deviation threshold. When dynamically adjusting the operating parameters of the ball mill, the execution cycles of the coarse adjustment command and the fine adjustment command are different.

9. An intelligent control system for the grinding process of calcium carbide solid waste, characterized in that, include: A processor, and a memory communicatively connected to the processor; The memory is provided with a computer-readable storage medium, and a computer program is stored on the computer-readable storage medium. When the processor processes a computer program stored on the computer-readable storage medium, it implements the method as described in any one of claims 1-8.

Images