Fault tracing method and device in sylvite production process and electronic equipment

Abstract

The invention discloses a fault tracing method and device in a sylvite production process and electronic equipment. The method comprises the following steps: collecting production data corresponding to a plurality of pieces of production equipment in a preset time period in a sylvite production process; according to the production data, target hidden representations corresponding to the multiple pieces of production equipment in a preset time period are determined, and then a fault track and a fault-free track in the sylvite production process are determined; based on the fault track and the fault-free track, potassium salt deviation values corresponding to the multiple pieces of production equipment in the preset time period are determined; based on the potassium salt deviation values corresponding to the multiple pieces of production equipment respectively, an instantaneous elasticity information set in the preset time period is determined; and based on the instantaneous elastic information sets corresponding to the plurality of predetermined time periods of the target batch, performing fault tracing on the potassium salt production process of the target batch, and determining fault production equipment. The technical problem that the traceability result of the fault production equipment in the potassium salt production process is not accurate in the prior art is solved.

Description

Technical Field

[0001] This application relates to the field of potash fertilizer production technology in salt lakes, and more specifically, to a method, apparatus, and electronic equipment for tracing the source of failures in the potash production process. Background Technology

[0002] The production process of potash fertilizer (i.e., potash) from salt lakes operates under the interaction of multiple physical fields, including vacuum heat exchange, crystal slurry circulation, and high-speed separation. This highly coupled process, coupled with the slow drift of parameters depending on the ionic strength of the raw materials and the ambient vacuum level, results in extremely scarce available fault samples for analysis. When slight thermal or solute imbalances occur during production, they ultimately manifest as fluctuations in the purity and yield of the finished potash product. However, the root cause often lies hidden in the subtle efficiency degradation of any upstream production equipment. Therefore, tracing the faults in the potash production process is crucial for ensuring high purity and yield of the finished potash product.

[0003] Existing technologies rely on empirical thresholds or purely data-driven anomaly detection methods. On the one hand, these methods struggle to maintain robustness in low-failure-sample scenarios. On the other hand, fault attribution results often fail to conform to the constraints of heat and solute conservation, leading to unreasonable attribution results for faulty production equipment in the potash production process. Therefore, related technologies suffer from inaccurate attribution results for faulty production equipment in the potash production process.

[0004] There is currently no effective solution to the above problems. Summary of the Invention

[0005] This application provides a method, apparatus, and electronic device for tracing faults in the potassium salt production process, so as to at least solve the technical problem of inaccurate tracing results of faulty production equipment in the potassium salt production process in related technologies.

[0006] According to one aspect of the embodiments of this application, a fault tracing method for a potash production process is provided, comprising: collecting production data corresponding to multiple production equipment during a predetermined time period in the potash production process; determining target hidden representations corresponding to the multiple production equipment during the predetermined time period based on the production data corresponding to the multiple production equipment, wherein the target hidden representations are used to describe the production data of the corresponding production equipment after considering the coupling effect between the corresponding production equipment and other production equipment, and after considering heat conservation and solute conservation, wherein the other production equipment are production equipment other than the corresponding production equipment among the multiple production equipment; determining fault trajectories and fault-free trajectories of the potash production process during the predetermined time period based on the target hidden representations corresponding to the multiple production equipment, wherein the fault trajectory is used to describe the change of production data over time under fault disturbance conditions in the potash production process, and the fault-free trajectory is used to describe the change of production data over time under fault disturbance conditions in the potash production process. This method describes the changes in production data over time in the potash production process under conditions of no fault disturbance. Based on fault trajectories and fault-free trajectories, it determines the potash deviation corresponding to multiple production equipment within a predetermined time period. Based on the potash deviation corresponding to multiple production equipment, it determines the instantaneous elasticity information set within the predetermined time period. The instantaneous elasticity information set includes instantaneous elasticity information corresponding to multiple production equipment, which is used to quantify the instantaneous impact of production equipment on potash deviation within the predetermined time period. By determining the instantaneous elasticity information set within the predetermined time period, it determines the instantaneous elasticity information sets corresponding to multiple predetermined time periods included in the potash production process of the target batch. Based on the instantaneous elasticity information sets corresponding to multiple predetermined time periods, it traces the fault source in the potash production process of the target batch to identify the faulty production equipment in the potash production process of the target batch.

[0007] According to another aspect of the embodiments of this application, a fault tracing device for a potassium salt production process is provided, comprising: a data acquisition module, configured to acquire production data corresponding to multiple production devices during a predetermined time period in the potassium salt production process; a first determination module, configured to determine target hidden representations corresponding to multiple production devices during the predetermined time period based on the production data corresponding to the multiple production devices, wherein the target hidden representations are used to describe the production data of the corresponding production device after considering the coupling effect between it and other production devices, and after considering heat conservation and solute conservation, wherein the other production devices are production devices other than the corresponding production device among the multiple production devices; and a second determination module, configured to determine fault trajectories and fault-free trajectories of the potassium salt production process during the predetermined time period based on the target hidden representations corresponding to the multiple production devices, wherein the fault trajectory is used to describe the change of production data over time under fault disturbance conditions in the potassium salt production process, and the fault-free trajectory is used to describe the potassium salt production process under fault disturbance conditions. The production data changes over time under conditions of no fault disturbance in the salt production process; the third determination module is used to determine the potassium salt deviation amount corresponding to multiple production equipment in a predetermined time period based on fault trajectories and fault-free trajectories; the fourth determination module is used to determine the instantaneous elasticity information set in the predetermined time period based on the potassium salt deviation amount corresponding to multiple production equipment, wherein the instantaneous elasticity information set includes the instantaneous elasticity information corresponding to multiple production equipment, and the instantaneous elasticity information is used to quantify the instantaneous impact of production equipment on potassium salt deviation amount in the predetermined time period; the fifth determination module is used to determine the instantaneous elasticity information set corresponding to multiple predetermined time periods in the potassium salt production process of the target batch by using the method of determining the instantaneous elasticity information set in the predetermined time period; the sixth determination module is used to trace the fault source in the potassium salt production process of the target batch based on the instantaneous elasticity information set corresponding to multiple predetermined time periods, and determine the faulty production equipment in the potassium salt production process of the target batch.

[0008] According to another aspect of the embodiments of this application, a non-volatile storage medium is provided, which stores multiple instructions adapted for a fault tracing method in a potassium salt production process, any one of which is loaded and executed by a processor.

[0009] According to another aspect of the embodiments of this application, an electronic device is provided, including: one or more processors and a memory, the memory being used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement any one of the following methods for fault tracing in a potassium salt production process.

[0010] According to another aspect of the embodiments of this application, a computer program product is provided, which, when executed on a data processing device, is suitable for performing the steps of a fault tracing method for a potash production process.

[0011] In this embodiment, production data corresponding to multiple production devices during a predetermined time period are collected during the potash production process. Based on the production data corresponding to the multiple production devices, target hidden representations corresponding to the multiple production devices are determined for each device during the predetermined time period. These target hidden representations describe the production data of the corresponding production device after considering the coupling effect with other production devices, as well as heat and solute conservation. The other production devices are those other than the corresponding production device among the multiple production devices. Based on the target hidden representations corresponding to the multiple production devices, fault trajectories and fault-free trajectories for the potash production process during the predetermined time period are determined. The fault trajectories describe the changes in production data over time when fault disturbances exist in the potash production process, while the fault-free trajectories describe the changes in production data when fault disturbances do not exist in the potash production process. The production data under fault disturbance conditions changes over time; based on fault trajectories and fault-free trajectories, the potassium salt deviation corresponding to multiple production equipment within a predetermined time period is determined; based on the potassium salt deviation corresponding to multiple production equipment, an instantaneous elasticity information set is determined within the predetermined time period, wherein the instantaneous elasticity information set includes instantaneous elasticity information corresponding to multiple production equipment, and the instantaneous elasticity information is used to quantify the instantaneous impact of production equipment on potassium salt deviation within the predetermined time period; by using the method of determining the instantaneous elasticity information set within the predetermined time period, the instantaneous elasticity information sets corresponding to multiple predetermined time periods included in the potassium salt production process of the target batch are determined; based on the instantaneous elasticity information sets corresponding to multiple predetermined time periods, fault source tracing is performed on the potassium salt production process of the target batch to identify the faulty production equipment in the potassium salt production process of the target batch. The goal is to collect production data from equipment during potash production, determine the fault trajectories and fault-free trajectories in the potash production process, analyze these trajectories to determine the instantaneous elasticity information of the equipment, and thus identify the faulty equipment in the potash production process. This aims to improve the accuracy of the tracing results for faulty equipment in the potash production process and solve the technical problem of inaccurate tracing results for faulty equipment in potash production in related technologies. Attached Figure Description

[0012] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0013] Figure 1 This is a flowchart of a method for tracing the source of failures in a potassium salt production process according to an embodiment of this application;

[0014] Figure 2 This is a flowchart of an optional method for tracing the source of failures in a potassium salt production process, provided according to an embodiment of this application.

[0015] Figure 3 This is a schematic diagram of an optional fault tracing device for a potassium salt production process according to an embodiment of this application. Detailed Implementation

[0016] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0017] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0018] For ease of description, the following explains some of the nouns or terms used in the embodiments of this application:

[0019] The Pitzer correction parameter is a correction factor mainly used to describe and calculate the activity coefficients of components in electrolyte solutions, and then to calculate the theoretical yield and purity of potassium salts under different conditions.

[0020] Hampel filtering is a nonlinear filtering technique used in signal processing, primarily for detecting and removing outliers in signals, i.e., data points that significantly deviate from normal values.

[0021] According to an embodiment of this application, a method embodiment for tracing the source of faults in a potassium salt production process is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0022] Figure 1 This is a flowchart of a fault tracing method for a potassium salt production process according to an embodiment of this application, such as... Figure 1 As shown, the method includes the following steps:

[0023] Step S102: Collect production data corresponding to multiple production equipment during a predetermined time period in the potassium salt production process;

[0024] It is understood that the potash production process involves multiple stages, such as evaporation, crystallization, and centrifugation. Each stage has corresponding production equipment, such as evaporation equipment in the evaporation stage, crystallization equipment in the crystallization stage, and centrifugation equipment in the centrifugation stage. Production data from these multiple production devices is collected over a predetermined time period, yielding separate production data for each device within that time period. This collection of production data provides a data foundation for subsequent fault tracing.

[0025] Optionally, online / offline data (i.e., production data) from the evaporation tank (i.e., evaporation equipment), crystallizer (i.e., crystallization equipment), and centrifuge (i.e., centrifugation equipment) are collected, and day / night cycle and shift information are added, concatenating them into a unified state vector. For the evaporation stage, the evaporation tank can be set as the upstream reference zone, and the temperature T is... E saline concentration C E Inbound traffic F Ei Export flow F Eo With vacuum degree V E Perform synchronous measurements to form evaporation segment vectors. This subvector contains the evaporation production data of the evaporation equipment, where... The vector dimension of the evaporation segment vector is 5. For the crystallization stage, the crystallizer can be set as the crystallization core region, and the inlet slurry temperature T... Ci 1. Outlet slurry temperature T Co Inlet saline concentration C Ci , outlet brine concentration C Co Slurry flow rate F C Vacuum degree V C Stirring rate R C and internal pressure P C Parallel measurements were performed to obtain the crystal segment vectors. This sub-vector contains the crystallization production data from the crystallization equipment. The vector representing the crystallization segment has a dimension of 8. During the centrifugation stage, the centrifuge can be set as the solid-liquid boundary zone, and the mother liquor flow rate F... M Moisture content of dried product D P Drum rotation speed ω F Three online indicators were measured. At the end of each batch, the purity (P) of the dried product was collected. Pu Production Y P With impurity content I PThree offline metrics are written into the centrifugal segment vector. This subvector contains the centrifugal production data of the centrifuge equipment, where... The vector dimension of the centrifugal segment is 6.

[0026] Optionally, the timestamp t can be expanded using a day-night cycle to obtain... Simultaneously, the shift number b and the cumulative duration τ of the batch are recorded and normalized to form a time-series context vector. Where h represents the unit hour. By concatenating the above four sub-vectors in time sequence using a synchronized time stamp, a 24-dimensional unified production state vector Φ0 can be obtained, which includes production data corresponding to multiple production devices.

[0027] Step S104: Based on the production data corresponding to multiple production equipment, determine the target hidden representation corresponding to multiple production equipment in a predetermined time period. The target hidden representation is used to describe the production data of the corresponding production equipment after considering the coupling effect between it and other production equipment, as well as the heat conservation and solute conservation. Other production equipment refers to the production equipment other than the corresponding production equipment among the multiple production equipment.

[0028] It is understandable that, based on the production data corresponding to multiple production devices, a target hidden representation is determined for each of these devices within a predetermined time period. This hidden representation describes the production data of the corresponding device after considering the coupling effects with other production devices, as well as the conservation of heat and solute. "Other devices" refers to all production devices other than those corresponding to the target hidden representation. By considering the coupling effects between devices, as well as the conservation of heat and solute, the target hidden representation can more accurately reflect the diffusion and change of faults during the production process, and make the fault analysis process conform to physical laws, thereby improving the accuracy and rationality of the results in determining the faulty production device.

[0029] In one optional embodiment, when multiple production devices include an evaporation device, a crystallization device, and a centrifugation device, the target hidden representation corresponding to each of the multiple production devices is determined based on the production data corresponding to each of the multiple production devices within a predetermined time period. This includes: determining the water evaporation rate and the instantaneous solute mass based on the evaporation production data of the evaporation device; determining the crystallization rate based on the crystallization production data of the crystallization device; obtaining the ideal evaporation trajectory of the evaporation device and the ideal crystallization trajectory of the crystallization device using a recursive Bayesian method based on the water evaporation rate, the instantaneous solute mass, the crystallization rate, and a preset first threshold, wherein the ideal evaporation trajectory describes the change of the evaporation production data over time under ideal conditions, and the ideal crystallization trajectory describes the change of the crystallization production data over time under ideal conditions; and determining the target hidden representation corresponding to each of the multiple production devices within the predetermined time period based on the ideal evaporation trajectory and the ideal crystallization trajectory.

[0030] It is understandable that, in the case of multiple production devices including evaporation equipment, crystallization equipment, and centrifugation equipment, the target hidden representation corresponding to each of the multiple production devices within a predetermined time period is determined as follows: First, based on the evaporation production data of the evaporation equipment, the water evaporation rate and instantaneous solute mass of the evaporation production stage are determined, and based on the crystallization production data of the crystallization equipment, the crystallization rate of the crystallization production stage is determined. Second, based on the aforementioned water evaporation rate, instantaneous solute mass, crystallization rate, and a preset first threshold, a recursive Bayesian method is used to obtain the ideal evaporation trajectory for the evaporation equipment, which describes the change of evaporation production data over time under ideal conditions, and the ideal crystallization trajectory for the crystallization equipment, which describes the change of crystallization production data over time under ideal conditions. Finally, based on the aforementioned ideal evaporation trajectory and ideal crystallization trajectory, the target hidden representation corresponding to each of the multiple production devices within the predetermined time period is determined. By using the recursive Bayesian method to generate the ideal evaporation trajectory and ideal crystallization trajectory, and then generating the target hidden representation that fully considers physical constraints and inter-device coupling effects, the accuracy of potassium salt fault tracing results can be enhanced.

[0031] Optionally, considering the uncertainties in the potassium salt production process, such as fluctuations in evaporation loss coefficient and impurity solubility, a recursive Bayesian method can be used to correct for uncertainties in the water evaporation rate, instantaneous solute mass, and crystallization rate. By using historical production data and real-time measurement data from the production equipment as observations, a likelihood function is constructed, the posterior distribution is estimated, and finally, the ideal evaporation trajectory and ideal crystallization trajectory of the evaporation and crystallization equipment under ideal conditions are generated.

[0032] Optionally, the evaporation rate and crystallization rate can be calculated using the material closure and heat transfer coupling formula. The evaporation loss coefficient and impurity solubility can be corrected using the recursive Bayesian method to generate ideal trajectories (including ideal evaporation trajectories and ideal crystallization trajectories). The aforementioned 24-dimensional state vector Φ0, along with the time label t0, is input and split into evaporator measuring point groups, crystallizer measuring point groups, and centrifuge measuring point groups in a predetermined order. The three measuring point groups are then uniformly labeled with the time label t0, thereby obtaining the production data corresponding to multiple production devices within a predetermined time period t0.

[0033] For the inlet flow rate F of the evaporation tank Ei Export flow F Eo and saline concentration C E Perform a one-minute sliding window material closure and calculate the moisture evaporation rate R. e With instantaneous solute mass M s Water evaporation rate R e It can be determined in the following way:

[0034] R e (k)=[F Ei (k)-F Eo (k)]×[1-C E (k)]

[0035] Among them, R e (k) represents the water evaporation rate (kg·min) at the k-th minute. -1 (kg per minute); F Ei (k), F Eo (k) represents the inlet and outlet flow rates of the evaporator at minute k (kg·min). -1 ); C E (k) represents the saline concentration (mass fraction) at minute k.

[0036] Heat exchange rate Q of the evaporation tank E Temperature difference ΔT between crystallizer and crystallizer C =T Ci -T Co Perform heat transfer coupling calculations to determine the crystallization rate R. c , put R c With the above R e M s Perform cross-checking of differences; if the deviation exceeds the limit ε m (i.e., preset the first threshold), then the heat transfer coefficient k will be automatically fine-tuned. h Crystallization rate R c It can be determined in the following way:

[0037] R c (k)=k h ×Q E (k)×ΔT C(k)

[0038] Among them, R c (k) represents the crystallization rate (kg·min) at minute k. -1 );k h For online adjustable heat transfer coefficient (kg·kJ) -1 ·℃ -1 (Kilograms per kilojoule per degree Celsius represents the mass (in kilograms) of material that can be processed or converted for every kilojoule of energy consumed and every degree Celsius change in temperature); Q E (k) represents the heat exchange in the evaporation tank at minute k (kg·min). -1 );ΔT C (k) represents the temperature difference (°C) in the crystallizer at minute k.

[0039] Alternatively, the evaporation loss coefficient β and the impurity solubility γ can be set as random parameters, and a recursive Bayesian method can be used to determine the evaporation loss coefficient β and the impurity solubility γ. e (k),R c [k] constructs a likelihood function for the observations, outputting the posterior distributions and confidence intervals of β and γ. - ,β + ] and [γ - ,γ + This allows for the incorporation of unmeasurable fluctuations from the field into rate estimation. For the uncertainty-corrected R... e (k) and R c (k) Integrate along the time axis to generate the ideal trajectory of the evaporation section. (i.e., ideal evaporation trajectory) and ideal crystallization trajectory (i.e., the ideal trajectory of crystallization) The posterior distributions of β and γ are also output.

[0040] In an optional embodiment, when the target hiding representation is represented in vector form, the target hiding representation corresponding to multiple production devices within a predetermined time period is determined based on the ideal evaporation trajectory and the ideal crystallization trajectory. This includes: determining attribute vectors corresponding to multiple production devices based on the ideal evaporation trajectory and the ideal crystallization trajectory, wherein the attribute vectors represent the production data of the production devices under ideal conditions; determining a summary vector corresponding to multiple production devices based on the attribute vectors corresponding to multiple production devices and a preset adjacency matrix, using a weighted calculation method, wherein the elements in the adjacency matrix represent the coupling strength between the corresponding production devices, and the summary vector represents the production data of the corresponding production device after considering the coupling effect with other production devices; determining an initial hiding representation corresponding to multiple production devices based on the summary vectors corresponding to multiple production devices and a preset trainable weight matrix, wherein the trainable weight matrix is ​​used to perform linear transformation and data amplification on the data in the summary vector; and correcting the initial hiding representations corresponding to multiple production devices to obtain the target hiding representations corresponding to multiple production devices within the predetermined time period.

[0041] It is understandable that, when the target hidden representation is represented in vector form, attribute vectors representing production data under ideal conditions are determined for each of the multiple production equipment based on the ideal evaporation trajectory and the ideal crystallization trajectory. Based on these attribute vectors and a pre-defined adjacency matrix, a weighted calculation method is used to determine the summary vector for each of the multiple production equipment. In the pre-defined adjacency matrix, rows and columns represent different production equipment, and each element represents the coupling strength between the production equipment corresponding to its row and column. This summary vector represents the production data of the corresponding production equipment after considering the coupling effect with other production equipment. Based on the summary vectors for each production equipment and a pre-defined trainable weight matrix used for linear transformation and data amplification of the data in the summary vectors, the initial hidden representations for each of the multiple production equipment are determined. These initial hidden representations do not consider heat and solute conservation, thus introducing some error. Based on heat and solute conservation, the initial hidden representations for each of the multiple production equipment are corrected to obtain the target hidden representations for each of the multiple production equipment within a predetermined time period. By using a pre-defined adjacency matrix, target hiding representation can accurately capture the coupling effect between production equipment, reflect the true state of the production equipment, and improve the accuracy of fault tracing results.

[0042] In one optional embodiment, the initial hidden representations corresponding to multiple production devices are corrected to obtain target hidden representations corresponding to the multiple production devices within a predetermined time period. This includes: determining the heat residual and material residual within the predetermined time period; for any initial hidden representation of any production device among the multiple production devices, correcting the initial hidden representation based on the heat residual, the heat capacity ratio of any production device, and a preset second threshold to obtain a first hidden representation, wherein the heat capacity ratio is used to quantify the contribution of any production device to the heat residual; correcting the initial hidden representation based on the material residual, the solute content ratio of any production device, and a preset third threshold to obtain a second hidden representation; determining any target hidden representation of any production device within the predetermined time period based on the first hidden representation and the second hidden representation; and determining the target hidden representations corresponding to the multiple production devices by using the method of determining any target hidden representation.

[0043] It is understood that the heat residual and material residual of the potash production process within a predetermined time period are determined. Based on the heat residual, the heat capacity ratio of any production equipment, and a preset second threshold, any initial hidden representation of any production equipment among multiple production equipment is corrected to obtain a first hidden representation of any production equipment. The aforementioned heat capacity ratio is used to quantify the contribution of any production equipment to the heat residual. Based on the material residual, the solute content ratio of any production equipment, and a preset third threshold, any initial hidden representation is corrected to obtain a second hidden representation of any production equipment. Based on the first and second hidden representations, any target hidden representation of any production equipment within the predetermined time period is determined. Using the method of determining any target hidden representation, the target hidden representations corresponding to multiple production equipment within the predetermined time period are determined. Through the calculation and correction of the heat residual and material residual, it is ensured that the obtained target hidden representation strictly adheres to the conservation of heat and solute, avoiding discrepancies between fault tracing results and physical logic, and improving the reliability and accuracy of fault tracing results.

[0044] In one optional embodiment, based on the heat residual, the heat capacity ratio of any production equipment, and a preset second threshold, any initial hidden representation is corrected to obtain a first hidden representation. This includes: if the heat residual is less than or equal to the preset second threshold, determining any initial hidden representation as the first hidden representation; if the heat residual is greater than the preset second threshold, correcting any initial hidden representation based on the heat residual and the heat capacity ratio to obtain a first corrected hidden representation; updating the heat residual based on the first corrected hidden representation to obtain an updated heat residual; if the updated heat residual is less than or equal to the preset second threshold, stopping the correction process, and determining the first corrected hidden representation obtained from the last correction as the first hidden representation.

[0045] It is understood that when the heat residual is less than or equal to a preset second threshold, it indicates that heat conservation is met, and in this case, any initial hidden representation is determined as the first hidden representation. When the heat residual is greater than the preset second threshold, it indicates that heat conservation is not met, and in this case, any initial hidden representation is corrected based on the heat residual and the heat capacity ratio to obtain a first corrected hidden representation for any production equipment. After obtaining the first corrected hidden representation, the heat residual is updated using the first corrected hidden representation to obtain the updated heat residual. The updated heat residual is compared with the preset second threshold. If the updated heat residual is less than or equal to the preset second threshold, the correction process stops, and the aforementioned first corrected hidden representation is determined as the first hidden representation; if the updated heat residual is greater than the preset second threshold, the above correction process continues until the obtained updated heat residual is less than or equal to the preset second threshold, at which point the correction process stops, and the first corrected hidden representation obtained from the last correction is determined as the first hidden representation. The above correction process ensures that the final result of the heat residual satisfies heat conservation, improving the rationality and accuracy of the fault tracing results.

[0046] Optionally, a second hidden representation based on solute conservation can be obtained by obtaining a first hidden representation. When the material residual is less than or equal to a preset third threshold, it indicates that solute conservation is met, and any initial hidden representation is determined as the second hidden representation. When the material residual is greater than the preset third threshold, it indicates that solute conservation is not met, and any initial hidden representation is corrected according to the material residual and the solute content ratio to obtain a second corrected hidden representation for any production equipment. After obtaining the second corrected hidden representation, the material residual is updated using the second corrected hidden representation to obtain the updated material residual. The updated material residual is compared with the preset third threshold. If the updated material residual is less than or equal to the preset third threshold, the correction process is stopped, and the above-mentioned second corrected hidden representation is determined as the second hidden representation; if the updated material residual is greater than the preset third threshold, the above correction process continues until the obtained updated material residual is less than or equal to the preset third threshold, at which point the correction process stops, and the second corrected hidden representation obtained from the last correction is determined as the second hidden representation.

[0047] Optionally, a graph convolutional network can be used to determine the target hidden representations corresponding to multiple production devices. The production devices are mapped as graph nodes. Based on the graph convolutional network, the initial hidden representations are corrected by writing back the heat residual and solute residual at each time slice (i.e., a predetermined time period). Then, fault perturbations are injected into the latent space to obtain fault-free trajectories and fault-containing trajectories (i.e., fault trajectories). The ideal trajectory... The initial state vector Φ0 is cached, time registration is performed on the three data streams, linear interpolation is applied to missing data points, and interval normalization is performed on dimensional differences. Each record is assigned a node label-time sequence number dual index to form a reference sequence Sref. The evaporator, crystallizer, and centrifuge are mapped to graph nodes v1, v2, and v3, respectively. A depth-first traversal is performed on the directions of aqueous stream, solid stream, and heat transfer, and the direction and intensity are written into the 3×3 adjacency matrix A according to the inlet-outlet rule. At the same time, the current measurement values ​​of each node in Sref are written into the node attribute matrix X (which includes attribute vectors corresponding to multiple production devices).

[0048] The node attribute matrix X is processed in three steps over time slices. First, the adjacency matrix A is used to sum the attribute vectors of the target node and its neighboring nodes by weight, resulting in a local summary (i.e., a summary vector) that includes the influence of the surrounding environment. Second, the summary vector is multiplied by the trainable weight matrix, and a new hidden representation (i.e., the initial hidden representation) is output based on the linear rectified activation function. Finally, the initial hidden representation is written back to the corresponding node and moved to the next time slice, and the cycle continues. This allows the initial hidden vector (i.e., the initial hidden representation in vector form) to absorb the adjacency coupling effect while maintaining the characteristics of the node, thus achieving dynamic linkage in continuous production.

[0049] A heat residual calculation based on heat conservation is performed. First, the heat input and output entries of the three nodes in the given time slice are traversed, and the heat residual is obtained using a "summation-difference" method. Then, the heat residual is quantitatively written back to the initial hidden vector of each node according to its heat capacity ratio, offsetting accumulated errors in real time. If the heat residual still exceeds a set threshold (i.e., a preset second threshold), the initial hidden vector is repeatedly updated, and the heat residual is also updated. The initial hidden representation is corrected through a "difference-allocation-writeback fine-tuning" method until convergence (i.e., the heat residual is less than the preset second threshold). Through this extremely short iteration, regardless of the number of convolution propagations within the graph convolutional network, the final output target hidden vector (i.e., the target hidden representation in vector form) strictly satisfies the requirement of heat balance (i.e., heat conservation), avoiding the problem of increasing errors in heat calculation over long sequences.

[0050] Perform material residual calculation based on solute conservation. Iterate through the solute inflow and outflow of the three nodes in the given time slice, calculating the material residual. Based on the current solute content percentage of each node, proportionally push the material residual back to the corresponding initial hidden vector, and iterate and correct until the material residual is less than the threshold ε. M (i.e., a preset third threshold) ensures that the material is closed at each step, preventing the problem of adding or subtracting material out of thin air during the convolution stacking process.

[0051] Step S106: Based on the target hiding representations corresponding to multiple production equipment, determine the fault trajectory and fault-free trajectory of the potassium salt production process within a predetermined time period. The fault trajectory is used to describe the change of production data over time when there is a fault disturbance in the potassium salt production process, and the fault-free trajectory is used to describe the change of production data over time when there is no fault disturbance in the potassium salt production process.

[0052] It is understandable that, based on the target hiding representations corresponding to multiple production devices, fault trajectories and fault-free trajectories for the potash production process within a predetermined time period are determined. The fault trajectories describe the changes in production data over time under fault disturbances in the potash production process, while the fault-free trajectories describe the changes in production data over time under fault-free disturbances. Through this process, not only can fault-free and fault trajectories be generated, but a precise comparison between these two trajectories can also be performed, providing more accurate and intuitive data support for identifying faulty production equipment in the potash production process.

[0053] In one optional embodiment, based on the target hiding representations corresponding to multiple production devices respectively, the fault trajectory and fault-free trajectory of the potash production process within a predetermined time period are determined, including: injecting fault perturbations into the target hiding representations corresponding to multiple production devices respectively to obtain perturbation hiding representations corresponding to multiple production devices respectively; determining fault-free trajectories based on the target hiding representations corresponding to multiple production devices respectively; and determining fault trajectories based on the perturbation hiding representations corresponding to multiple production devices respectively.

[0054] It can be understood that after obtaining the target hidden representations corresponding to multiple production equipment, fault perturbations, such as blockages, heat exchange failures, and vacuum fluctuations, are injected into these representations to obtain perturbation hidden representations for each equipment. The target hidden representations are then reverse-mapped to obtain the fault trajectory of the potash production process over a predetermined time period. Simultaneously, the perturbation hidden representations are reverse-mapped to obtain the fault-free trajectory of the potash production process over the same time period. This reverse mapping process is the decoding process, used to restore the target and perturbation hidden representations to intuitive production data. The flexibility of perturbation injection allows the fault tracing process to adapt to different types of faults and production environments, enhancing generalization ability and predictive ability for unknown faults. Furthermore, it provides rich fault sample data for the fault tracing process.

[0055] Optionally, typical fault disturbance vectors (i.e., fault disturbances represented in vector form) can be injected into the node latent space. Based on an expert knowledge base, common fault disturbances such as blockage, heat transfer failure, and vacuum fluctuations are mapped to parameter attenuation ratios α. f Perform an offset operation on the specified node, i.e., retain (1-α).f The original hidden vector (i.e., the target hidden representation z) f ), and superimposed α f z f The fault disturbance is obtained by deriving the disturbance hidden representation (1-α). f )z f +α f z f The offset amplitude in the above process is finely adjustable. By changing only the feature dimensions related to the failure mode, it can inherit the completed heat and material correction results without violating conservation relationships. This small-amplitude perturbation based on physical constraints can still generate engineering-usable failure trajectories in scenarios with extremely low measured failure sample sizes.

[0056] Optionally, the offset latent space (including the perturbation hidden representations corresponding to multiple production devices) and the unoffset latent space (i.e., the aforementioned node latent space, including the target hidden representations corresponding to multiple production devices) are decoded simultaneously. Then, the two sets of hidden vectors (target hidden representation and perturbation hidden representation) are restored to a readable sequence of physical quantities through reverse mapping, thereby obtaining the fault trajectory. (i.e., fault trajectory) and fault-free trajectory Output according to node-time dual index.

[0057] Step S108: Based on the fault trajectory and the fault-free trajectory, determine the potassium salt deviation corresponding to multiple production equipment in the predetermined time period.

[0058] It is understandable that, based on fault trajectories and fault-free trajectories, the potassium salt deviation amount corresponding to multiple production equipment in the potassium salt production process during a predetermined time period can be determined. This could be expressed as potassium salt purity deviation (characterized by the purity of the dried potassium salt) or potassium salt yield deviation (characterized by the potassium salt output). By comparing fault trajectories and fault-free trajectories, the potassium salt deviation amount of each production equipment during the predetermined time period can be accurately quantified, thus providing a quantitative basis for identifying faulty production equipment.

[0059] In one optional embodiment, the potassium salt deviation amount corresponding to multiple production devices within a predetermined time period is determined based on the fault trajectory and the fault-free trajectory. This includes: determining the theoretical potassium salt amount of the centrifuge device included in the multiple production devices within the predetermined time period based on the fault trajectory and the fault-free trajectory; determining the potassium salt deviation amount of the centrifuge device within the predetermined time period based on the theoretical potassium salt amount and the actual potassium salt amount of the centrifuge device; and performing reverse propagation of the potassium salt deviation amount of the centrifuge device based on the allocation weight to obtain the potassium salt deviation amount corresponding to multiple production devices within the predetermined time period. The reverse propagation starts from the centrifuge device and ends at the evaporation device included in the multiple production devices. The allocation weight is used to quantify the mass flow rate and heat flow power among the production devices.

[0060] It is understandable that, based on the fault trajectory and the fault-free trajectory, the theoretical potassium salt content of the centrifuges included in multiple production equipment within a predetermined time period is determined. Simultaneously, the actual potassium salt content of the centrifuges within the predetermined time period is measured. The difference between the theoretical and actual potassium salt content is used to obtain the potassium salt deviation of the centrifuges within the predetermined time period. Based on the weighting of the material flow and heat flow power allocation among the production equipment, and combined with the directed process flow diagram of the potassium salt production process, the potassium salt deviation of the centrifuges is propagated backwards to obtain the potassium salt deviation corresponding to each of the multiple production equipment within the predetermined time period. This backward propagation direction is the opposite direction of the potassium salt production process in the directed process flow diagram, i.e., from the centrifuges to the evaporators. This backward propagation deviation mechanism enables the potassium salt deviation to be traced back from the centrifuges to the faulty production equipment, providing a clear path for precise fault location and improving the efficiency and accuracy of fault tracing.

[0061] Optionally, based on the production process flow diagram, starting from the centrifuge, the potassium salt deviation from the centrifuge is propagated backward through the crystallizer (i.e., the crystallizing equipment) and the evaporation tank (i.e., the evaporation equipment), according to the weighting among the various production equipment. This propagation is then carried out at each step, with the potassium salt deviation being rationally allocated and adjusted based on the state of the previous production equipment and the coupling strength between it and the current production equipment.

[0062] Optionally, by comparing fault-free trajectories with fault-containing trajectories, and combining the phase balance database with the filtered measured dry product purity and yield, the purity deviation and yield deviation (i.e., the potassium salt deviation represented by dry product purity and the potassium salt deviation represented by yield) can be calculated, and a node-time-fault tensor can be formed.

[0063] Optionally, the fault-free trajectory With fault trajectory Synchronous reading, and extracting the following two types of fields second by second from the three nodes of evaporator v1, crystallizer v2 and centrifuge v3: (1) Liquid phase comprehensive salt mass fraction Represents the instantaneous concentration of dissolved solids, where, This indicates that at time slice t, node v i (1) The total mass fraction of salt in the liquid phase; (2) The liquid phase temperature in, This indicates that at time slice t, node v i The liquid phase temperature. The two types of field values ​​extracted above are filled with missing values ​​using bidirectional linear interpolation, and then normalized based on the minimum-maximum factor to ensure the values ​​fall within the [0, 1] interval. The processed data is then written into the spatiotemporal index matrix M according to row → time slice number, column → node number, layer → {c, θ}. ct .

[0064] Optionally, M can be used ct The database of multi-component phase equilibrium of brine is accessed line by line. This database was built offline based on many years of sampling data and covers KCl-NaCl-MgC. l2 -CaC l2 The database contains data on a pentagonal system of potassium chloride (potassium chloride, sodium chloride, magnesium chloride, calcium chloride, and water), including concentrations ranging from 0 to 30 wt% and temperatures from 5 to 120 °C. The data is stored in the form of an equation of state with Pitzer correction parameters. During a query, the database first locates the nearest concentration-temperature grid, then performs two-dimensional spline interpolation on eight thermodynamic data points, ultimately outputting the theoretical potassium chloride crystal (i.e., potassium salt) content ρ. th (v i ,t), where ρ th (v i ,t) represents the purity of potassium chloride as a dry product, measured over time t in the production equipment v. i The theoretical amount of potassium salt.

[0065] Optionally, the purity P of the dry potassium salt product, pushed in real time by the centrifuge PLC bus (Programmable Logic Controller Bus), can be... m (t) and potassium salt production Y m (t)) Read the row storage table, where P m (t) represents the actual amount of potassium salt in the centrifuge, characterized by the dry purity of potassium salt at time slice t. m (t) represents the actual amount of potassium salt produced by the centrifuge at time slice t, characterized by potassium salt yield. Hampel filtering is first applied to both sequences to remove spikes, followed by an exponential moving average with a damping coefficient of 0.3 to reduce high-frequency noise. For occasional null values, three-point proximity backfilling is used to ensure time sequence integrity. The processed measured sequences are then resampled to the value of ρ. th A time axis with completely identical (v3,t) is used to generate a reference set S. meas .

[0066] The theoretical crystalline phase content (i.e., the theoretical potassium salt content of the centrifuge) at each centrifuge node is compared second by second with the measured index (i.e., the actual potassium salt content of the centrifuge). The potassium salt deviation of the centrifuge is calculated, including the potassium salt purity deviation ΔP(t), characterized by the purity of the dried potassium salt, and the potassium salt yield deviation ΔY(t), characterized by the potassium salt production rate. The potassium salt purity deviation ΔP(t) and the potassium salt yield deviation ΔY(t) can be determined as follows:

[0067] [ΔP(t),ΔY(t)]=[ρ th (v3,t)-P m (t),Yth (t)-Y m (t)]

[0068] Where, ρ th (v3,t) represents the theoretical percentage of KCl crystal phase at time slice t (i.e., the theoretical amount of potassium salt at time slice t, characterized by the dry purity of potassium chloride); Y th (t) is the basis The theoretical potassium salt quantity of the centrifuge, characterized by potassium salt yield, is calculated at time slice t; ΔP(t) and ΔY(t) are the potassium salt purity deviation of the centrifuge, characterized by potassium salt dry product purity, and the potassium salt yield deviation of the centrifuge, characterized by potassium salt yield, respectively, at time slice t.

[0069] ΔP(t) and ΔY(t) are expanded in a three-dimensional manner according to node-time-fault category: the first dimension corresponds to v1, v2, and v3; the second dimension is filled with all T time slices; and the third dimension maps the injected N types of fault disturbance labels. The two types of deviations (i.e., potassium salt purity deviation and potassium salt yield deviation) are written into a tensor of form 3×T×N, and the node's Chinese alias, sampling period, and data integrity flag are recorded in the header. The dry product purity deviation tensor and the yield deviation tensor are concatenated along the node dimensions to generate a single quality-yield deviation tensor Ω.

[0070] Step S110: Based on the potassium salt deviation amount corresponding to multiple production equipment, determine the instantaneous elasticity information set within a predetermined time period. The instantaneous elasticity information set includes instantaneous elasticity information corresponding to multiple production equipment. The instantaneous elasticity information is used to quantify the instantaneous influence of the production equipment on the potassium salt deviation amount within the predetermined time period.

[0071] It is understandable that, based on the potassium salt deviation amounts corresponding to multiple production equipment, the instantaneous elasticity information corresponding to each production equipment within a predetermined time period is determined, thus forming an instantaneous elasticity information set for that predetermined time period. This instantaneous elasticity information is used to quantify the instantaneous impact of the production equipment on the potassium salt deviation amounts within the predetermined time period. The instantaneous elasticity information set not only provides the determination of the faulty production equipment but also provides quantitative information on the degree of impact of the production equipment on the potassium salt deviation amounts, enhancing the interpretability of the fault tracing results.

[0072] Optionally, the potassium salt deviation of the centrifuge can be propagated in reverse according to the material flow rate and heat flow power based on the process flow diagram to obtain the instantaneous elasticity information of each node. This information is then accumulated into batch elasticity (i.e., cumulative elasticity information) and combined with the production reduction coefficient to generate comprehensive elasticity information for each production device.

[0073] Optionally, the quality-yield deviation tensor Ω can be correlated with the directed process flow graph G. p Load them together. Reorder Ω along the node dimension so that the internal node indexes match G.p The v1 (evaporation tank), v2 (crystallizer), and v3 (centrifuge) are consistent. The quality-yield deviation tensor Ω with aligned nodes is then compared with the directed process flow diagram G. p = (V, E) performs topology propagation simultaneously, where V = {v1, v2, v3} represents the evaporator, crystallizer, and centrifuge, respectively. Ω carries the node, time, fault, and deviation type (dry product purity | yield) in four dimensions, respectively. Elements in E represent the mass flow and heat flow power between production equipment. Read each edge e in the directed process graph. i→j Real-time material flow With heat flow power The two readings are linearly scaled and then summed to obtain the allocated weights. Among them, e i→j This represents the edge between node i and node j in the directed process flow diagram. This represents the mass flow between node i and node j. κ represents the heat flux power between node i and node j. i→j (t) represents the weight allocated between node i and node j in time slice t. Normalization is performed on all κ entering the same merge node so that any time slice satisfies ∑ i κ i→j (t) = 1, ensuring that both material balance and energy balance are satisfied during the difference backtracking process.

[0074] Optionally, centrifuge node v3 can be used as the starting point for reverse propagation. Its potassium salt deviation vector in Ω (i.e., the potassium salt deviation represented by a two-dimensional vector, with the elements of the vector being the potassium salt purity deviation and the potassium salt yield deviation, respectively) can be split line by line along the reverse edge, thereby propagating the potassium salt deviation of the centrifuge equipment in reverse. The steps of reverse propagation are: (1) Determine the current downstream node v j (2) According to each κ i→j (t) Split the potassium salt deviation to the direct upstream node v i (3) Remove the downstream nodes and edges that have been split from the list to be processed, and add the newly obtained upstream nodes to the queue to be processed; (4) Repeat steps (1)-(3) until the source is traced back to the evaporation pool node v1. After each splitting action, verify the difference between the received potassium salt deviation and the required received potassium salt deviation of the upstream node. If the difference exceeds the preset threshold, write back the potassium salt deviation corresponding to the difference to this node according to the proportion of each sub-edge (i.e., the weight of the distribution) to ensure that the cumulative error does not spread. After the entire traversal is completed, the potassium salt deviation vector dvi(t) of the production equipment that changes with time is obtained. Each element is bound to its reverse propagation path in the directed process flow diagram. Here, dvi(t) represents the potassium salt deviation vector at node v1 in time slice t.i The potassium salt deviation vector.

[0075] Optionally, the potassium salt deviation vector of the production equipment is dvi(t) = [d P vi(t),d Y vi(t)] and the original absolute deviation ωvi(t) in the quality-yield deviation tensor Ω at the same node and simultaneously occupies [|Ω] P vi,t|,|Ω Y The indexes are aligned using vi,t|]; missing entries are filled using linear interpolation; and abnormal spikes are removed using a three-standard-deviation rule, where d P vi(t) represents v at time slice t. i The deviation in potassium salt purity, d Y vi(t) represents v at time slice t. i The potash production deviation, ωvi(t), represents the value of v at node t in the quality-yield deviation tensor Ω. i Potassium salt deviation, |Ω P vi,t| represents the value at time slice t in the quality-yield deviation tensor Ω. i The deviation in potassium salt purity, |Ω Y vi,t| represents the value at time slice t in the quality-yield deviation tensor Ω. i The deviation in potassium salt production. The ratio of the two is calculated by component, and node v is obtained. i The instantaneous elastic vector ξvi(t) (i.e., the instantaneous elastic information represented in vector form) can be determined as follows:

[0076]

[0077] Where ξvi(t) represents node v at time slice t. i The instantaneous elasticity information is represented in vector form, where ε is a minimal positive number to prevent the denominator from being zero; the upper and lower components of the instantaneous elasticity vector ξvi(t) are respectively represented at node v in time slice t. i Real-time elasticity information regarding dry product purity and real-time elasticity information regarding production output are generated. ξvi(t) is written into a local elasticity table B, where the row index is set to the production equipment number, the column index is set to the timestamp, the cell stores the two-dimensional instantaneous elasticity vector, and metadata such as sampling interval and data quality flags are appended to the table header.

[0078] Step S112: Using the method of determining the instantaneous elastic information set of a predetermined time period, determine the instantaneous elastic information set corresponding to each of the multiple predetermined time periods included in the production process of the target batch of potassium salt.

[0079] It is understandable that by using a predetermined set of instantaneous elasticity information for a specific time period, the instantaneous elasticity information sets corresponding to multiple predetermined time periods included in the production process of the target batch of potassium salt can be determined. By continuously analyzing the instantaneous elasticity information sets of production equipment over multiple time periods, the performance trend of the production equipment can be predicted, reducing the risk of production interruptions due to equipment failure.

[0080] Step S114: Based on the instantaneous elastic information sets corresponding to multiple predetermined time periods, trace the fault source of the potash production process of the target batch and determine the faulty production equipment in the potash production process of the target batch.

[0081] It is understandable that, based on the instantaneous elasticity information sets corresponding to multiple predetermined time periods, fault tracing can be performed on the potash production process of the target batch to identify the faulty production equipment. By continuously tracking and analyzing the instantaneous elasticity information sets of the production equipment at different time periods, the faulty production equipment in the potash production process of the target batch can be accurately identified, avoiding false alarms or missed detections.

[0082] In one optional embodiment, based on the instantaneous elasticity information sets corresponding to multiple predetermined time periods, fault tracing is performed on the potash production process of the target batch to determine the faulty production equipment in the potash production process of the target batch. This includes: for any production equipment among multiple production equipment, determining the instantaneous elasticity information of any production equipment in the multiple predetermined time periods based on the instantaneous elasticity information sets corresponding to multiple predetermined time periods; determining the cumulative elasticity information of any production equipment in the target batch based on the instantaneous elasticity information corresponding to multiple predetermined time periods, wherein the cumulative elasticity information is used to quantify the cumulative impact of any production equipment in the target batch on the potash deviation; and determining the faulty production equipment in any production equipment in the target batch based on the cumulative elasticity information and the output reduction coefficient of any production equipment. The process involves several steps: First, a comprehensive elasticity information is used to quantify the impact of any production equipment on the potash deviation in the target batch, after considering production reduction factors. Based on this comprehensive elasticity information, a first correction factor and a second correction factor are used to determine the composite risk coefficient of any production equipment in the target batch. The first correction factor adjusts for the impact of equipment failure on the potash deviation, and the second correction factor adjusts for the impact of production delays caused by equipment failure diagnosis on the potash deviation. The composite risk coefficients for multiple production equipment in the target batch are then determined using the same method. Finally, based on these composite risk coefficients, the faulty production equipment in the potash production process of the target batch is identified.

[0083] It is understood that, for any one of multiple production devices, based on the instantaneous elasticity information corresponding to the production device in multiple predetermined time periods, cumulative elasticity information is determined to quantify the cumulative impact of any production device on the potassium salt deviation in the target batch. Based on the aforementioned cumulative elasticity information and the production reduction coefficient of any production device, comprehensive elasticity information is determined. This comprehensive elasticity information represents the impact of any production device on the potassium salt deviation in the target batch after considering the production reduction factor. Based on the comprehensive elasticity information, combined with a first correction factor used to adjust the impact of any production device malfunction on the potassium salt deviation, and a second correction factor used to adjust the impact of production lag caused by fault diagnosis on the potassium salt deviation, a composite risk coefficient for any production device is obtained. Using the method of determining the composite risk coefficient of any production device, composite risk coefficients corresponding to multiple production devices in the target batch are determined, and the production device corresponding to the largest composite risk coefficient is identified as the malfunctioning production device in the potassium salt production process of the target batch. By combining instantaneous elasticity information, cumulative elasticity information, production reduction coefficient, and composite risk coefficient, multi-dimensional fault detection can be achieved. This allows fault tracing in the potash production process to not only focus on the immediate performance of production equipment but also consider its long-term cumulative impact, as well as the direct economic losses and indirect time costs caused by the fault, thereby improving the comprehensiveness and accuracy of fault tracing.

[0084] Optionally, the local elasticity table B is extracted row by row, accumulated along the time axis, and multiplied by the sampling interval Δt (i.e., the time interval between two adjacent time slices), thereby converting the discrete instantaneous elasticity information into a continuous contribution; thus obtaining the batch cumulative elasticity vector. (i.e., cumulative elasticity information represented in vector form), in, Represents node v i The cumulative elasticity vector is given, where the first component represents the total impact of the production equipment on the dry purity of potassium salt in this batch, and the second component represents the total impact on the yield of potassium salt. The positive and negative signs directly indicate an increase or decrease. The multiple production equipment are then assigned their respective... Batch-level elasticity table T is generated by longitudinally splicing together according to the production equipment sequence. batch Each row is appended with a batch number-equipment name dual index. If an absolute value of a component in the cumulative elasticity vector of a certain production equipment shows an increasing trend across three consecutive batches, the corresponding cell is highlighted to alert maintenance personnel to pay attention to chronic performance degradation or potential faults.

[0085] T batch The production reduction factor λv is applied to the row and pre-defined production volume. i Multiplying these values ​​yields comprehensive elasticity information, which is then used to form a comprehensive elasticity table T. mix, where λv i Represents node v i The production reduction factor. Through the above process, the disturbance intensity of production equipment on quality (measured by dry product purity and yield) and the weighting of the yield loss caused by the same production equipment are combined into a unified measurement system, allowing a single table to simultaneously reflect both shocks. For the comprehensive elasticity table T... mix The dry product purity-yield combined index is sorted in descending order, and the sorting results are written into the equipment comprehensive flexibility list. It also includes node labels, comprehensive resilience information values, and fault category masks.

[0086] Optionally, a simplified fault list can be output by weighting the overall elasticity of production equipment spare parts costs and testing delay factors, normalizing it, and then outputting the list in descending order of risk (i.e., composite risk coefficient) values. The overall elasticity table is then aligned row by row with flowchart nodes v1, v2, and v3. For each row's overall elasticity vector (i.e., overall elasticity information represented in vector form), a unique node number, Chinese name, and physical location are written, thus establishing the node. Bias affects the mapping table. Two metrics are collected for each node in the mapping table: the recent average replacement cost of spare parts for the same model of production equipment, and the average detection latency of the sensing-diagnostic link. The raw values ​​of these two metrics are divided by the mean of all nodes to obtain a dimensionless correction factor with consistent dimensions. (i.e., node v) i The first correction factor) and (i.e., node v) i (the second correction factor).

[0087] Alternatively, node v can be i The comprehensive elasticity vector Synchronous multiplication and Node-by-node generation of composite risk coefficients (i.e., node v) i (composite risk coefficient), for all Perform range normalization to compress the result to the 0–1 interval. Then, perform range normalization on the normalized result. Sort the nodes in descending order of their numerical values, and then write the sort number, node name, and risk score (i.e., composite risk coefficient value) into the simplified fault list. The production batch number and generation time are appended to the top of the list. The list is designed for on-site operations and maintenance, and its structure retains only the necessary fields. Readers can quickly identify high-risk nodes without much calculation, thereby enabling real-time priority scheduling based on physical conservation on continuous potash production lines with low failure rates, uncertain parameters, and highly coupled multi-unit structures.

[0088] Through the above steps S102 to S114, the goal is to collect production data from production equipment during the potash production process, determine the fault trajectory and fault-free trajectory of the potash production process, analyze the fault trajectory and fault-free trajectory to determine the instantaneous elasticity information of the production equipment, and thus identify the faulty production equipment in the potash production process. This achieves the technical effect of improving the accuracy of the tracing results of faulty production equipment in the potash production process, thereby solving the technical problem of inaccurate tracing results of faulty production equipment in the potash production process in related technologies.

[0089] Based on the above embodiments and optional embodiments, this application proposes an implementation method for an optional fault tracing method in the potash production process. This implementation method can be understood as an artificial intelligence-based fault tracing method for the production process of salt lake potash fertilizer (i.e., potash). Figure 2 This is a flowchart of an optional fault tracing method for a potassium salt production process according to an embodiment of this application. The steps of the method include:

[0090] Step S1: Collect online / offline data (i.e., production data) from the evaporation tank (i.e., evaporation equipment), crystallizer (i.e., crystallization equipment), and centrifuge (i.e., centrifugation equipment), and add day-night cycle and shift information, splicing them into a unified state vector.

[0091] Step S11: Set the evaporation tank as the upstream reference zone and set the temperature T. E saline concentration C E Inbound traffic F Ei Export flow F Eo With vacuum degree V E Perform synchronous measurements to form evaporation segment vectors. This subvector contains the evaporation production data of the evaporation equipment.

[0092] Step S12: Set the crystallizer as the crystallization core zone and set the inlet slurry temperature T. Ci 1. Outlet slurry temperature T Co Inlet saline concentration C Ci , outlet brine concentration C Co Slurry flow rate F C Vacuum degree V C Stirring rate R C and internal pressure P C Parallel measurements were performed to obtain the crystal segment vectors. This sub-vector contains the crystallization production data from the crystallization equipment.

[0093] Step S13: Set the centrifuge to the solid-liquid separation zone and adjust the mother liquor flow rate F. M Moisture content of dried product D P Drum rotation speed ω FThree online indicators were measured. At the end of each batch, the purity (P) of the dried product was collected. Pu Production Y P With impurity content I P Three offline metrics are written into the centrifugal segment vector. This subvector contains the centrifugal production data of the centrifugal equipment.

[0094] Step S14: Perform day-night cycle expansion on the timestamp t to obtain... Simultaneously, the shift number b and the cumulative duration τ of the batch are recorded and normalized to form a time-series context vector.

[0095] Step S15: The four sub-vectors are concatenated in sequence using a synchronized time stamp to obtain a 24-dimensional unified production state vector Φ0, which includes production data corresponding to multiple production devices.

[0096] Step S2: Calculate the evaporation rate and crystallization rate according to the material closure and heat transfer coupling formula, and use the recursive Bayes method to correct the evaporation loss coefficient and impurity solubility to generate ideal trajectories (including ideal evaporation trajectory and ideal crystallization trajectory).

[0097] Step S21: Input the 24-dimensional state vector Φ0 output from step S1 along with the time label t0, and split it into the evaporator measuring point group, the crystallizer measuring point group and the centrifuge measuring point group in a predetermined order. Then, mark the three measuring point groups as the time label t0, and obtain the production data corresponding to multiple production equipment in the predetermined time period t0.

[0098] Step S22, adjust the inlet flow rate F of the evaporation tank. Ei Export flow F Eo and saline concentration C E Perform a one-minute sliding window material closure and calculate the moisture evaporation rate R. e With instantaneous solute mass M s Water evaporation rate R e The method for determining the value is the same as in the above embodiments, and will not be repeated here.

[0099] Step S23, heat exchange Q of the evaporation tank E Temperature difference ΔT between crystallizer and crystallizer C =T Ci -T Co Perform heat transfer coupling calculations to determine the crystallization rate R. c , put R c R obtained in step S22 e M s Perform cross-checking of differences; if the deviation exceeds the limit ε m (i.e., preset the first threshold), then the heat transfer coefficient k will be automatically fine-tuned. h Crystallization rate Rc The method for determining the value is the same as in the above embodiments, and will not be repeated here.

[0100] Step S24: Set the evaporation loss coefficient β and impurity solubility γ as random parameters, and use the recursive Bayesian method, with [R... e (k),R c [k] constructs a likelihood function for the observations, outputting the posterior distributions and confidence intervals of β and γ. - ,β + ] and [γ - ,γ + This allows for the incorporation of unmeasurable fluctuations in the field into rate estimation.

[0101] Step S25, for the uncertainty-corrected R e (k) and R c (k) Integrate along the time axis to generate the ideal trajectory of the evaporation section. (i.e., ideal evaporation trajectory) and ideal crystallization trajectory (i.e., the ideal trajectory of crystallization) The posterior distributions of β and γ are also output.

[0102] Step S3: Map the production equipment as graph nodes. Based on the graph convolutional network, use the heat residual and solute residual (i.e., material residual) to write back and correct the initial hidden representation in each time slice (i.e., a predetermined time period). Then inject fault perturbation into the latent space to obtain fault-free trajectory and fault-containing trajectory (i.e., fault trajectory).

[0103] Step S31, the ideal trajectory output in step S2 The initial state vector Φ0 is cached, time registration is performed on the three data streams, linear interpolation is applied to missing data points, and interval normalization is performed on dimensional differences. Each record is assigned a node label-time sequence number dual index to form a reference sequence Sref.

[0104] Step S32: Map the evaporator, crystallizer, and centrifuge to graph nodes v1, v2, and v3, respectively. Perform a depth-first traversal of the aqueous stream, solid stream, and heat transfer directions, and write the directions and intensities into the 3×3 adjacency matrix A according to the inlet-outlet rule. Simultaneously, write the current measurement values ​​of each node in Sref into the node attribute matrix X (which includes attribute vectors corresponding to multiple production devices).

[0105] Step S33 involves performing three operations on the node attribute matrix X over time slices. First, the adjacency matrix A is used to weight and summarize the attribute vectors of the target node and its neighboring nodes to obtain a local summary (i.e., a summary vector) that includes the influence of the surrounding environment. Second, the summary vector is multiplied by the trainable weight matrix, and a new hidden representation (i.e., the initial hidden representation) is output based on the linear rectified activation function. Finally, the initial hidden representation is written back to the corresponding node and moved to the next time slice, continuing the loop. This allows the initial hidden vector (i.e., the initial hidden representation in vector form) to absorb the adjacency coupling effect while maintaining the characteristics of the node, thus achieving dynamic linkage in continuous production.

[0106] Step S34: Calculate the heat residual based on heat conservation. First, iterate through the heat input and output entries of the three nodes in this time slice, and obtain the heat residual using a "summation-difference" method. Then, quantitatively write the heat residual back to the initial hidden vector of each node according to the heat capacity ratio of each node to offset the accumulated error in real time. If the heat residual still exceeds the set threshold (i.e., the preset second threshold), the initial hidden vector is repeatedly updated, and the heat residual is updated as well. The initial hidden representation is corrected through a "difference-allocation-writeback fine-tuning" method until convergence (i.e., the heat residual is less than the preset second threshold).

[0107] Through this extremely short iteration, no matter how many times the graph convolutional network propagates, the final output target hidden vector (i.e. the target hidden representation in vector form) strictly satisfies the requirement of thermal balance (i.e. heat conservation), avoiding the problem that the error increases with the calculation of heat in long sequences.

[0108] Step S35: Calculate the material residual based on solute conservation. Iterate through the solute inflow and outflow of the three nodes in the current time slice, calculating the material residual. Based on the current solute content ratio of each node, proportionally push the material residual back to the corresponding initial hidden vector, and iterate and correct until the material residual is less than the threshold ε. M (i.e., a preset third threshold) is achieved. Material closure is ensured at each step to prevent arbitrary additions or subtractions during convolution stacking.

[0109] Step S36: Inject typical fault disturbance vectors (i.e., fault disturbances represented in vector form) into the node latent space. Based on the expert knowledge base, common fault disturbances such as blockage, heat transfer failure, and vacuum fluctuations are mapped to parameter attenuation ratios α. f Perform an offset operation on the specified node, i.e., retain (1-α). f The original hidden vector (i.e., the target hidden representation z) f ), and superimposed α f z f The fault disturbance is obtained by deriving the disturbance hidden representation (1-α). f )z f +αf z f The offset amplitude in the above process is finely adjustable. By changing only the feature dimensions related to the failure mode, the heat and material correction results completed in steps S34 and S35 can be inherited without disrupting the conservation relationship. This small-amplitude perturbation based on physical constraints can still generate engineering-usable failure trajectories in scenarios with extremely low measured failure sample sizes.

[0110] Step S37: Simultaneously decode the offset latent space (including the perturbation hidden representations corresponding to multiple production devices) and the unoffset latent space (i.e., the aforementioned node latent space, including the target hidden representations corresponding to multiple production devices). Then, restore the two sets of hidden vectors (target hidden representation and perturbation hidden representation) into readable physical quantity sequences through reverse mapping, and obtain the fault trajectory. (i.e., fault trajectory) and fault-free trajectory Output according to node-time dual index.

[0111] Step S4: Compare the fault-free trajectory with the faulty trajectory, combine the phase balance database and the filtered measured dry product purity and yield, calculate the purity deviation and yield deviation (i.e., the potassium salt deviation represented by dry product purity and the potassium salt deviation represented by yield), and form the node-time-fault tensor.

[0112] Step S41, generate the fault-free trajectory in step S3. With fault trajectory Synchronously read in the data and extract the following two types of fields second by second from the three nodes: evaporator v1, crystallizer v2, and centrifuge v3:

[0113] (1) Mass fraction of total salt content in liquid phase Represents the instantaneous concentration of dissolved solids, where, This indicates that at time slice t, node v i The total mass fraction of salts in the liquid phase;

[0114] (2) Liquid phase temperature in, This indicates that at time slice t, node v i The liquid phase temperature.

[0115] The extracted field values ​​are then filled with missing values ​​using bidirectional linear interpolation, followed by minimum-maximum normalization to ensure the values ​​fall within the [0, 1] interval. The processed data is then written into the spatiotemporal index matrix M according to row → time slice number, column → node number, and layer → {c, θ}. ct .

[0116] Step S42, for M ctThe database of multi-component phase equilibrium of brine is accessed line by line. This database was built offline based on many years of sampling data and covers KCl-NaCl-MgC. l2 -CaC l2 The pentagonal system of -H2O (potassium chloride-sodium chloride-magnesium chloride-calcium chloride-water) includes data in the concentration range of 0–30wt% and the temperature range of 5–120℃. The data in the database is stored in the form of state equations plus Pitzer correction parameters.

[0117] The query first locates the nearest concentration-temperature grid, then performs two-dimensional spline interpolation on eight thermodynamic data points, and finally outputs the theoretical potassium chloride crystal (i.e., potassium salt) content ρ. th (v i ,t), where ρ th (v i ,t) represents the purity of potassium chloride as a dry product, measured over time t in the production equipment v. i The theoretical amount of potassium salt.

[0118] Step S43: The purity P of the potassium salt dry product, pushed in real time via the centrifuge PLC bus, is... m (t) and potassium salt production Y m (t)) Read the row storage table, where P m (t) represents the actual amount of potassium salt in the centrifuge, characterized by the dry purity of potassium salt at time slice t. m (t) represents the actual amount of potassium salt produced by the centrifuge at time slice t; Hampel filtering is first performed on the two sequences to remove spikes, and then exponential moving average with a damping coefficient of 0.3 is used to weaken high-frequency noise; for occasional null values, three-point proximity backfill is called to ensure the integrity of the time sequence.

[0119] Resample the processed measured sequence to ρ th A time axis with completely identical (v3,t) is used to generate a reference set S. meas .

[0120] Step S44 involves comparing the theoretical crystalline phase content (i.e., the theoretical potassium salt content of the centrifuge equipment) with the measured index (i.e., the actual potassium salt content of the centrifuge equipment) second by second, and calculating the potassium salt deviation of the centrifuge equipment. This includes the potassium salt purity deviation ΔP(t), characterized by the purity of the dried potassium salt, and the potassium salt yield deviation ΔY(t), characterized by the potassium salt production rate. The determination methods for the potassium salt purity deviation ΔP(t) and the potassium salt yield deviation ΔY(t) are the same as in the above embodiments, and the number of steps will not be repeated.

[0121] Step S45: Expand ΔP(t) and ΔY(t) in a three-dimensional manner according to node-time-fault category: the first dimension corresponds to v1, v2, and v3; the second dimension is filled with all T time slices; and the third dimension maps the N fault disturbance labels injected in step S3. Write the two types of deviations (i.e., potassium salt purity deviation and potassium salt production deviation) into a tensor of 3×T×N and record the node's Chinese alias, sampling period, and data integrity flag in the header.

[0122] Step S46: The dry product purity deviation tensor and the yield deviation tensor are spliced ​​along the node dimension to generate a single quality-yield deviation tensor Ω.

[0123] Step S5: Based on the process flow diagram, the potassium salt deviation of the centrifuge is propagated in reverse according to the material flow rate and heat flow power to obtain the instantaneous elasticity information of each node. The information is accumulated into batch elasticity (i.e., cumulative elasticity information) and combined with the production reduction coefficient to generate comprehensive elasticity information of each production equipment.

[0124] Step S51: Combine the quality-yield deviation tensor Ω generated in step S4 with the directed process flow diagram G. p Load them together. Reorder Ω along the node dimension so that the internal node indexes match G. p The v1 (evaporation tank), v2 (crystallizer), and v3 (centrifuge) are consistent.

[0125] Step S52: Connect the node-aligned quality-yield deviation tensor Ω to the directed process flow graph G. p = (V, E) simultaneously performs topology propagation, where V = {v1, v2, v3} represents the evaporator, crystallizer, and centrifuge respectively, Ω carries the node, time, fault, and deviation type (dry product purity | yield) in four dimensions respectively, and the elements in E are used to represent the mass flow and heat flow power between production equipment.

[0126] Read each edge e in the directed process flow diagram i→j Real-time material flow With heat flow power The two readings are linearly scaled and then summed to obtain the allocated weights. Among them, e i→j This represents the edge between node i and node j in the directed process flow diagram. This represents the mass flow between node i and node j. κ represents the heat flux power between node i and node j. i→j (t) represents the weight allocated between node i and node j in time slice t. Normalization is performed on all κ entering the same merge node so that any time slice satisfies ∑ i κ i→j(t) = 1, ensuring that both material balance and energy balance are satisfied during the difference backtracking process.

[0127] Using centrifuge node v3 as the starting point for reverse propagation, its potassium salt deviation vector in Ω (i.e., the potassium salt deviation represented by a two-dimensional vector, with the elements of the vector being the potassium salt purity deviation and the potassium salt yield deviation, respectively) is split line by line along the reverse edge, and then the potassium salt deviation of the centrifuge equipment is propagated in reverse. The steps of reverse propagation are: (1) Determine the current downstream node v j (2) According to each κ i→j (t) Split the potassium salt deviation to the direct upstream node v i (3) Remove the downstream nodes and edges that have been split from the list to be processed, and add the newly obtained upstream nodes to the queue to be processed; (4) Repeat steps (1)-(3) until the source is traced back to the evaporation pool node v1.

[0128] After each split operation, the difference between the received potassium salt deviation and the expected received potassium salt deviation of the upstream node is verified. If the difference exceeds the preset threshold, the potassium salt deviation corresponding to the difference is written back to this node according to the proportion of each sub-edge (i.e., the weighting) to ensure that the cumulative error does not spread.

[0129] After the entire traversal is completed, the potassium salt deviation vector dvi(t) of the production equipment changes over time. Each element in this vector is bound to its backward propagation path in the directed process flow diagram, where dvi(t) represents the deviation at time slice t at node v. i The potassium salt deviation vector.

[0130] Step S53, the potassium salt deviation vector dvi(t) of the production equipment output in step S52 is [d P vi(t),d Y vi(t)] and the original absolute deviation ωvi(t) in the quality-yield deviation tensor Ω at the same node and simultaneously occupies [|Ω] P vi,t|,|Ω Y The indexes are aligned using vi,t|]; missing entries are filled using linear interpolation; and abnormal spikes are removed using a three-standard-deviation rule, where d P vi(t) represents v at time slice t. i The deviation in potassium salt purity, d Y vi(t) represents v at time slice t. i The potash production deviation, ωvi(t), represents the value of v at node t in the quality-yield deviation tensor Ω. i Potassium salt deviation, |Ω P vi,t| represents the value at time slice t in the quality-yield deviation tensor Ω. iThe deviation in potassium salt purity, |Ω Y vi,t| represents the value at time slice t in the quality-yield deviation tensor Ω. i The deviation in potassium salt production. The ratio of the two is calculated by component, and node v is obtained. i The instantaneous elastic vector ξvi(t) is (i.e., the instantaneous elastic information represented in vector form). The method for determining the instantaneous elastic vector ξvi(t) is the same as in the above embodiment, and will not be repeated here.

[0131] Write ξvi(t) into the local elasticity table B. Set the row index of the table to the production equipment number, the column index to the timestamp, and the cell to store the two-dimensional instantaneous elasticity vector. Add the sampling interval and data quality flag metadata to the table header.

[0132] Step S54: Extract the local elasticity table B row by row, accumulate the data along the time axis, and multiply by the sampling interval Δt (i.e., the time interval between two adjacent time slices) to convert the discrete instantaneous elasticity information into a continuous contribution; thus obtaining the batch cumulative elasticity vector. (i.e., cumulative elasticity information represented in vector form), in, Represents node v i The cumulative elasticity vector is given, where the first component represents the total impact of the production equipment on the dry purity of potassium salt in this batch, and the second component represents the total impact on the yield of potassium salt. The positive and negative signs directly indicate an increase or decrease. The multiple production equipment are then assigned their respective... Batch-level elasticity table T is generated by longitudinally splicing together according to the production equipment sequence. batch Each row is appended with a batch number-equipment name dual index. If an absolute value of a component in the cumulative elasticity vector of a certain production equipment shows an increasing trend across three consecutive batches, the corresponding cell is highlighted to alert maintenance personnel to pay attention to chronic performance degradation or potential faults.

[0133] Step S55, T batch The production reduction factor λv is applied to the row and pre-defined production volume. i Multiplying these values ​​yields comprehensive elasticity information, which is then used to form a comprehensive elasticity table T. mix , where λv i Represents node v i The production reduction factor. Through the above process, the intensity of the disturbance to quality (measured by dry product purity and yield) by the production equipment is combined with the weight of the production loss caused by the same production equipment into a unified measurement system, so that a single table can reflect both impacts simultaneously.

[0134] Step S56, for the comprehensive elasticity table T mix The dry product purity-yield combined index is sorted in descending order, and the sorting results are written into the equipment comprehensive flexibility list. It also includes node labels, comprehensive resilience information values, and fault category masks.

[0135] Step S6: Introduce the weighted comprehensive elasticity of spare parts costs and testing delay factors of production equipment, normalize it, and output a simplified fault list in descending order of risk (i.e., composite risk coefficient) value.

[0136] Step S61: Align the integrated elasticity table output in step S5 row by row with the flowchart nodes v1, v2, and v3. Write the unique node number, Chinese name, and physical location into each row of the integrated elasticity vector (i.e., the integrated elasticity information represented in vector form), and establish the node. Deviation affects the mapping table.

[0137] Step S62: Collect two indicators for each node in the mapping table: the recent average replacement cost of spare parts for the same model of production equipment, and the average detection latency of the sensing-diagnosis link. Divide the raw values ​​of the two indicators by the mean of all nodes to obtain a dimensionless correction factor with consistent dimensions. (i.e., node v) i The first correction factor) and (i.e., node v) i (the second correction factor).

[0138] Step S63, move node v i The comprehensive elasticity vector Synchronous multiplication and Node-by-node generation of composite risk coefficients (i.e., node v) i (composite risk coefficient), for all Perform range normalization to compress the results to the 0–1 range.

[0139] Step S64, normalize the Sort the nodes in descending order of their numerical values, and then write the sort number, node name, and risk score (i.e., composite risk coefficient value) into the simplified fault list. The production batch number and generation time are appended to the top of the list. The list is designed for on-site operations and maintenance, and its structure retains only the necessary fields. Readers can quickly identify high-risk nodes without much calculation, thereby enabling real-time priority scheduling based on physical conservation on continuous potash production lines with low failure rates, uncertain parameters, and highly coupled multi-unit structures.

[0140] The above optional implementation methods achieve at least the following effects: the heat-solute dual conservation recursive estimation and physical constraint graph convolution are coupled into one, and through sliding window closure, heat transfer-crystallization rate joint correction and small-amplitude physical consistency perturbation (i.e. fault perturbation) injection, a pair of physically perfectly matched dual benchmarks, namely, fault-free trajectory and observed fault trajectory containing fault perturbation, are generated in real time and synchronously. This effectively avoids the dual dilemma of very few fault samples and difficulty in calibrating mechanism parameter drift in continuous production lines. It can continuously provide a reliable comparison without sacrificing conservation accuracy, and lay a highly reliable baseline for quality-yield anomaly identification. By embedding the dry product purity-yield deviation derived from the dual trajectory (i.e., the fault-free trajectory and the fault trajectory) into the topological reverse decomposition framework, and combining conservation weighting, node elastic accumulation, and economic-timeliness composite correction, an equipment elastic list is constructed that can locate, quantify contribution, and prioritize risks with one click. The list not only gives the actual percentage impact of each node on the deviation, but also integrates two types of operating cost factors: spare parts cost and diagnostic lag. This significantly improves the transparency of fault root cause explanation and the accuracy of decision priority, providing a directly executable quantitative basis for intelligent operation and maintenance and resource optimization scheduling of continuous production lines of salt lake potash fertilizer.

[0141] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0142] This embodiment also provides a fault tracing device for a potassium salt production process. This device is used to implement the above embodiments and preferred embodiments, and will not be repeated as already described. As used below, the terms "module" and "device" can refer to a combination of software and / or hardware that performs a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0143] According to an embodiment of this application, an apparatus embodiment for implementing a fault tracing method in a potash production process is also provided. Figure 3 This is a schematic diagram of a fault tracing device for a potassium salt production process according to an embodiment of this application, as shown below. Figure 3 As shown, the fault tracing device for the above-mentioned potassium salt production process includes a data acquisition module 302, a first determination module 304, a second determination module 306, a third determination module 308, a fourth determination module 310, a fifth determination module 312, and a sixth determination module 314. The device will be described below.

[0144] Data acquisition module 302 is used to collect production data corresponding to multiple production equipment during a predetermined time period in the potash production process;

[0145] The first determining module 304 is connected to the data acquisition module 302 and is used to determine the target hidden representation of the multiple production equipment respectively within a predetermined time period based on the production data corresponding to the multiple production equipment. The target hidden representation is used to describe the production data of the corresponding production equipment after considering the coupling effect between it and other production equipment, as well as the heat conservation and solute conservation. Other production equipment refers to the production equipment other than the corresponding production equipment among the multiple production equipment.

[0146] The second determining module 306, connected to the first determining module 304, is used to determine the fault trajectory and the fault-free trajectory of the potassium salt production process within a predetermined time period based on the target hiding representations corresponding to multiple production equipment respectively. The fault trajectory is used to describe the change of production data over time when there is a fault disturbance in the potassium salt production process, and the fault-free trajectory is used to describe the change of production data over time when there is no fault disturbance in the potassium salt production process.

[0147] The third determining module 308, connected to the second determining module 306, is used to determine the potassium salt deviation amount corresponding to multiple production equipment in a predetermined time period based on the fault trajectory and the fault-free trajectory.

[0148] The fourth determining module 310, connected to the third determining module 308, is used to determine the instantaneous elasticity information set within a predetermined time period based on the potassium salt deviation amount corresponding to multiple production equipment. The instantaneous elasticity information set includes instantaneous elasticity information corresponding to multiple production equipment, and the instantaneous elasticity information is used to quantify the instantaneous influence of the production equipment on the potassium salt deviation amount within the predetermined time period.

[0149] The fifth determining module 312, connected to the fourth determining module 310, is used to determine the instantaneous elastic information sets corresponding to the multiple predetermined time periods included in the production process of the target batch of potassium salt by adopting the method of determining the instantaneous elastic information set of the predetermined time period.

[0150] The sixth determining module 314, connected to the fifth determining module 312, is used to trace the source of faults in the potash production process of the target batch based on the instantaneous elastic information sets corresponding to multiple predetermined time periods, and to determine the faulty production equipment in the potash production process of the target batch.

[0151] This application provides a fault tracing device for a potassium salt production process. By setting up a data acquisition module 302, a first determination module 304, a second determination module 306, a third determination module 308, a fourth determination module 310, a fifth determination module 312, and a sixth determination module 314, it can achieve the purpose of collecting production data of production equipment in the potassium salt production process, determining the fault trajectory and fault-free trajectory of the potassium salt production process, analyzing the fault trajectory and fault-free trajectory to determine the instantaneous elasticity information of the production equipment, and thus determining the faulty production equipment in the potassium salt production process. This achieves the technical effect of improving the accuracy of the fault tracing results for the potassium salt production process, and solves the technical problem of inaccurate fault tracing results for the potassium salt production process in related technologies.

[0152] It should be noted that the above modules can be implemented by software or hardware. For example, for the latter, it can be implemented in the following ways: the above modules can be located in the same processor; or the above modules can be located in different processors in any combination.

[0153] It should be noted that the data acquisition module 302, the first determining module 304, the second determining module 306, the third determining module 308, the fourth determining module 310, the fifth determining module 312, and the sixth determining module 314 mentioned above correspond to steps S102 to S114 in the embodiments. The examples and application scenarios implemented by the above modules and their corresponding steps are the same, but they are not limited to the content disclosed in the above embodiments. It should be noted that the above modules, as part of the device, can run in a computer terminal.

[0154] It should be noted that the optional or preferred implementation methods of this embodiment can be found in the relevant descriptions in the embodiments, and will not be repeated here.

[0155] The fault tracing device for the above-mentioned potassium salt production process may also include a processor and a memory. The data acquisition module 302, the first determination module 304, the second determination module 306, the third determination module 308, the fourth determination module 310, the fifth determination module 312, and the sixth determination module 314 are all stored in the memory as program units. The processor executes the above-mentioned program units stored in the memory to realize the corresponding functions.

[0156] The processor contains a core that retrieves the corresponding program unit from memory. One or more cores may be configured. Memory may include non-persistent memory in computer-readable media, random access memory (RAM), and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory includes at least one memory chip.

[0157] This application provides a non-volatile storage medium storing a program that, when executed by a processor, implements a fault tracing method for the potassium salt production process.

[0158] This application provides an electronic device, which includes a processor, a memory, and a program stored in the memory and executable on the processor. When the processor executes the program, it performs the following steps: collecting production data corresponding to multiple production devices during a predetermined time period in the potassium salt production process; determining target hidden representations corresponding to the multiple production devices during the predetermined time period based on the production data, wherein the target hidden representations describe the production data of the corresponding production device after considering the coupling effect with other production devices, and considering heat conservation and solute conservation, and the other production devices are the production devices other than the corresponding production device among the multiple production devices; and determining fault trajectories and fault-free trajectories in the potassium salt production process during the predetermined time period based on the target hidden representations corresponding to the multiple production devices, wherein the fault trajectories describe the production data under fault disturbance conditions in the potassium salt production process. The data changes over time. Fault-free trajectories describe the changes in production data over time under conditions where there are no fault disturbances in the potash production process. Based on fault trajectories and fault-free trajectories, the potash deviation amounts corresponding to multiple production equipment within a predetermined time period are determined. Based on the potash deviation amounts corresponding to multiple production equipment, an instantaneous elasticity information set is determined within the predetermined time period. This instantaneous elasticity information set includes instantaneous elasticity information corresponding to multiple production equipment, which is used to quantify the instantaneous impact of production equipment on potash deviation amounts within the predetermined time period. Using the method of determining the instantaneous elasticity information set for the predetermined time period, the instantaneous elasticity information sets corresponding to multiple predetermined time periods within the potash production process of the target batch are determined. Based on the instantaneous elasticity information sets corresponding to multiple predetermined time periods, fault tracing is performed on the potash production process of the target batch to identify the faulty production equipment. The equipment mentioned in this paper can be servers, PCs, etc.

[0159] This application also provides a computer program product, which, when executed on a data processing device, is suitable for executing an initialization program with the following method steps: collecting production data corresponding to multiple production devices during a predetermined time period in the potassium salt production process; determining target hidden representations corresponding to the multiple production devices during the predetermined time period based on the production data corresponding to the multiple production devices, wherein the target hidden representations are used to describe the production data of the corresponding production device after considering the coupling effect with other production devices, as well as considering heat conservation and solute conservation, and the other production devices are the production devices other than the corresponding production device among the multiple production devices; and determining fault trajectories and fault-free trajectories of the potassium salt production process during the predetermined time period based on the target hidden representations corresponding to the multiple production devices, wherein the fault trajectories are used to describe the changes in production data over time under fault disturbance conditions in the potassium salt production process. The fault-free trajectory describes the change of production data over time under the condition that there are no fault disturbances in the potash production process. Based on the fault trajectory and the fault-free trajectory, the potash deviation corresponding to multiple production equipment in a predetermined time period is determined. Based on the potash deviation corresponding to multiple production equipment, the instantaneous elasticity information set in the predetermined time period is determined. The instantaneous elasticity information set includes the instantaneous elasticity information corresponding to multiple production equipment, which is used to quantify the instantaneous impact of production equipment on the potash deviation in the predetermined time period. By determining the instantaneous elasticity information set in the predetermined time period, the instantaneous elasticity information sets corresponding to multiple predetermined time periods in the potash production process of the target batch are determined. Based on the instantaneous elasticity information sets corresponding to multiple predetermined time periods, the fault source of the potash production process of the target batch is traced to determine the faulty production equipment in the potash production process of the target batch.

[0160] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0161] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0162] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0163] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0164] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0165] Memory may include non-persistent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, like read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0166] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0167] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0168] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0169] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A method for tracing the source of failures in a potassium salt production process, characterized in that, include: Collect production data from multiple production devices during a predetermined time period in the potassium salt production process; Based on the production data corresponding to the multiple production equipment, a target hidden representation corresponding to the multiple production equipment is determined during the predetermined time period. The target hidden representation is used to describe the production data of the corresponding production equipment after considering the coupling effect between it and other production equipment, as well as the heat conservation and solute conservation. The other production equipment refers to the production equipment other than the corresponding production equipment among the multiple production equipment. Based on the target hiding representations corresponding to the multiple production equipment respectively, the fault trajectory and the fault-free trajectory of the potassium salt production process are determined in the predetermined time period. The fault trajectory is used to describe the change of production data over time when there is a fault disturbance in the potassium salt production process, and the fault-free trajectory is used to describe the change of production data over time when there is no fault disturbance in the potassium salt production process. Based on the fault trajectory and the fault-free trajectory, determine the potassium salt deviation amount corresponding to the multiple production equipment in the predetermined time period; Based on the potassium salt deviation amount corresponding to the multiple production equipment, an instantaneous elasticity information set is determined for the predetermined time period. The instantaneous elasticity information set includes instantaneous elasticity information corresponding to the multiple production equipment. The instantaneous elasticity information is used to quantify the instantaneous influence of the production equipment on the potassium salt deviation amount during the predetermined time period. The instantaneous elastic information sets corresponding to multiple predetermined time periods included in the production process of the target batch of potassium salt are determined by using the instantaneous elastic information set for the predetermined time period. Based on the instantaneous elastic information sets corresponding to the multiple predetermined time periods, fault tracing is performed on the potash production process of the target batch to determine the faulty production equipment in the potash production process of the target batch.

2. The method according to claim 1, characterized in that, When the plurality of production equipment includes evaporation equipment, crystallization equipment, and centrifugation equipment, determining the target hiding representation corresponding to each of the plurality of production equipment within the predetermined time period based on the production data corresponding to each of the plurality of production equipment includes: Based on the evaporation production data of the evaporation equipment, the water evaporation rate and the instantaneous solute mass are determined; Based on the crystallization production data of the crystallization equipment, the crystallization rate is determined; Based on the water evaporation rate, the instantaneous solute mass, the crystallization rate, and a preset first threshold, the ideal evaporation trajectory of the evaporation device and the ideal crystallization trajectory of the crystallization device are obtained using the recursive Bayesian method. The ideal evaporation trajectory is used to describe the change of the evaporation production data over time under ideal conditions, and the ideal crystallization trajectory is used to describe the change of the crystallization production data over time under ideal conditions. Based on the ideal evaporation trajectory and the ideal crystallization trajectory, the target hidden representations corresponding to the multiple production devices are determined respectively during the predetermined time period.

3. The method according to claim 2, characterized in that, When the target hiding representation is characterized in vector form, determining the target hiding representation corresponding to the plurality of production devices respectively during the predetermined time period based on the ideal evaporation trajectory and the ideal crystallization trajectory includes: Based on the ideal evaporation trajectory and the ideal crystallization trajectory, attribute vectors corresponding to multiple production devices are determined, wherein the attribute vectors are used to represent the production data of the production devices under ideal conditions; Based on the attribute vectors corresponding to the multiple production devices and the preset adjacency matrix, a weighted calculation method is used to determine the summary vectors corresponding to the multiple production devices. The elements in the adjacency matrix represent the coupling strength between the corresponding production devices, and the summary vector is used to represent the production data of the corresponding production device after considering the coupling effect between it and other production devices. Based on the summary vectors corresponding to the multiple production devices and the preset trainable weight matrix, the initial hidden representations corresponding to the multiple production devices are determined. The trainable weight matrix is ​​used to perform linear transformation and data amplification on the data in the summary vector. The initial hidden representations corresponding to the plurality of production equipment are corrected to obtain the target hidden representations corresponding to the plurality of production equipment during the predetermined time period.

4. The method according to claim 3, characterized in that, The step of correcting the initial hidden representations corresponding to the plurality of production equipment to obtain the target hidden representations corresponding to the plurality of production equipment during the predetermined time period includes: Determine the heat residual and material residual during the predetermined time period; For any initial hidden representation of any production equipment among the plurality of production equipment, based on the heat residual, the heat capacity ratio of any production equipment, and a preset second threshold, the initial hidden representation is corrected to obtain a first hidden representation, wherein the heat capacity ratio is used to quantify the contribution of any production equipment to the heat residual. Based on the material residual, the solute content ratio of any production equipment, and a preset third threshold, the initial hidden representation is corrected to obtain a second hidden representation; Based on the first hidden representation and the second hidden representation, determine any target hidden representation of any production equipment during the predetermined time period; The target hiding representations corresponding to the plurality of production devices are determined by determining any one of the target hiding representations.

5. The method according to claim 4, characterized in that, The first hidden representation is obtained by correcting any initial hidden representation based on the heat residual, the heat capacity ratio of any production equipment, and a preset second threshold, including: If the heat residual is less than or equal to the preset second threshold, then any initial hidden representation is determined as the first hidden representation; If the heat residual is greater than the preset second threshold, the initial hidden representation is corrected based on the heat residual and the heat capacity ratio to obtain a first corrected hidden representation; Based on the first corrected hidden representation, the heat residual is updated to obtain the updated heat residual; If the updated heat residual is less than or equal to the preset second threshold, the correction process is stopped, and the first correction hidden representation obtained from the last correction is determined as the first hidden representation.

6. The method according to claim 1, characterized in that, The step of determining the fault trajectory and fault-free trajectory of the potassium salt production process within the predetermined time period based on the target hiding representations corresponding to the multiple production devices includes: Inject fault perturbations into the target hiding representations corresponding to the multiple production devices respectively to obtain perturbation hiding representations corresponding to the multiple production devices respectively; The fault-free trajectory is determined based on the target hiding representations corresponding to the multiple production equipment; The fault trajectory is determined based on the disturbance hiding representations corresponding to the multiple production equipment.

7. The method according to claim 1, characterized in that, The step of determining the potassium salt deviation for each of the multiple production devices within the predetermined time period based on the fault trajectory and the fault-free trajectory includes: Based on the fault trajectory and the fault-free trajectory, the theoretical potassium salt content of the centrifuges included in the plurality of production equipment in the predetermined time period is determined. Based on the theoretical potassium salt content of the centrifuge and the actual potassium salt content of the centrifuge, the potassium salt deviation of the centrifuge during the predetermined time period is determined. Based on the allocation weight, the potassium salt deviation of the centrifuge is backpropagated to obtain the potassium salt deviation corresponding to each of the multiple production equipment in the predetermined time period. The backpropagation starts from the centrifuge and ends at the evaporation equipment included in the multiple production equipment. The allocation weight is used to quantify the material flow and heat flow power among the production equipment.

8. The method according to any one of claims 1 to 7, characterized in that, The step of tracing the fault sources in the potash production process of the target batch based on the instantaneous elastic information sets corresponding to the multiple predetermined time periods, and determining the faulty production equipment in the potash production process of the target batch, includes: For any one of the plurality of production equipment, based on the instantaneous elasticity information sets corresponding to the plurality of predetermined time periods, the instantaneous elasticity information of any one of the production equipment corresponding to the plurality of predetermined time periods is determined. Based on the instantaneous elasticity information corresponding to the multiple predetermined time periods, the cumulative elasticity information of any production equipment in the target batch is determined, wherein the cumulative elasticity information is used to quantify the cumulative impact of any production equipment in the target batch on the potassium salt deviation. Based on the cumulative elasticity information and the output reduction factor of any production equipment, the comprehensive elasticity information of any production equipment in the target batch is determined, wherein the comprehensive elasticity information is used to quantify the degree of influence of any production equipment on the potassium salt deviation in the target batch after considering the output reduction factor; Based on the comprehensive elasticity information, the first correction factor and the second correction factor of any production equipment determine the composite risk coefficient of any production equipment in the target batch, wherein the first correction factor is used to adjust the impact of the failure of any production equipment on the amount of potassium salt deviation, and the second correction factor is used to adjust the impact of the production delay caused by the fault diagnosis of any production equipment on the amount of potassium salt deviation. The composite risk coefficients corresponding to multiple production equipment in the target batch are determined by using the method of determining the composite risk coefficient of any one of the production equipment; Based on the composite risk coefficients corresponding to the multiple production equipment, the faulty production equipment in the potash production process of the target batch is determined.

9. A fault tracing device for a potassium salt production process, characterized in that, include: The data acquisition module is used to collect production data from multiple production devices during a predetermined time period in the potash production process. The first determining module is used to determine the target hidden representation corresponding to the multiple production equipment respectively in the predetermined time period based on the production data corresponding to the multiple production equipment respectively. The target hidden representation is used to describe the production data of the corresponding production equipment after considering the coupling effect between it and other production equipment, as well as the heat conservation and solute conservation. The other production equipment is the production equipment other than the corresponding production equipment among the multiple production equipment. The second determining module is used to determine the fault trajectory and the fault-free trajectory of the potassium salt production process in the predetermined time period based on the target hiding representation corresponding to the plurality of production equipment respectively. The fault trajectory is used to describe the change of production data over time when there is a fault disturbance in the potassium salt production process, and the fault-free trajectory is used to describe the change of production data over time when there is no fault disturbance in the potassium salt production process. The third determining module is used to determine the potassium salt deviation amount corresponding to the multiple production equipment in the predetermined time period based on the fault trajectory and the fault-free trajectory. The fourth determining module is used to determine the instantaneous elasticity information set in the predetermined time period based on the potassium salt deviation amount corresponding to the plurality of production equipment respectively. The instantaneous elasticity information set includes the instantaneous elasticity information corresponding to the plurality of production equipment respectively. The instantaneous elasticity information is used to quantify the instantaneous influence of the production equipment on the potassium salt deviation amount in the predetermined time period. The fifth determining module is used to determine the instantaneous elastic information sets corresponding to multiple predetermined time periods included in the production process of the target batch of potassium salt by using the instantaneous elastic information set of the predetermined time period; The sixth determining module is used to trace the fault source of the potash production process of the target batch based on the instantaneous elastic information sets corresponding to the multiple predetermined time periods, and to determine the faulty production equipment in the potash production process of the target batch.

10. An electronic device, characterized in that, include: One or more processors and a memory, the memory being used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement the fault tracing method for the potash production process according to any one of claims 1 to 8.

Images