Corrosion-resistant turbine vacuum system for wet-process phosphoric acid production and control method thereof

Abstract

The application discloses a kind of corrosion-resistant turbine vacuum systems for wet-process phosphoric acid production and control method thereof, it is related to wet-process phosphoric acid production vacuum technical field, system is by front pretreatment system, waste heat recovery and temperature control system, corrosion-resistant turbine fan and intelligent control system constitute, and intelligent control system includes acoustic perception subsystem, multiscale acoustic sparse reconstruction module, microvibration monitoring subsystem, time sequence polarization analysis module, state coupling module and self-aware control kernel.Formation fan health probability cloud picture data and vacuum behavior prediction data by reconstructing three-phase flow space state field and structure state field, self-aware control kernel linkage adjusts turbine fan speed, heat exchange temperature and spraying intensity, realize filtering vacuum degree stability, inhibit scale and cavitation and extend impeller life, give consideration to energy consumption reduction and operation reliability improvement.

Description

Technical Field

[0001] This invention relates to the field of vacuum technology for wet-process phosphoric acid production, specifically a corrosion-resistant turbine vacuum system and its control method for wet-process phosphoric acid production. Background Technology

[0002] In the wet-process phosphoric acid production, phosphate rock reacts with sulfuric acid to generate a slurry containing solid particles, fluorinated acid mist, and various impurities. Solid-liquid separation is then achieved through a vacuum filtration process. To maintain a stable pressure differential during filter cake formation and filtrate discharge, the production unit typically incorporates a vacuum system at the filtration end. This system draws the fluorinated and acidic tail gas from the filtration equipment, gas-liquid separator, and scrubbing tower before discharging it into subsequent treatment stages. This type of tail gas is characterized by high temperature, high water content, high acid mist concentration, fine and hard solid particles, and complex composition. The vacuum equipment and pipelines operate under conditions of strong corrosion, multiphase flow erosion, and scaling. Therefore, the vacuum system must provide sufficient suction capacity while also ensuring corrosion resistance, anti-scaling capabilities, and long-term stable operation.

[0003] Existing production lines widely employ vacuum units consisting of water ring vacuum pumps, steam jet devices, or Roots vacuum pumps connected in series with water ring vacuum pumps. These are supplemented with corrosion-resistant lining materials and simple washing and demisting structures. Pressure transmitters are placed at key locations in the vacuum pipeline, and traditional PID closed-loop control is used to maintain the filtration vacuum within a given range. Some units have vibration monitoring points installed on the vacuum pump or fan base for over-limit alarms and shutdown protection. Through these methods, existing technology can, to a certain extent, meet the basic requirements for vacuum and safety in wet-process phosphoric acid plants.

[0004] However, the aforementioned existing technologies have significant limitations. First, vacuum monitoring based on single-point pressure signals cannot reflect the spatial distribution of gas, droplets, and solid particles within the vacuum pipeline. It cannot identify droplet clusters, high-scouring zones of solid particles, or areas with abnormally high local liquid content near the gas-liquid interface, making it difficult to detect potential scaling and erosion problems in a timely manner. Vacuum fluctuations are often only detected when the problem has progressed to a more severe stage. Second, conventional vibration monitoring mostly uses overall amplitude or simple spectrum analysis, which is highly sensitive to strong background noise and operating disturbances. It cannot stably extract subtle features related to material fatigue from high-order modal time series, making it difficult to assess the fatigue risk and potential crack location of corrosion-resistant turbine impellers early. Equipment operation relies more on experience and regular maintenance. Third, traditional PID control usually only uses the vacuum deviation as the control variable, lacking the support of the three-phase flow spatial state field and structural state field. The control strategy cannot simultaneously consider multiple objective constraints such as vacuum stability, impeller life, and scaling control, easily leading to problems such as frequent adjustments, large vacuum fluctuations, increased local scaling, or fan overload. On the other hand, existing technologies lack self-sensing control and self-learning mechanisms, and relevant parameters mainly rely on manual tuning. Under conditions of raw material grade fluctuations, changes in operating load, and equipment aging, it is difficult to adapt to new operating conditions in a timely manner, and it is difficult to continuously optimize operational reliability and energy efficiency.

[0005] Therefore, there is an urgent need for a corrosion-resistant turbine vacuum system and its control method for wet-process phosphoric acid production that integrates an acoustic sensing subsystem, a multi-scale acoustic sparse reconstruction module, a micro-vibration monitoring subsystem, a time-series polarization analysis module, a state coupling module, and a self-sensing control kernel. By constructing a three-phase flow spatial state field and a structural state field, the system generates fan health probability cloud map data and vacuum behavior prediction data. Under complex multiphase corrosion conditions, it enables precise perception of vacuum behavior and equipment health status, as well as coordinated control under multi-objective constraints, fundamentally improving vacuum stability, suppressing scaling and cavitation, and extending impeller life. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings of the prior art and propose a corrosion-resistant turbine vacuum system and its control method for wet-process phosphoric acid production, so as to solve the above-mentioned problems.

[0007] The objective of this invention is achieved through the following technical solution: a corrosion-resistant turbine vacuum system for wet-process phosphoric acid production, comprising a pretreatment system connected to the wet-process phosphoric acid filtration process, a waste heat recovery and temperature control system, a corrosion-resistant turbine fan, and an intelligent control system connected to the aforementioned components for signal and control connections. The intelligent control system includes an acoustic sensing subsystem, a multi-scale acoustic sparse reconstruction module, a micro-vibration monitoring subsystem, a temporal polarization analysis module, a state coupling module, and a self-sensing control kernel. The acoustic sensing subsystem is configured in the pretreatment system and the corrosion-resistant turbine fan... The system is connected to the vacuum pipeline corresponding to the turbine inlet and signal-connected to the multi-scale acoustic sparse reconstruction module. The multi-scale acoustic sparse reconstruction module is configured to reconstruct the distribution of gas phase, droplets, and solid particles within the vacuum pipeline under sparse constraints and non-uniform sampling weights to form a three-phase flow spatial state field. It also sets weighting coefficients for the micro-vibration monitoring data based on this three-phase flow spatial state field. The micro-vibration monitoring subsystem is installed on the corrosion-resistant turbine fan casing and signal-connected to the time-series polarization analysis module. The time-series polarization analysis module is configured to perform micro-vibration monitoring under the constraints of the three-phase flow spatial state field. The subsystem collects micro-vibration signals, performs modal decomposition to obtain modal time series, and performs time-series polarization transformation to filter structural information channels and construct a structural state field characterizing impeller stress, fouling initiation, and cavitation risk. This structural state field is then fed back to the multi-scale acoustic sparse reconstruction module to adjust the sparsity weights. The state coupling module is signal-connected to the multi-scale acoustic sparse reconstruction module and the time-series polarization analysis module, and is configured to calculate wind turbine health probability cloud map data and vacuum behavior prediction data within a preset time window based on the three-phase flow spatial state field and the structural state field. The self-induction... The control kernel is signal-connected to the state coupling module and connected to the pretreatment system, the waste heat recovery and temperature control system, and the corrosion-resistant turbine fan control. It is configured to generate control commands based on the fan health probability cloud map data and the vacuum behavior prediction data to adjust the speed of the corrosion-resistant turbine fan, the heat exchange temperature of the waste heat recovery and temperature control system, and the spray intensity and working status of the pretreatment system. This is to simultaneously achieve stable filtration vacuum, impeller life meeting predetermined requirements, and suppression of scaling and cavitation development under fluctuating load conditions in the wet phosphoric acid filtration process.

[0008] After each update of the three-phase flow spatial state field, the multi-scale acoustic sparse reconstruction module sends information representing droplet cluster regions, high-concentration solid particle regions, and local liquid content anomaly regions in the three-phase flow spatial state field to the temporal polarization analysis module. The temporal polarization analysis module assigns prior structural information weights to the modal time series of the corresponding time period based on the information, and after constructing the structural state field, feeds back information on impeller high-stress regions and high-risk fouling initiation regions to the multi-scale acoustic sparse reconstruction module to adjust the sparse weights of the corresponding spatial units. Thus, a closed-loop iteration is formed through the bidirectional correction between the three-phase flow spatial state field and the structural state field.

[0009] The non-uniform sampling weights used in the multi-scale acoustic sparse reconstruction module are determined by a combination of acoustic propagation sensitivity, fluid turbulence intensity in the pre-processing system, and vacuum pipeline structural characteristics. This results in sensors in the acoustic sensing subsystem near the gas-liquid separation section, near the gas-liquid interface, and in areas prone to condensation having a greater weight in the reconstruction of the three-phase flow spatial state field than other sensors.

[0010] The multi-scale acoustic sparse reconstruction module is configured to hierarchically expand the acoustic field signal acquired by the acoustic sensing subsystem on a multi-scale acoustic basis with scale index and spatial index, and apply penalty weights to the high-scale coefficients to enhance the regions corresponding to local strong impact sound and droplet cluster sound in the three-phase flow spatial state field, and to explicitly identify the droplet cluster region and the solid particle impact region.

[0011] The micro-vibration monitoring subsystem is configured to perform multimodal decomposition on the micro-vibration signals collected by the vibration sensor to distinguish the impeller speed-related fundamental frequency mode, the mode generated by fluid excitation, and the higher-order mode related to material fatigue. The time-series polarization analysis module only performs the time-series polarization transformation on the time series of the higher-order modes related to material fatigue and gives the fatigue risk index of each impeller region in the structural state field.

[0012] When performing the time-series polarization transformation, the temporal polarization analysis module groups the higher-order mode time series into channels of different polarization layers according to the time sequence. By calculating the structural information reliability index of each channel, channels with reliability higher than a set threshold are marked as structural information channels, and channels with reliability lower than the threshold are marked as noise channels. The structural state field is constructed only based on the structural information channels.

[0013] The state coupling module is configured to map the three-phase flow spatial state field and the structural state field into a multi-dimensional system state vector including vacuum degree, local liquid content, local particle concentration, impeller stress level and fouling initiation probability. The multi-dimensional system state vector is accumulated within a preset time window to form a multi-dimensional system state sequence, and the fan health probability cloud map data and vacuum behavior prediction data are calculated based on the multi-dimensional system state sequence.

[0014] The self-sensing control kernel is configured to generate a feedforward control quantity based on the vacuum behavior prediction data, and to superimpose the feedforward control quantity with the feedback control quantity calculated based on the current filtration vacuum deviation to form a comprehensive control quantity. The comprehensive control quantity is used to drive the frequency converter of the corrosion-resistant turbine fan to adjust the fan speed and simultaneously adjust the heat exchanger temperature of the waste heat recovery and temperature control system and the spray opening sequence of the pretreatment system. When generating the comprehensive control quantity, the self-sensing control kernel uses the upper limit of impeller fatigue failure probability, the upper limit of scale germination probability, and the upper limit of preset vacuum fluctuation amplitude in the fan health probability cloud map data as constraints.

[0015] The self-sensing control kernel includes a self-learning unit, which is configured to associate and store the wind turbine health probability cloud map data, the vacuum behavior prediction data, and actual fault events and maintenance records at the end of the operating cycle. In subsequent operating cycles, it updates the non-uniform sampling weight, the structural information channel reliability threshold, and the weight coefficients of the feedforward control quantity and the feedback control quantity online based on the associated storage results. When it detects that the three-phase flow space state field is missing, the structural state field is missing, or the state coupling module does not output the wind turbine health probability cloud map data and the vacuum behavior prediction data, it prohibits the output of steady-state operation control commands and only outputs protective shutdown control commands or load reduction operation control commands.

[0016] A control method for a corrosion-resistant turbine vacuum system in wet-process phosphoric acid production, applied to a corrosion-resistant turbine vacuum system in wet-process phosphoric acid production, includes steps S1 to S5. Step S1 involves acquiring acoustic field signals using an acoustic sensing subsystem installed in the pre-treatment system and on the vacuum pipeline corresponding to the inlet of the corrosion-resistant turbine fan, and inputting these signals into a multi-scale acoustic sparse reconstruction module. Step S2 involves the multi-scale acoustic sparse reconstruction module reconstructing the three-phase flow spatial state field under sparse constraints and non-uniform sampling weights, and setting weight coefficients for the micro-vibration monitoring subsystem data. Step S3 involves acquiring impeller micro-vibration signals using a micro-vibration monitoring subsystem installed on the casing of the corrosion-resistant turbine fan, and analyzing these signals in the three-phase flow using a time-series polarization analysis module. Under the constraint of the spatial state field, the temporal polarization transformation is performed to construct the structural state field and feed it back to the multi-scale acoustic sparse reconstruction module; in step S4, the state coupling module calculates the fan health probability cloud map data and the vacuum behavior prediction data based on the three-phase flow spatial state field and the structural state field; in step S5, the self-sensing control kernel generates control commands for linkage adjustment of the speed of the corrosion-resistant turbine fan, the heat exchange temperature of the waste heat recovery and temperature control system, and the spray intensity and working status of the pretreatment system, so as to maintain the filtration vacuum within the set range and suppress scaling, corrosion and cavitation development during the operation of the wet phosphoric acid filtration process.

[0017] The beneficial effects of this invention are: This invention introduces an acoustic sensing subsystem, a multi-scale acoustic sparse reconstruction module, and a micro-vibration monitoring subsystem into the vacuum side of the wet phosphoric acid filtration process. This constructs a three-phase flow spatial state field and a structural state field, transforming the previously invisible black-box distribution of gas phase, droplets, and solid particles within the vacuum pipeline, as well as impeller stress and fatigue risks, into visible and quantifiable spatial field information. This allows for accurate identification of droplet cluster regions, high-concentration solid particle regions, and high-stress impeller regions even under complex multiphase conditions, effectively suppressing scaling, erosion, and early impeller fatigue failure in the suction pipeline. The multi-scale acoustic sparse reconstruction module employs a multi-scale acoustic basis with scale and spatial indices, combined with non-uniform sampling weights, to sparsely reconstruct the acoustic field signal. Compared to traditional methods relying solely on single-point vacuum measurement, this improves spatial resolution near the gas-liquid interface and areas prone to condensation, providing a physically meaningful three-phase flow spatial state field input for subsequent control. The temporal polarization analysis module performs temporal polarization transformation on the high-order mode time series obtained by the micro-vibration monitoring subsystem, and constructs the structural state field by screening through the structural information channel. It can stably extract high-order vibration characteristics related to material fatigue under the conditions of time-related mechanical noise and operating condition disturbance, providing a reliable basis for the assessment of impeller fatigue risk index.

[0018] The state coupling module performs a unified mapping of the three-phase flow spatial state field and the structural state field, forming a multi-dimensional system state vector and its time series, which in turn generates wind turbine health probability cloud map data and vacuum behavior prediction data. Based on this, the self-sensing control kernel simultaneously generates feedforward control quantity and feedback control quantity, and introduces the upper limit of impeller fatigue failure probability and upper limit of scale initiation probability in the wind turbine health probability cloud map data and the vacuum degree fluctuation constraint in the vacuum behavior prediction data into the control decision. This enables the corrosion-resistant turbine wind turbine speed, waste heat recovery and temperature control system heat exchange temperature and pretreatment system spray intensity to achieve coordinated optimization under the premise of meeting the multi-objective constraints of vacuum stability, equipment life and scale control. Compared with conventional PID control based solely on vacuum degree closed loop, this can significantly reduce vacuum fluctuations and reduce the risk of unplanned downtime. The self-learning unit set in the self-sensing control kernel uses wind turbine health probability cloud map data, vacuum behavior prediction data, and fault and maintenance records to update the non-uniform sampling weight, structural information channel reliability threshold, and feedforward and feedback control weight coefficients online. This enables the multi-scale acoustic sparse reconstruction module, temporal polarization analysis module, and control strategy to automatically adjust parameters according to changes in raw material quality and equipment aging, achieving adaptive optimization under long-term operating conditions, thereby further improving the robustness and energy efficiency of the system. Attached Figure Description

[0019] Figure 1 The process of this invention Figure 1 ; Figure 2 The process of this invention Figure 2 ; Figure 3 The process of this invention Figure 3 . Detailed Implementation

[0020] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Example 1 like Figure 1As shown, in this embodiment, the corrosion-resistant turbine vacuum system for wet-process phosphoric acid production is arranged on the vacuum side of the wet-process phosphoric acid filtration process. The wet-process phosphoric acid filtration process is connected to the pre-treatment system via a vacuum pipeline. The pre-treatment system is connected to the inlet of the corrosion-resistant turbine blower via a vacuum pipeline. The exhaust side of the corrosion-resistant turbine blower is connected to the atmospheric side or the tail gas treatment device via a waste heat recovery and temperature control system to provide stable suction capacity under conditions of feed fluctuations, changes in filter cake thickness, and changes in filtrate viscosity. The pre-treatment system is internally equipped with a spray section, a demisting section, and a solid particle interception section for preliminary washing and demisting of fluorine-containing and acidic gases, reducing the droplet and solid particle load entering the corrosion-resistant turbine blower. The waste heat recovery and temperature control system includes a heat exchanger connected to the turbine blower exhaust and a heat exchange loop connected to the slurry pipeline of the wet phosphoric acid acidolysis process. By controlling the flow rate and temperature of the heat exchange medium, the system can regulate the exhaust temperature of the turbine blower and utilize the waste heat from the upstream process. The corrosion-resistant turbine blower uses a corrosion-resistant impeller and casing structure. The impeller is connected to the drive motor through a coupling, and the drive motor speed is controlled by a frequency converter.

[0022] The intelligent control system is located in the control room and is connected to the frequency converter, valve actuator, acoustic sensing subsystem, and micro-vibration monitoring subsystem via industrial Ethernet. The acoustic sensing subsystem is installed on the outer wall of the vacuum pipeline at the outlet section of the pre-treatment system and the inlet section of the corrosion-resistant turbine fan. Several acoustic sensors are arranged axially and circumferentially and are connected to the multi-scale acoustic sparse reconstruction module to continuously collect the sound field signals generated by the flowing medium within the vacuum pipeline. The multi-scale acoustic sparse reconstruction module treats the acoustic sampling data as a joint temporal and spatial signal. Through discretization using a preset time sampling period, the spatial distribution signal within each sampling period is sent to the multi-scale acoustic processing unit. Based on a pre-constructed multi-scale acoustic basis, and under the influence of sparse constraints and non-uniform sampling weights, the relative distribution information of gas phase, droplets, and solid particles on the spatial discrete unit of the vacuum pipeline is obtained, thus forming a three-phase flow spatial state field.

[0023] After completing the calculation of the three-phase flow spatial state field at each time step, the multi-scale acoustic sparse reconstruction module marks the droplet cluster region, the high-concentration solid particle region, and the region where the local liquid content is significantly higher than the average level, and sets the weighting coefficients for the micro-vibration monitoring data accordingly. These weighting coefficients correspond to the expected intensity of droplet impact and particle scouring experienced by the impeller at different time periods, and are used to adjust the contribution of data from different time periods in the time-series polarization analysis module. The micro-vibration monitoring subsystem is installed on the outer surface of the corrosion-resistant turbine fan casing. It collects micro-vibration signals through vibration sensors arranged in the axial and radial directions, and after local amplification and filtering, sends them to the time-series polarization analysis module at a fixed sampling frequency.

[0024] The temporal polarization analysis module receives micro-vibration data from the micro-vibration monitoring subsystem within each time window. It then uses a modal decomposition algorithm to decompose this data into fundamental frequency modes related to impeller speed, modes generated by fluid excitation, and higher-order modes related to material fatigue. In this embodiment, temporal polarization analysis is performed only on the time series of higher-order modes related to material fatigue. Based on the droplet cluster region and solid particle impact region provided by the three-phase flow spatial state field, the temporal polarization analysis module divides the higher-order mode time series into several time periods on the time axis. Higher prior structural information weights are assigned to the time periods corresponding to strong droplet cluster regions and strong particle impact regions, and these time periods are layered and combined according to a predetermined polarization structure to form channel signals. The structural information reliability index is calculated based on the stability of the channel signals, their sensitivity to stress changes, and noise levels. This index is compared with a set threshold, and channels with high reliability are designated as structural information channels, while the rest are designated as noise channels. Statistical analysis is performed solely based on the structural information channel to generate stress levels, fatigue accumulation indices, fouling initiation risks, and cavitation risks at discrete locations on the impeller. This constructs a structural state field, which is then returned to the multi-scale acoustic sparse reconstruction module. This module is used to adjust the sparsity weights of the acoustic features in subsequent time steps, thus establishing a closed-loop iterative relationship between the three-phase flow spatial state field and the structural state field.

[0025] The state coupling module maps the three-phase flow spatial state field and the structural state field into a multi-dimensional system state vector at each time step. The system state vector is defined on each spatial unit and represented by the following formula: in, Representing spatial units In time The local vacuum level (in Pa). This indicates the local liquid content of the unit (dimensionless, ranging from 0 to 1). This indicates the local particle concentration in the unit (in mg / m³). This indicates the stress level (in MPa) at the impeller or pipeline location where the unit is located. This represents the probability of scale formation in this unit (dimensionless, ranging from 0 to 1). The state coupling module accumulates the system state vectors of each spatial unit within a preset time window according to the time sampling step, forming a multidimensional system state sequence set. (in For the first (Sampling time points) are used to characterize the overall operating status of the turbine vacuum system within the specified time window. Based on the multidimensional system state sequence, the state coupling module constructs a probability distribution reflecting the equipment health level, evaluates the combined relationship between stress level, fatigue accumulation, and fouling initiation probability, generates fan health probability cloud map data, and combines the time variation trends of vacuum degree and local liquid content to extrapolate several time steps after the preset time window to obtain vacuum behavior prediction data for each spatial unit.

[0026] The self-sensing control kernel receives fan health probability cloud map data and vacuum behavior prediction data, and converts them into control commands for the corrosion-resistant turbine fan, waste heat recovery and temperature control system, and pretreatment system. In this embodiment, the self-sensing control kernel constructs a control input vector in each control cycle, represented as follows: in, For corrosion-resistant turbine fans to last for a period of time Target speed (in revolutions per minute). The target temperature at the outlet of the heat exchanger in the waste heat recovery and temperature control system (in °C). The spray intensity of the pretreatment system is a normalized value, ranging from 0 to 1. The self-sensing control kernel constructs an objective function based on the fan health probability cloud map data and vacuum behavior prediction data within the current time window. This function constrains the allowable fluctuation range of the filtration vacuum degree near the set value, while limiting the fatigue failure probability and fouling initiation probability of key impeller components to no more than the set upper limit. Based on this, the control input vector that minimizes the vacuum degree error is selected. The control relationship can be expressed as follows: in, This represents the parameter that makes the function reach its minimum value. The cost function is composed of the superposition of vacuum deviation term, impeller fatigue risk term, and fouling risk term. Let be the system state vector. To control the input vector, The time variable is used. The self-sensing control kernel uses the optimal control input vector obtained from the solution as the control command for the current control cycle. It adjusts the speed of the corrosion-resistant turbine fan by communicating with the frequency converter, controls the temperature of the heat exchanger by adjusting the heat exchange medium valve, and controls the spray intensity and spray sequence by adjusting the frequency of the spray pump and the opening degree of the spray valve. In this way, it can achieve stable vacuum, impeller life meeting the predetermined requirements, and controlled scaling and cavitation development under load fluctuations.

[0027] The control method executes data acquisition, data processing, state coupling, and control output sequentially. In step S1, the control system triggers the acoustic sensing subsystem to acquire sound field signals within each sampling period. The acquired data is scaled and filtered before being input into the multi-scale acoustic sparse reconstruction module. In step S2, the multi-scale acoustic sparse reconstruction module solves the three-phase flow spatial state field under sparse constraints and non-uniform sampling weights, and updates the weight coefficients of the micro-vibration monitoring data. In step S3, the micro-vibration monitoring subsystem acquires impeller micro-vibration signals. The temporal polarization analysis module performs temporal polarization transformation considering the constraints of the three-phase flow spatial state field, constructs the structural state field, and feeds it back to the multi-scale acoustic sparse reconstruction module. In step S4, the state coupling module summarizes the three-phase flow spatial state field and structural state field generated in multiple sampling periods within a preset time window, constructs a multi-dimensional system state vector and a multi-dimensional system state sequence, and calculates the wind turbine health probability cloud map data and vacuum behavior prediction data. In step S5, the self-sensing control kernel solves for the optimal control input vector based on the fan health probability cloud map data and vacuum behavior prediction data corresponding to the current time window, generates control commands, and performs linkage adjustment on the speed of the corrosion-resistant turbine fan, the heat exchange temperature of the waste heat recovery and temperature control system, and the spray intensity and working status of the pretreatment system. The new operating conditions are reflected in the data of the next sampling cycle through the acoustic sensing subsystem and the micro-vibration monitoring subsystem, forming a closed-loop control process.

[0028] Example 2 like Figure 1 and Figure 2 As shown, in this embodiment, based on the overall system structure described in Embodiment 1, the internal implementation of the acoustic sensing subsystem and the multi-scale acoustic sparse reconstruction module, the construction of non-uniform sampling weights, and the bidirectional correction process between the three-phase flow spatial state field and the structural state field are explained in detail.

[0029] The acoustic sensing subsystem arranges several acoustic sensors at intervals along the axial and circumferential directions of the vacuum pipeline from the outlet of the pretreatment system to the inlet of the corrosion-resistant turbine fan, prioritizing placement at gas-liquid separation points, near the gas-liquid interface, and in sections prone to condensation. Through on-site pipeline structure mapping, the vacuum pipeline is divided into several spatial units along its length, with each unit corresponding to several acoustic sensors or combinations thereof. The acoustic sensing subsystem synchronously acquires the sound field signal within the pipeline within a fixed time step. By index The raw acoustic signal collected by the sensor is Normalized acoustic sampling data were obtained after bandpass filtering and amplitude normalization. .

[0030] The multi-scale acoustic sparse reconstruction module constructs a set of multi-scale acoustic bases with scale index and spatial index. ,in For scale indexing, For spatial unit indexing, Let be the axial coordinate of the vacuum tube. The sound field within the vacuum tube is considered as a linear combination of acoustic bases at various scales. At a discrete time step, several multi-scale acoustic coefficients are used. With corresponding acoustic basis functions Weighted and represented. The acoustic sensing subsystem at each location. Collected from The acoustic coefficients and multi-scale acoustic coefficients satisfy an approximately linear relationship, which can be written in vector form as follows: in, The observation vector is composed of normalized sampled data from each sensor (dimensional: , (Total number of sensors) This is a coefficient vector composed of the acoustic coefficients of each spatial unit at each scale. For the reason The coefficient matrix formed For the first The discrete time step is used. Considering the spatial sparsity and local aggregation characteristics of the influence of gas phase, droplets, and solid particles on the sound field, and assuming that the high-scale acoustic coefficients of most spatial units are close to zero, with significant high-scale coefficients only appearing in droplet cluster regions and high-concentration solid particle regions, a sparse constraint reconstruction method is employed to solve the problem. .

[0031] To reflect the differences in contribution from different sensor locations, the multi-scale acoustic sparse reconstruction module introduces non-uniform sampling weights into the observation data. This is done for each acoustic sensor index. Define acoustic propagation sensitivity Fluid turbulence intensity in the pretreatment system and vacuum pipeline structural characteristic coefficients , The transmission gain from the sound source to the sensor location is obtained by calculating the linearized acoustic model. Calculated from the Reynolds number at the installation location and the amplitude of local velocity fluctuations. The weighting is calculated based on the shape, height difference, and cross-sectional variation of adjacent pipe components. The non-uniform sampling weighting is constructed as follows: in, For acoustic sensor indexing, For sensors Non-uniform sampling weights, The acoustic propagation sensitivity at this sensor location, The fluid turbulence intensity at the installation location. The characteristic coefficient of the vacuum pipeline structure. , , The weighting coefficients set during the system debugging phase.

[0032] Within the coefficient space, to highlight the ability of high-scale acoustic components to characterize droplet cluster sound and solid particle impact sound, while suppressing high-scale noise, a scale penalty matrix is ​​constructed. , where scale index The corresponding scale penalty coefficient is In the lower scale part Smaller, in the high-scale part The size is relatively large. A sparse reconstruction objective function is constructed by combining non-uniform sampling weights and scale penalties, and at each time step... Solution: in, This represents the parameter that makes the function reach its minimum value. The scaling penalty matrix, For the coefficient vector Norm, for Norm, It is a diagonal weight matrix. To balance sparsity and fitting error, a regularization parameter is used. For the first Each discrete time step.

[0033] To identify droplet cluster regions, high-concentration solid particle regions, and regions with anomalous local liquid content in the three-phase flow spatial state field, the multi-scale acoustic sparse reconstruction module calculates characteristic indicators such as high-scale acoustic energy, local liquid content, and local particle concentration for each spatial cell and sets corresponding thresholds. Spatial cells with high-scale energy exceeding the threshold are marked as droplet cluster regions, cells with local particle concentration exceeding the threshold are marked as high-concentration solid particle regions, and cells with local liquid content significantly higher than the long-term average and exceeding the incremental threshold are marked as regions with anomalous local liquid content. The multi-scale acoustic sparse reconstruction module sends the spatial and temporal distribution of these regions and their corresponding time indices to the temporal polarization analysis module.

[0034] After receiving the regional information, the temporal polarization analysis module converts it into prior weights for the modal time series of the micro-vibration monitoring subsystem. The time series of higher-order modes of impeller micro-vibration is divided into several time segments on the time axis, each corresponding to a period of operation. The projections of the droplet cluster region, the high-concentration solid particle region, and the local abnormal liquid content region on time are matched with the higher-order mode time series to obtain the intensity indices of droplet clustering and particle impact in each time segment. The temporal polarization analysis module defines prior weights for structural information for each time segment. The higher value is taken during the time period that includes the above-mentioned effects, and the lower value is taken during other time periods. This is also considered when dividing channels and filtering structural information channels. By introducing channel reliability calculation, the contribution of data from high-risk time periods to the structural state field becomes higher.

[0035] After the structural state field is constructed, the temporal polarization analysis module statistically analyzes the stress levels and fatigue accumulation indices at discrete circumferential and radial locations of the impeller, defines high-stress regions and high-risk regions for fouling initiation, and feeds their spatial indices and risk indices back to the multi-scale acoustic sparse reconstruction module. In each spatial unit... Define sparse weight coefficients at the point This is used to adjust the sparse penalty intensity of the space unit when the high-stress area of ​​the impeller or the high-risk area for fouling initiation covers the space unit. At that time, the sparsity weight of the spatial unit is increased according to the risk index, and the update formula is as follows: in, For spatial unit indexing, These are the basic sparse weights when no structural state field feedback is obtained. For the updated sparse weights, This is the stress risk index corresponding to this space unit. As an indicator of the risk of scale formation, , This is the weighting amplification factor.

[0036] Through the above mechanism, within each control cycle, the three-phase flow spatial state field is constructed in the multi-scale acoustic sparse reconstruction module through non-uniform sampling weights and scale penalties. The temporal polarization analysis module uses the high-risk region information provided by this state field to assign prior structural information weights to the modal time series, thereby constructing the structural state field. The high-stress and high-risk regions identified by the structural state field then react on the multi-scale acoustic sparse reconstruction module through the sparse weight update mechanism, making the reconstruction of the three-phase flow spatial state field in subsequent time steps more concentrated in these regions, forming a closed-loop iterative process of bidirectional correction.

[0037] Example 3 like Figures 1 to 3 As shown, in this embodiment, based on the systems described in Embodiments 1 and 2, the specific implementation of the micro-vibration monitoring subsystem, the time-series polarization analysis module, and the self-sensing control kernel is explained in detail. The multi-modal decomposition method of micro-vibration signals, the structural information channel screening method, and the control law and self-learning mechanism based on health probability cloud map and vacuum behavior prediction are given.

[0038] The micro-vibration monitoring subsystem is installed at the axial and radial positions of the corrosion-resistant turbine fan casing. The vibration sensors employ a corrosion-resistant encapsulation structure and are rigidly connected to the casing via threads or welding. The micro-vibration monitoring subsystem acquires vibration signals at a fixed sampling frequency, forming a time series containing multiple sampling points within each sampling period. This time series is represented as a sequence of length [length missing] within a time window. Discrete time series (in For sampling point index, , The time window length, (Sampling interval). This time series is sent to the multimode decomposition unit inside the time-series polarization analysis module, where it is decomposed into impeller speed-related fundamental frequency modes, fluid-excited modes, and higher-order modes related to material fatigue through spectral analysis and narrowband filtering. Based on the actual speed of the corrosion-resistant turbine fan... Calculate the angular frequency (unit: revolutions per minute). (Unit: radians / second), in the frequency domain Perform Discrete Fourier Transform to locate the position. The fundamental frequency components and their low-order harmonics are centered on this, and a narrowband filter is designed in conjunction with the fluid excitation frequency band. The set of high-order frequency bands related to material fatigue is denoted as... The time-domain components within this frequency band are inversely transformed to obtain the higher-order mode time series. Within a time window, the high-order mode time series from different measurement points are combined into a vector form. Each dimension corresponds to a measuring point on the impeller or casing.

[0039] The time-series polarization analysis module obtains Then, the higher-order mode time series within the time window are grouped and polarized according to time order. Polarization channels can be constructed by combining adjacent samples, and recursively combined in a multi-layer polarization structure to obtain multiple channel signals. All channel signals are uniformly numbered as follows: (in For channel indexing, Indicates channel At the sampling point (Higher-order mode values ​​at the location). To evaluate the ability of each channel to represent structural information, a structural information reliability index is introduced for each channel. The reliability within the current time window is denoted as . It can be composed of channel variance, correlation coefficient between channel and impeller stress level or fouling initiation probability sequence, and noise variance estimate, etc. The time-series polarization analysis module uses structural information and channel reliability thresholds. Filter each channel, when At that time, the channel Marked as a structural information channel, when At that time, the channel Marked as a noise channel, only the set of structure information channels is used. Construct the structural state field.

[0040] After channel filtering is completed, the time-series polarization analysis module reconstructs the structural information channel signals according to the impeller spatial discrete units. Based on the energy and accumulated cyclic stress levels of higher-order modes at different spatial locations, fatigue damage indices are calculated, and each impeller discrete location is indexed as follows: The spatial unit performs a weighted average of the higher-order modal energies corresponding to all structural information channels within the current time window, assigning greater weight to channels with higher reliability to obtain the fatigue risk index distribution of each region of the impeller. This distribution is integrated with information on fouling initiation risk and cavitation risk to form a structural state field. The spatial index and risk value of high-risk regions are then returned to the multi-scale acoustic sparse reconstruction module to enhance the acoustic resolution of these regions.

[0041] After the state coupling module completes the fusion of the three-phase flow spatial state field and the structural state field, it obtains the fan health probability cloud map data and vacuum behavior prediction data. The former gives the spatial distribution of the impeller fatigue failure probability and fouling initiation probability of each spatial unit in the future, while the latter gives the predicted time series of the overall filtration vacuum degree and local vacuum fluctuations. Based on these data, the self-sensing control kernel constructs feedforward control and feedback control quantities, and superimposes them to form a comprehensive control quantity, which is used to drive the corrosion-resistant turbine fan frequency converter, the heat exchanger of the waste heat recovery and temperature control system, and the spray actuator of the pretreatment system.

[0042] The self-sensing control kernel uses the corrosion-resistant turbine fan speed, heat exchanger outlet temperature, and spray intensity as control input vectors. ,in This is an index for the discrete control cycle. Based on vacuum behavior prediction data, the state coupling module provides a prediction sequence of the filtered vacuum degree within several prediction steps. The self-sensing control kernel will then compare the predicted vacuum degree with the setpoint. Comparison, constructing the feedforward control vector That is, through the feedforward gain matrix The predicted vacuum deviation is mapped as a feedforward correction for fan speed, heat exchanger outlet temperature, and spray intensity. Meanwhile, the self-sensing control kernel is based on the currently measured filtration vacuum. With set value The deviation between them constructs a feedback control vector. Through the feedback gain matrix The current vacuum deviation is converted into a correction to the control input. Under the constraints of impeller fatigue failure probability, fouling initiation probability, and vacuum fluctuation amplitude not exceeding a preset upper limit, the feedforward control quantity and the feedback control quantity are superimposed to obtain the comprehensive control quantity. It is used to generate execution instructions for frequency converters, heat exchangers, and spray systems.

[0043] To enable the system to adapt to changes in raw material quality, operating conditions, and equipment aging during long-term operation, a self-learning unit is incorporated into the self-sensing control core. This unit updates the non-uniform sampling weights, structural information channel reliability thresholds, and weight coefficients of feedforward and feedback control variables online. At the end of each operating cycle, the self-learning unit reads the fan health probability cloud map data, vacuum behavior prediction data, and actual fault events and maintenance records for that cycle, forming training samples through timestamps and spatial location correlation. For the non-uniform sampling weights, the self-learning unit adjusts the weights of the corresponding acoustic sensors based on the actual impeller damage and fouling locations, using the following update formula: in, For acoustic sensor indexing, The sampling weights are used at the beginning of the current operating cycle. The sampling weights are updated after the period ends. The learning rate for updating the sampling weights (usually taken as...) ), For position The associated actual degree of damage or scaling, To be based on the current model at the location The predicted damage or scaling indicators. For the structural information channel reliability threshold. The self-learning unit performs gradient adjustment by statistically analyzing the contributions of the structural information channel and the noise channel to the true positive rate and false positive rate of fault warning over a period of time, and records the target true positive rate as... The actual true positive rate is recorded as The following update relationship is adopted: in, and These represent the reliability thresholds of the structural information channels before and after the update. The learning rate updated for the threshold (usually taken as...) ), The target true positive rate is set in advance. This represents the actual true positive rate calculated based on historical early warning results within the current statistical time window. Regarding the weights of the feedforward and feedback control variables, if the system's response to vacuum disturbances is consistently slow or exhibits significant overshoot over a long period, the self-learning unit fine-tunes the feedforward gain based on the observed overshoot, and the maximum vacuum overshoot corresponding to the current control cycle is recorded as... The update relationship is as follows: in, and These represent the feedforward gain matrices before and after the update (their dimensions are the same as the control input vector's dimension). The learning rate for updating the feedforward gain (usually taken as...) ), This is an index for discrete control cycles. In the control cycle The maximum overshoot of the filtration vacuum obtained from internal statistics. To and An identity matrix with matching dimensions. Through the above update mechanism, the self-learning unit gradually corrects the sampling weights, the reliability threshold of the structural information channel, and the magnitude of the feedforward gain without changing the basic control structure, enabling the system to maintain its adaptability to fault risks and vacuum dynamic characteristics during long-term operation.

[0044] In terms of operational safety, the self-learning unit also monitors the integrity of the three-phase flow spatial state field, structural state field, and the fan health probability cloud map data and vacuum behavior prediction data output by the state coupling module. When any of the above data fails to update successfully or is missing in a certain control cycle, the self-sensing control kernel triggers protection logic, prohibits the output of control commands for long-term steady-state operation, and switches the control mode to protective shutdown mode or load reduction operation mode. In protective shutdown mode, the fan speed is set to a safe low speed or shutdown state, the heat exchanger load is reduced, and the spray intensity of the pre-treatment system is reduced; in load reduction operation mode, the minimum production load is maintained under safety constraints, and the normal self-sensing control mode is restored after the sensing data returns to normal. Through the above mechanism, the micro-vibration monitoring subsystem, the time-series polarization analysis module, and the self-sensing control kernel form a closed-loop link from high-order modal micro-vibration data to structural state sensing and then to comprehensive control execution, and the self-learning unit realizes long-term adaptive adjustment of control parameters and sensing weights.

[0045] Example 4 In this embodiment, a comparative and ablation experimental platform was built on the same wet-process phosphoric acid filtration production line. The filtration area, raw material grade, slurry temperature, filtration vacuum setpoint, shift organization, and maintenance rules were kept consistent; only the vacuum system structure and control strategy were changed. Quantifiable data were collected through long-term continuous operation to verify the overall improvement of this invention in three indicators: vacuum stability, impeller life, and scaling control, and to verify the necessity of multi-scale acoustic sparse reconstruction, temporal polarization analysis, and self-sensing control.

[0046] In the comparative experiment, the control group used a commonly used liquid ring vacuum pump with a pressure sensor and conventional PID control, without setting up an acoustic sensing subsystem, three-phase flow spatial state field, structural state field, fan health probability cloud map data, and vacuum behavior prediction data, relying solely on single-point feedback control based on filtration vacuum degree. The experimental group used the corrosion-resistant turbine vacuum system and intelligent control system described in Examples 1-3, recording key indicators under the same filtration capacity and operating time conditions. The comparative experiment period was 6000 hours of continuous operation, and the average data from three parallel production lines was used for each group. The data are shown in Table 1.

[0047] Table 1. Comparison data between conventional liquid ring vacuum systems and the corrosion-resistant turbine vacuum system of this invention.

[0048] As shown in Table 1, this invention achieves better performance in vacuum stability (fluctuation ±2.1 kPa, exceeding limit by 0.6%), lifespan (12800 h vs 8500 h), scaling (0.25 vs 0.65 mm / 1000h, less unplanned downtime), and energy consumption (15.6 vs 18.2 kWh / t). It is significantly superior to conventional liquid ring vacuum systems in terms of performance, meeting process constraints while also considering energy efficiency.

[0049] To verify the necessity of the key modules, multiple ablation schemes were designed. The technical requirements were defined as: S (vacuum fluctuation ≤ ±3 kPa and over-limit percentage ≤ 2%), T (lifespan prediction ≥ 12,000 h), and Z (scale growth ≤ 0.35 mm / 1000 h and scale-related shutdowns ≤ 1 time within 6000 h). Without changing the process conditions, only the module configuration of the intelligent control system was adjusted, resulting in four control schemes. Each scheme was run for 6000 hours on two parallel production lines. The results are shown in Table 2.

[0050] Table 2. Ablation experimental results of multi-scale acoustic sparse reconstruction, temporal polarization analysis, and self-sensing control kernel.

[0051] The results in Table 2 are summarized as follows: A (all modules) simultaneously satisfy S / T / Z; B (no three-phase flow spatial state field) S and Z are not satisfied; C (no structured state field) T is not satisfied; D (no self-sensing control, only PID) S is not satisfied. The absence of any module will result in at least one indicator failing to meet the requirements.

[0052] In summary, both comparative and ablation experiments demonstrate that the three-phase flow spatial state field, structural state field, and self-sensing control kernel respectively support vacuum stability, lifetime, and fouling control; multi-objective constraints require the coordinated achievement of all three; the absence of any module will result in at least one indicator failing to be met. The overall solution of this invention exhibits significant comprehensive advantages compared to conventional systems.

[0053] The above description is merely a preferred embodiment of the present invention. It should be understood that the present invention is not limited to the forms disclosed herein and should not be construed as excluding other embodiments. It can be used in various other combinations, modifications, and environments, and can be altered within the scope of the concept described herein through the above teachings or related technologies or knowledge. Modifications and variations made by those skilled in the art that do not depart from the spirit and scope of the present invention should be within the protection scope of the appended claims.

Claims

1. A corrosion-resistant turbine vacuum system for wet-process phosphoric acid production, characterized in that, The system includes a pretreatment system connected to the wet-process phosphoric acid filtration process, a waste heat recovery and temperature control system, a corrosion-resistant turbine fan, and an intelligent control system that is connected to the aforementioned components for signal and control. The intelligent control system includes an acoustic sensing subsystem, a multi-scale acoustic sparse reconstruction module, a micro-vibration monitoring subsystem, a time-series polarization analysis module, a state coupling module, and a self-sensing control kernel. The acoustic sensing subsystem is located on the vacuum pipelines corresponding to the inlets of the pretreatment system and the corrosion-resistant turbine fan and is connected to the multi-scale acoustic sparse reconstruction module for signal transmission. The multi-scale acoustic sparse reconstruction module is configured to reconstruct the distribution of gas phase, droplets, and solid particles within the vacuum pipeline under sparse constraints and non-uniform sampling weights to form a three-phase flow spatial state field. It then sets weighting coefficients for the micro-vibration monitoring data based on this three-phase flow spatial state field. The micro-vibration monitoring subsystem is mounted on the corrosion-resistant turbine fan casing and connected to the time-series polarization analysis module. The time-series polarization analysis module is configured to perform modal decomposition on the micro-vibration signals acquired by the micro-vibration monitoring subsystem under the constraints of the three-phase flow spatial state field. The temporal polarization transformation is performed on the state-time series to filter structural information channels and construct a structural state field characterizing impeller stress, fouling initiation, and cavitation risk. The structural state field is then fed back to the multi-scale acoustic sparse reconstruction module to adjust the sparsity weights. The state coupling module is signal-connected to the multi-scale acoustic sparse reconstruction module and the temporal polarization analysis module. It is configured to calculate the fan health probability cloud map data and vacuum behavior prediction data within a preset time window based on the three-phase flow spatial state field and the structural state field. The self-sensing control kernel is signal-connected to the state coupling module and connected to the pre-processing system, the waste heat recovery and temperature control system, and the corrosion-resistant turbine fan control system. It is configured to generate control commands based on the fan health probability cloud map data and vacuum behavior prediction data to coordinate and adjust the speed of the corrosion-resistant turbine fan, the heat exchange temperature of the waste heat recovery and temperature control system, and the spray intensity and working status of the pre-processing system. This is to simultaneously achieve stable filtration vacuum, impeller life meeting predetermined requirements, and suppression of fouling and cavitation development under fluctuating load conditions in the wet phosphoric acid filtration process.

2. The corrosion-resistant turbine vacuum system for wet-process phosphoric acid production as described in claim 1, characterized in that, After each update of the three-phase flow spatial state field, the multi-scale acoustic sparse reconstruction module sends information representing droplet cluster regions, high-concentration solid particle regions, and local liquid content anomaly regions in the three-phase flow spatial state field to the temporal polarization analysis module. The temporal polarization analysis module assigns prior structural information weights to the modal time series of the corresponding time period based on the information, and after constructing the structural state field, feeds back information on impeller high-stress regions and high-risk fouling initiation regions to the multi-scale acoustic sparse reconstruction module to adjust the sparse weights of the corresponding spatial units. Thus, a closed-loop iteration is formed through the bidirectional correction between the three-phase flow spatial state field and the structural state field.

3. The corrosion-resistant turbine vacuum system for wet-process phosphoric acid production as described in claim 1, characterized in that, The non-uniform sampling weights used in the multi-scale acoustic sparse reconstruction module are determined by a combination of acoustic propagation sensitivity, fluid turbulence intensity in the pre-processing system, and vacuum pipeline structural characteristics. This results in sensors in the acoustic sensing subsystem near the gas-liquid separation section, near the gas-liquid interface, and in areas prone to condensation having a greater weight in the reconstruction of the three-phase flow spatial state field than other sensors.

4. The corrosion-resistant turbine vacuum system for wet-process phosphoric acid production as described in claim 1, characterized in that, The multi-scale acoustic sparse reconstruction module is configured to hierarchically expand the acoustic field signal acquired by the acoustic sensing subsystem on a multi-scale acoustic basis with scale index and spatial index, and apply penalty weights to the high-scale coefficients to enhance the regions corresponding to local strong impact sound and droplet cluster sound in the three-phase flow spatial state field, and to explicitly identify the droplet cluster region and the solid particle impact region.

5. The corrosion-resistant turbine vacuum system for wet-process phosphoric acid production as described in claim 1, characterized in that, The micro-vibration monitoring subsystem is configured to perform multimodal decomposition on the micro-vibration signals collected by the vibration sensor to distinguish the impeller speed-related fundamental frequency mode, the mode generated by fluid excitation, and the higher-order mode related to material fatigue. The time-series polarization analysis module only performs the time-series polarization transformation on the time series of the higher-order modes related to material fatigue and gives the fatigue risk index of each impeller region in the structural state field.

6. The corrosion-resistant turbine vacuum system for wet-process phosphoric acid production as described in claim 5, characterized in that, When performing the time-series polarization transformation, the temporal polarization analysis module groups the higher-order mode time series into channels of different polarization layers according to the time sequence. By calculating the structural information reliability index of each channel, channels with reliability higher than a set threshold are marked as structural information channels, and channels with reliability lower than the threshold are marked as noise channels. The structural state field is constructed only based on the structural information channels.

7. The corrosion-resistant turbine vacuum system for wet-process phosphoric acid production as described in claim 1, characterized in that, The state coupling module is configured to map the three-phase flow spatial state field and the structural state field into a multi-dimensional system state vector including vacuum degree, local liquid content, local particle concentration, impeller stress level and fouling initiation probability. The multi-dimensional system state vector is accumulated within a preset time window to form a multi-dimensional system state sequence, and the fan health probability cloud map data and vacuum behavior prediction data are calculated based on the multi-dimensional system state sequence.

8. The corrosion-resistant turbine vacuum system for wet-process phosphoric acid production as described in claim 1, characterized in that, The self-sensing control kernel is configured to generate a feedforward control quantity based on the vacuum behavior prediction data, and to superimpose the feedforward control quantity with the feedback control quantity calculated based on the current filtration vacuum deviation to form a comprehensive control quantity. The comprehensive control quantity is used to drive the frequency converter of the corrosion-resistant turbine fan to adjust the fan speed and simultaneously adjust the heat exchanger temperature of the waste heat recovery and temperature control system and the spray opening sequence of the pretreatment system. When generating the comprehensive control quantity, the self-sensing control kernel uses the upper limit of impeller fatigue failure probability, the upper limit of scale germination probability, and the upper limit of preset vacuum fluctuation amplitude in the fan health probability cloud map data as constraints.

9. The corrosion-resistant turbine vacuum system for wet-process phosphoric acid production as described in claim 8, characterized in that, The self-sensing control kernel includes a self-learning unit, which is configured to associate and store the wind turbine health probability cloud map data, the vacuum behavior prediction data, and actual fault events and maintenance records at the end of the operating cycle. In subsequent operating cycles, it updates the non-uniform sampling weight, the structural information channel reliability threshold, and the weight coefficients of the feedforward control quantity and the feedback control quantity online based on the associated storage results. When it detects that the three-phase flow space state field is missing, the structural state field is missing, or the state coupling module does not output the wind turbine health probability cloud map data and the vacuum behavior prediction data, it prohibits the output of steady-state operation control commands and only outputs protective shutdown control commands or load reduction operation control commands.

10. A control method for a corrosion-resistant turbine vacuum system used in wet-process phosphoric acid production, characterized in that, The corrosion-resistant turbine vacuum system for wet-process phosphoric acid production as described in claim 1 includes steps S1 to S5. In step S1, an acoustic sensing subsystem installed on the vacuum pipeline corresponding to the inlet of the pretreatment system and the corrosion-resistant turbine blower is used to collect sound field signals and input them into the multi-scale acoustic sparse reconstruction module. Step S2: The multi-scale acoustic sparse reconstruction module reconstructs the three-phase flow spatial state field under sparse constraints and non-uniform sampling weights, and sets the weight coefficients of the micro-vibration monitoring subsystem data. Step S3: The micro-vibration monitoring subsystem installed on the corrosion-resistant turbine fan casing collects impeller micro-vibration signals, and performs the time-series polarization transformation in the time-series polarization analysis module under the constraints of the three-phase flow spatial state field to construct the structural state field and feed it back to the multi-scale acoustic sparse reconstruction module. Step S4: The state coupling module calculates the fan health probability cloud map data and the vacuum behavior prediction data based on the three-phase flow spatial state field and the structural state field. Step S5: The self-sensing control kernel generates control commands based on the fan health probability cloud map data and the vacuum behavior prediction data to coordinate the adjustment of the corrosion-resistant turbine fan speed, the heat exchange temperature of the waste heat recovery and temperature control system, and the spray intensity and working state of the pretreatment system, so as to maintain the filtration vacuum within the set range and suppress scaling, corrosion and cavitation development during the operation of the wet phosphoric acid filtration process.

Images