Phosphate production monitoring and control method based on artificial intelligence

Abstract

The invention belongs to the technical field of phosphate production, and particularly relates to a phosphate production monitoring and control method based on artificial intelligence, which comprises the steps of data acquisition, data processing, data analysis, data output and feedback control. According to the method, multi-source sensor fusion and dynamic parameter calculation are carried out, reaction dynamic states are directly captured, preprocessing delay caused by interference is avoided, process adjustment response time is compressed, on the basis of joint optimization of a quality prediction sub-module and an energy consumption evaluation sub-module, and automatic compensation of an AI model on interference factors is combined, so that the detection precision is improved, and the detection precision is improved. Meanwhile, the equipment maintenance frequency is reduced, a multi-source heterogeneous data fusion platform is constructed, the whole-process quality-energy consumption relation is mapped in real time through a dynamic production state matrix, an AI model is driven to generate optimization instructions, data islands are eliminated, and process optimization decisions are synchronized in real time.

Description

Technical Field

[0001] The present invention belongs to the field of phosphate production, and particularly relates to its monitoring and control, and specifically discloses a phosphate production monitoring and control method based on artificial intelligence. Background Art

[0002] Phosphate is an important chemical raw material, widely used in food additives, new energy materials, and agricultural fertilizers. Its industrial production mainly relies on phosphate rock processing. The current mainstream processes include wet process, thermal process and semi-aqueous process.

[0003] Existing phosphate production monitoring technologies mainly rely on a combination of offline laboratory analysis and basic sensor networks: intermittent testing of the purity of raw materials, intermediates and finished products through chemical titration, spectrophotometry or ion chromatography, while conventional temperature and pressure sensors are used to collect process parameters. However, these methods have significant flaws - laboratory testing takes up to several hours, sensor data is not linked to component analysis, resulting in poor real-time performance and limited accuracy, and multi-source data is not integrated to form "information islands", which makes it difficult to support dynamic optimization.

[0004] Poor real-time performance: Fe 3+ , turbidity, and organic interference sources require complex pretreatment, reducing the real-time performance of online monitoring.

[0005] Limited accuracy: High-precision methods have high maintenance costs, and rapid screening methods lack accuracy and are difficult to meet the needs of the entire process.

[0006] Data silos: Laboratory, sensor, and manually recorded data are not integrated, resulting in delayed process optimization.

[0007] Therefore, a real-time monitoring, high-precision and parameter-linked intelligent monitoring and control method is needed to solve the above problems. Summary of the Invention

[0008] In view of this, the present invention proposes a phosphate production monitoring and control method based on artificial intelligence. The reactor temperature, air pressure, pH value, phosphate concentration and impurity ion concentration data are collected in real time through multi-source sensors. The temperature change rate, pressure fluctuation variance, and pH root mean square deviation dynamic parameters are calculated, and the reaction efficiency and product purity are integrated to construct a comprehensive quality index. The dynamic production status matrix is established in combination with the total energy consumption. The matrix is then input into a pre-trained artificial intelligence optimization model, and the quality prediction and energy consumption evaluation sub-modules are integrated to generate real-time optimization instructions for temperature regulation, raw material addition rate and stirring intensity. The closed-loop dynamic adjustment of the process parameters is realized through the controller, and finally the coordinated optimization of production quality, energy efficiency and stability is achieved.

[0009] The purpose of the present invention can be achieved by the following technical solutions:

[0010] A phosphate production monitoring and control method based on artificial intelligence is characterized by comprising the following steps:

[0011] S1. Data acquisition: The sensor group collects multi-source data from the phosphate production process in real time, including reactor temperature data, air pressure data, pH value data, phosphate concentration data, impurity ion concentration data, and energy consumption data;

[0012] S2. Data processing: Process the collected multi-source data once to obtain the temperature change rate, pressure fluctuation variance, and root mean square deviation of pH value, and calculate the reaction efficiency, product purity and total energy consumption;

[0013] S3. Data Analysis: Integrate reaction efficiency and product purity to obtain comprehensive quality indicators; calculate energy efficiency evaluation models through total energy consumption measurement and construct a dynamic production status matrix;

[0014] S4. Data output: The dynamic production state matrix is input into the artificial intelligence optimization model, and the production parameter adjustment instructions are output; the artificial intelligence optimization model is obtained by training historical production data and corresponding optimal control parameters, and integrates the quality prediction submodule and the energy consumption assessment submodule;

[0015] S5. Feedback control: According to the adjustment instructions, the controller dynamically adjusts the reactor temperature, raw material addition rate and stirring intensity to achieve closed-loop control of phosphate production.

[0016] Combining all the above technical solutions, the present invention has the following positive effects:

[0017] 1. The present invention combines multiple source sensors with dynamic parameter calculation to directly capture the reaction dynamics and avoid Fe 3+ / Pretreatment delay caused by turbidity interference compresses the process adjustment response time.

[0018] 2. The present invention is based on the joint optimization of the quality prediction submodule and the energy consumption assessment submodule, combined with the automatic compensation of interference factors by the AI model, to achieve improved detection accuracy without the need for high-cost chromatography / titration equipment, while reducing the frequency of equipment maintenance.

[0019] 3. The present invention constructs a multi-source heterogeneous data fusion platform, which maps the quality-energy consumption relationship of the entire process in real time through a dynamic production status matrix, drives the AI model to generate optimization instructions, eliminates data silos, and synchronizes process optimization decisions in real time. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0021] Attachment Figure 1 It is a step diagram of the present invention.

[0022] Attachment Figure 2 Flowchart of the present invention. DETAILED DESCRIPTION

[0023] The technical solutions of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0024] See also Figure 2 As shown, the present invention proposes a phosphate production monitoring and control method based on artificial intelligence, including a data acquisition end, a data processing end, a data analysis end, a data output end and a feedback control end.

[0025] like Figure 1 As shown, the specific implementation steps of the present invention include the following steps:

[0026] S1. Data acquisition: The sensor group collects multi-source data from the phosphate production process in real time, including reactor temperature data, air pressure data, pH value data, phosphate concentration data, impurity ion concentration data and energy consumption data.

[0027] It should be noted that the real-time collection of multi-source data by the sensor group is specifically as follows:

[0028] The PT100 temperature sensor collects reaction temperature data, the piezoelectric pressure sensor collects air pressure data, the glass electrode pH meter detects pH data in real time, the UV spectrophotometer analyzes phosphate concentration data online, the ion selective electrode array detects impurity ion concentration data, and the Hall current sensor collects stirring motor power consumption data. Combined with the steam flow meter, thermal energy consumption data is calculated, where the energy consumption data includes heating power, stirring power, and material conveying power.

[0029] It should be noted that the multi-source data needs to be collected synchronously, based on hardware timestamp alignment to achieve millisecond-level synchronization of devices with different sampling frequencies, and the cubic spline interpolation method is used to reconstruct a unified time series for non-equally spaced sampling data.

[0030] After collecting the data, the basic data is cleaned and calibrated as follows:

[0031] First, the signal is denoised, specifically:

[0032] Wavelet threshold denoising was used on pH sensor data to eliminate electrode polarization interference; Kalman filtering was implemented on temperature sensor data to eliminate measurement lag caused by thermal inertia; and a sliding window quartile method with a window length of 30 seconds was used to detect and eliminate outliers.

[0033] Then calibrate and compensate the sensor, specifically:

[0034] A temperature-pressure cross-compensation model was established to correct the measurement deviation of the pressure sensor under high temperature fluctuation conditions; dynamic calibration was implemented for the phosphate concentration probe: standard solution was injected every 2 hours for online calibration.

[0035] S2. Data processing: Process the collected multi-source data once to obtain the temperature change rate, pressure fluctuation variance, and root mean square deviation of pH value, and calculate the reaction efficiency, product purity and total energy consumption.

[0036] It should be noted that the temperature change rate is:

[0037]

[0038] Where n represents the total number of temperature data, Represents the rate of temperature change at any time, T i -T i-1 It represents the temperature change, which is the temperature difference between adjacent times. Δt represents the time interval, that is, the time difference between two temperature measurements.

[0039] Characterizes the severity of temperature changes over time and reflects the thermodynamic stability of the process; excessively high temperature change rates cause the reaction path to deviate from expectations, triggering side reactions or crystallization defects.

[0040] The pressure fluctuation variance is specifically:

[0041]

[0042] in m represents the total amount of pressure data, X i It is any pressure data.

[0043] Quantifying the discreteness of pressure parameters reflects the mechanical stability of the reaction system; pressure fluctuations will destroy the gas-liquid equilibrium or cause increased equipment vibration.

[0044] The larger the variance, the more severe the pressure fluctuation; the smaller the variance, the more stable the pressure; if the variance is 0, it means that all pressure data are exactly the same and there is no fluctuation.

[0045] The root mean square deviation of pH value is specifically:

[0046]

[0047] Where N is the total number of pH value data points, e i =PH 实际 -PH 目标 , represents the deviation of the i-th sample.

[0048] It reflects the degree of deviation between the actual value and the target value, measures the dynamic fluctuation of the pH value, quantifies the pH stability in the reactor, and prevents local over-acidity / over-alkalinity from causing crystallization defects. pH deviation can also lead to decreased reaction selectivity or the formation of by-products.

[0049] It should be noted that the reaction efficiency is calculated as follows:

[0050]

[0051] Where V is the reaction volume, k1, k2, and k3 are weight coefficients obtained through regression fitting of historical data to ensure the balance of the effects of various parameters on efficiency.

[0052] Based on the molar conservation principle of chemical reactions, C·V represents the actual amount of phosphate produced in the reaction system and is a core indicator reflecting the effective output of the reaction. Its calculation is based on the basic definition of substance purity, namely the ratio of the mass of a pure substance to the total mass.

[0053] k1·v represents the negative impact of the temperature change rate on the reaction efficiency, which is adjusted by the weight coefficient k1. Temperature fluctuations can lead to side reactions or uneven crystal growth, reducing the reaction efficiency;

[0054] k2·σ 2 The variance of pressure fluctuation reflects the reaction stability. High-pressure fluctuation will destroy the reaction equilibrium, and its inhibitory effect on efficiency needs to be quantified by variance.

[0055] The root mean square deviation of k3·RpH characterizes the dynamic changes of the acid-base environment and directly affects the reaction rate and product selectivity.

[0056] The interference of temperature, pressure and pH fluctuations on the reaction is comprehensively suppressed, and the model's generalization ability for complex working conditions is improved through adaptive weight distribution.

[0057] The specific product purity calculation model is:

[0058]

[0059] where ∑C i is the total concentration of impurity ions, P s The target pressure is set to a value, and a pressure variance correction term is introduced to suppress the impact of process fluctuations on purity.

[0060] It is the core calculation item of purity. Based on the law of conservation of mass, the purity of the substance is directly defined as the mass ratio of the active ingredient to the total ingredients.

[0061] The ratio of pressure variance to target pressure setting value is introduced to suppress the interference of pressure fluctuation on crystallization process. When the actual pressure fluctuation σ 2 The closer to the set value P s , the smaller the purity loss.

[0062] Based on the traditional purity calculation, a penalty term for pressure control deviation is added to avoid crystal defects caused by equipment vibration or flow abnormalities.

[0063] The total energy consumption is calculated as follows:

[0064]

[0065] P h is the heating power, P m is the stirring power, P p The material conveying power is calculated by integrating the real-time energy consumption sensor data. The integration form follows the law of conservation of energy and accumulates the instantaneous power data of the heating, stirring and conveying links in real time.

[0066] The total energy consumption is the time integral of the power of heating, stirring, conveying and other sub-items, which conforms to the cumulative calculation logic of energy consumption per unit product, that is, total energy consumption = the sum of energy consumption of each link.

[0067] Through real-time sensor data integration, the energy consumption distribution of each process stage can be accurately tracked, providing a basis for energy efficiency optimization.

[0068] S3. Data analysis: The reaction efficiency and product purity are integrated to obtain a comprehensive quality index; the energy efficiency evaluation model is calculated through total energy consumption measurement, and a dynamic production status matrix is constructed.

[0069] It should be noted that the comprehensive quality indicators are:

[0070]

[0071] in η ’ , P ′ The reaction efficiency η and purity P are normalized by Min-Max to eliminate dimensional differences and ensure fair weighting of multiple objective parameters.

[0072] It represents the ratio of actual energy consumption to benchmark energy consumption. Negative correction items reflect the constraints on energy consumption.

[0073] α, β, and γ are determined using the entropy weight method, and weights are dynamically assigned based on the information entropy of each indicator to avoid subjective bias and satisfy α+β+γ=1.

[0074] The three heterogeneous objectives of efficiency, purity and energy consumption are unified into a single evaluation indicator, supporting intelligent algorithms to quickly find the optimal solution.

[0075] The energy efficiency evaluation model is as follows:

[0076]

[0077] Among them, M p is the product quality, Q·M p It represents the product of comprehensive quality and output, reflects the total effective output, and conforms to the definition of "output value per unit of energy consumption" in industrial production.

[0078] E t For total energy consumption, the energy efficiency index EEI is constructed in the form of a ratio, which conforms to the general definition of energy utilization efficiency, that is, the ratio of useful energy output to total energy consumption, ensuring the consistency of the evaluation system.

[0079] Establish quantitative standards for energy utilization efficiency and drive the adjustment of process parameters towards high energy efficiency.

[0080] S4. Data output: The dynamic production status matrix is input into the artificial intelligence optimization model, and the production parameter adjustment instructions are output; the artificial intelligence optimization model is obtained through historical production data and corresponding optimal control parameter training, and integrates the quality prediction submodule and the energy consumption assessment submodule.

[0081] It should be noted that constructing a multidimensional time series matrix M(t)∈R N×T ,in:

[0082] The row vector contains the following feature dimensions:

[0083] Process parameters: temperature T, pressure P, pH value, original flow rate;

[0084] Quality indicators: Q, EEI,

[0085] Equipment status: stirring motor current, heater power;

[0086] The column vector is the time window span.

[0087] Quality prediction submodule: Receives the process parameters and equipment status data of the matrix M(t), and maps the fully connected layer to the predicted value of the quality indicator Q.

[0088] Energy consumption assessment submodule: uses the network to input the energy consumption time series data of the past c minutes and outputs the total energy consumption trend curve of the next d minutes. If the predicted energy consumption surge exceeds the threshold, an early warning signal is triggered.

[0089] S5. Feedback control: According to the adjustment instructions, the controller dynamically adjusts the reactor temperature, raw material addition rate and stirring intensity to achieve closed-loop control of phosphate production.

[0090] It should be noted that the model outputs control instructions, and its value range is constrained as follows:

[0091] Temperature adjustment: Modulate the on-off ratio of the solid-state relay and adjust the power of the heating rod within the range of -x≤ΔT≤+x, where the value of x is between 1℃ and 2℃ to prevent crystallization defects caused by sudden changes;

[0092] Raw material flow rate: The piezoelectric ceramic valve adjusts its opening according to Δv, and its range is ±y rated flow, where the value of y is between 8% and 10% to prevent material imbalance;

[0093] Stirring speed: The inverter drives the motor speed, the range of which is ±z baseline speed, where the value of z is between 15%-20% to prevent eddy current cavitation.

[0094] The closed-loop control of phosphate production is as follows:

[0095] Real-time monitoring: The dynamic matrix M(t) is updated every 30 seconds and the EEI is recalculated.

[0096] Abnormal self-recovery: If the EEI drops by more than a after three consecutive adjustments, where the value of a is between 5% and 7%, the model weight will be rolled back to the previous stable version. At the same time, the expert diagnosis mode will be triggered and the abnormal data will be recorded in the knowledge base.

[0097] Human-machine collaboration: The model outputs the confidence level (0-100%) for each instruction. Manual overwriting of low-confidence instructions (i.e., those with a confidence level less than b) can be performed by humans, switching to preset safety parameters. The value of b ranges from 80% to 90%. A visual dashboard displays EEI trends, equipment health, and quality prediction deviations in real time.

Claims

1. A phosphate production monitoring and control method based on artificial intelligence, characterized in that: The specific steps include: S1. Data acquisition: The sensor group collects multi-source data from the phosphate production process in real time, including reactor temperature data, air pressure data, pH value data, phosphate concentration data, impurity ion concentration data, and energy consumption data; S2. Data processing: Process the collected multi-source data once to obtain the temperature change rate, pressure fluctuation variance, and root mean square deviation of pH value, and calculate the reaction efficiency, product purity and total energy consumption; S3. Data Analysis: Integrate reaction efficiency and product purity to obtain comprehensive quality indicators; calculate energy efficiency evaluation models through total energy consumption measurement and construct a dynamic production status matrix; S4. Data output: The dynamic production state matrix is input into the artificial intelligence optimization model, and the production parameter adjustment instructions are output; the artificial intelligence optimization model is obtained by training historical production data and corresponding optimal control parameters, and integrates the quality prediction submodule and the energy consumption assessment submodule; S5. Feedback control: According to the adjustment instructions, the controller dynamically adjusts the reactor temperature, raw material addition rate and stirring intensity to achieve closed-loop control of phosphate production.

2. The artificial intelligence-based phosphate production monitoring and control method according to claim 1, wherein: The multi-source data collection is specifically as follows: A PT100 temperature sensor collects reaction temperature data, a piezoelectric pressure sensor collects air pressure data, a glass electrode pH meter detects pH data in real time, a UV spectrophotometer analyzes phosphate concentration data online, an ion-selective electrode array detects impurity ion concentration data, a Hall current sensor collects stirring motor power consumption data, and a steam flow meter is used to calculate heat energy consumption data. The energy consumption data includes heating power, stirring power and material conveying power.

3. The artificial intelligence-based phosphate production monitoring and control method according to claim 1, wherein: The process of processing multi-source data once is as follows: The temperature change rate is specifically: Where n represents the total number of temperature data, Represents the rate of temperature change at any time, T i -T i-1 It represents the temperature change, which is the temperature difference between adjacent times, and Δt represents the time interval, that is, the time difference between two temperature measurements; Characterizes the severity of temperature changes over time and reflects the thermodynamic stability of the process; The pressure fluctuation variance is specifically: in m represents the total amount of pressure data, X i is any pressure data; Quantify the discrete degree of pressure parameters to reflect the mechanical stability of the reaction system; The larger the variance, the more severe the pressure fluctuation; the smaller the variance, the more stable the pressure; if the variance is 0, it means that all pressure data are exactly the same and there is no fluctuation; The root mean square deviation of the pH value is specifically: Where N is the total number of pH value data points, e i =PH 实际 -PH 目标 , represents the deviation of the i-th sample; It reflects the degree of deviation between the actual value and the target value, measures the dynamic fluctuation of the pH value, and quantifies the pH stability in the reactor.

4. The artificial intelligence-based phosphate production monitoring and control method according to claim 3, wherein: The reaction efficiency is calculated as follows: Where V is the reaction volume, k1, k2, k3 are weight coefficients obtained by regression fitting of historical data; C·V represents the actual amount of phosphate generated in the reaction system and is the core indicator reflecting the effective output of the reaction; k1·v represents the negative impact of the temperature change rate on the reaction efficiency, which is adjusted by the weight coefficient k1. Temperature fluctuations can lead to side reactions or uneven crystal growth, reducing the reaction efficiency; k2·σ 2 Pressure fluctuation variance reflects reaction stability. High pressure fluctuation will destroy the reaction equilibrium, and its inhibitory effect on efficiency needs to be quantified by variance. The root mean square deviation of k3·RpH characterizes the dynamic changes of the acid-base environment and directly affects the reaction rate and product selectivity; The product purity calculation model is specifically: where ∑C i is the total concentration of impurity ions, P s Set the target pressure value; It is the core calculation item of purity and directly adopts the definition of material purity, that is, the mass ratio of active ingredients to total ingredients; The ratio of pressure variance to target pressure setting value is introduced to suppress the interference of pressure fluctuation on crystallization process. When the actual pressure fluctuation σ 2 The closer to the set value P s , the smaller the purity loss; The total energy consumption is calculated as follows: P h is the heating power, P m is the stirring power, P p The material conveying power is calculated by integrating the real-time energy consumption sensor data; The total energy consumption is the time integral of the power of heating, stirring, conveying and other sub-items, which conforms to the cumulative calculation logic of energy consumption per unit product, that is, total energy consumption = the sum of energy consumption of each link.

5. The artificial intelligence-based phosphate production monitoring and control method according to claim 1, wherein: The dynamic production status matrix is specifically: Construct a multidimensional time series matrix M(t)∈R N×T ,in: The row vector contains the following feature dimensions: Process parameters include temperature T, pressure P, pH value and original flow rate; quality indicators include Q, EEI and The device status includes the stirring motor current and heater power; the column vector is the time window span; Quality prediction submodule: Receives process parameters and equipment status data of the matrix M(t), and maps the fully connected layer to the predicted value of the quality indicator Q; Energy consumption assessment submodule: uses the network to input the energy consumption time series data of the past c minutes and outputs the total energy consumption trend curve of the next d minutes. If the predicted energy consumption surge exceeds the threshold, an early warning signal is triggered.

6. The artificial intelligence-based phosphate production monitoring and control method according to claim 1, wherein: The controller is used to dynamically adjust the reactor temperature, raw material addition rate and stirring intensity as follows: Temperature adjustment: Modulate the on-off ratio of the solid-state relay and adjust the power of the heating rod within the range of -x≤ΔT≤+x, where the value of x is between 1℃ and 2℃; Raw material flow rate: The piezoelectric ceramic valve adjusts its opening according to Δv, which is within the range of ±y rated flow, where the value of y is between 8% and 10%; Stirring speed: The speed of the inverter-driven motor, which ranges from ±z baseline speed, where the value of z is between 15% and 20%.

Images