A dipotassium hydrogen phosphate storage and dispensing management system

Abstract

The application relates to the technical field of material management, in particular to a dipotassium hydrogen phosphate storage and distribution management system, which comprises a load bias feature extraction module, a turning section risk identification module, a work station passing feature classification module, a path stability evaluation module and a path combination optimization module. In the loading process, the material mass distribution and the gravity center offset state are dynamically extracted and normalized, the turning response change in the unloading path and the unloading end passing offset trend are combined, a path whole-process passing trend sequence is constructed, the passing state is discriminated and the path passing level is evaluated through trend consistency and fluctuation characteristics, and in the path combination stage, the path is optimized based on trend abnormal section replacement and trend sequence combination evaluation value, so that the passing risk state of the dipotassium hydrogen phosphate from the whole process of loading to unloading under the conditions of multiple work stations and multiple paths can be dynamically classified, trend identified and path selection optimized.

Description

Technical Field

[0001] This invention relates to the field of materials management technology, and in particular to a storage and distribution management system for dipotassium hydrogen phosphate. Background Technology

[0002] The field of materials management technology involves the systematic management of various materials throughout the entire process of procurement, transportation, storage, distribution and use. Its core aspects include the classification management of materials, inventory control, optimization of inbound and outbound processes, and material scheduling and allocation strategies, so as to achieve standardized, transparent and efficient allocation of materials.

[0003] The traditional dipotassium hydrogen phosphate storage and distribution management system refers to a systematic approach for the storage and on-demand distribution management of dipotassium hydrogen phosphate, a chemical material widely used in agriculture, industry, and other fields. This includes the management of dipotassium hydrogen phosphate storage containers, material weighing and loading records, batch information registration, usage statistics, and distribution process control.

[0004] Traditional dipotassium hydrogen phosphate storage and distribution management systems primarily focus on material storage, weighing records, and batch registration. They lack the ability to identify material loading imbalances and their impact on subsequent route characteristics. In complex scenarios with multiple routes, workstations, and turning points, they struggle to dynamically perceive load change trends and route response characteristics during transportation. This can lead to frequent vehicle deviations, unstable steering, or abnormal unloading at some routes, affecting the balanced distribution of materials and transportation efficiency. For example, some routes may be reachable in batch scheduling, but uneven loading at the beginning can cause abnormal load responses in the middle, resulting in decreased end-point efficiency or frequent unloading failures. The system cannot provide route status trend judgments or dynamic adjustment criteria, limiting the improvement of overall transportation management capabilities. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of existing technologies by proposing a storage and distribution management system for dipotassium hydrogen phosphate.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a dipotassium hydrogen phosphate storage and distribution management system, the system comprising:

[0007] The off-center loading feature extraction module monitors the mass flow rate of the conveyor belt output section of the silo during the loading process of dipotassium hydrogen phosphate, extracts the axial mass distribution and the offset of the center of gravity, and constructs a loading feature dataset.

[0008] Based on the loading feature dataset, the turning section risk identification module identifies the load response risk change trend of all turning sections in the path to the use area and obtains the turning section risk identification result.

[0009] Based on the risk identification results of the turning section, the station access feature classification module analyzes the access response characteristics of the unloading station section corresponding to the last segment of each path stored to the use area, and obtains the unloading station access offset feature set;

[0010] The path stability assessment module constructs a traffic trend sequence of the complete path to the use area based on the unloading station traffic offset feature set, and obtains the dipotassium hydrogen phosphate management results.

[0011] The present invention is improved in that the loading feature dataset includes axial mass distribution, gravity center offset, and cross-sectional mass flow rate; the turning section risk identification result includes the turning section location and load response risk change trend; the unloading station passage offset feature set includes the unloading station section and passage response characteristics; and the dipotassium hydrogen phosphate management result includes the storage to use area path and passage trend sequence.

[0012] The present invention is improved in that the off-center loading feature extraction module includes:

[0013] The cross-sectional quality acquisition submodule acquires mass flow density data collected from the left, middle and right positions of the silo conveyor belt outlet during the loading of dipotassium hydrogen phosphate. It calculates the difference between the mass flow density values ​​of the left and right sections, normalizes the difference into a width lateral mass offset value according to the width of the conveyor belt, and generates a lateral mass offset value sequence.

[0014] The roller stress analysis submodule obtains the axial stress collected by the stress sensors at the front and rear shafts of the conveying roller under the time period corresponding to the transverse mass offset value sequence, converts the stress value difference at both ends into the longitudinal gravity center offset distance, normalizes it according to the conveying length, and obtains the longitudinal gravity offset value sequence.

[0015] The off-center loading factor generation submodule calculates the ratio of each pair of values ​​based on the time-point data of the lateral mass offset value sequence and the longitudinal gravity offset value sequence, and constructs an offset factor vector. Based on the principal component contribution sorting, it selects a specified off-center loading factor to construct the loading principal vector and obtains the loading feature dataset.

[0016] The present invention is improved in that the turning section risk identification module includes:

[0017] The steering path extraction submodule obtains all steering section numbers stored in the usage area path, and collects the measured front and rear axle load change sequences of vehicles passing through each steering section in the scheduling batch, and constructs the steering section axle load change sequence in chronological order.

[0018] The response difference fitting submodule calls the axle load change sequence of the steering section, matches the offset factor vector sequence in the loading feature dataset within the corresponding time period, calculates the change slope between two adjacent points in each sequence based on a fixed time window, constructs the axle load change slope sequence and the off-center load factor change slope sequence, calculates the difference between the two slopes at the same time point and summarizes them into a trend slope residual sequence.

[0019] The risk trend classification submodule constructs a feature vector based on the mean residual and fluctuation range of each segment in the trend slope residual sequence, and inputs it into the support vector machine to perform risk level discrimination and boundary division, thereby obtaining the risk identification result of the turning segment.

[0020] The present invention is improved in that the workstation access feature classification module includes:

[0021] The unloading station positioning submodule extracts the end unloading station number of each path in the path to the use area, establishes the correlation between the station number and the risk residual curve, and obtains the unloading station segment matching set.

[0022] The passage response difference calculation submodule calls the unloading station section matching set to obtain the residual trend sequence corresponding to the end of each station path in the risk identification result of the turning section. It calls the offset factor vector in the dipotassium hydrogen phosphate loading feature data set to extract the local slope change trend in the same time period, calculates the curve offset direction consistency rate of the two trends in the same window, and generates the station passage trend consistency rate sequence.

[0023] The workstation status classification submodule classifies the response status according to the trend consistency rate of each unloading workstation in the workstation traffic trend consistency rate sequence, sets the classification boundary according to the Bayesian discrimination rule, and outputs the unloading workstation traffic offset feature set.

[0024] The present invention is improved in that the path stability evaluation module includes:

[0025] The path structure reorganization submodule extracts all unloading station numbers in each path stored in the usage area based on the unloading station passage offset feature set, and arranges the corresponding station segment numbers in the path order to generate a path structure sequence set;

[0026] The trend sequence construction submodule calls the path structure sequence set, matches the passage trend status of each workstation segment in the passage offset feature set of the unloading workstation, and combines them to form a path passage trend sequence set;

[0027] The path level assessment submodule performs graded processing based on the trend change direction, number of jumps, and state fluctuation amplitude of each path in the path traffic trend sequence set, and outputs the dipotassium hydrogen phosphate management results.

[0028] The present invention is improved by further including a path combination optimization module, which performs cross-combination and recombination optimization on each traffic trend sequence in the dipotassium hydrogen phosphate management results to obtain an optimized management path;

[0029] The optimized management path includes cross-combined path segments and reorganized path structure.

[0030] The present invention is improved in that the path combination optimization module includes:

[0031] The trend fluctuation screening submodule selects the path segment with the largest overall trend change based on the management results of dipotassium hydrogen phosphate, and records the corresponding storage tank start point and unloading station end point to generate a set of abnormal path change segments.

[0032] The path segment replacement submodule calls the path change abnormal segment set, filters other path segments with the same start and end points from the currently available paths, replaces the abnormal segments in the original path, and then recombines them to generate a path trend combination sequence set.

[0033] The combined scheme ranking submodule calculates the evaluation value of each combined path based on the mean squared error and the maximum single-point trend jump value of each sequence in the path trend combined sequence set, and weights them according to the two indicators with a set proportional factor. It then outputs the path sequence with the smallest evaluation value to obtain the optimized management path.

[0034] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0035] In this invention, by dynamically extracting and normalizing the material mass distribution and center of gravity offset during loading, and combining the changes in turning response and the passage offset trend at the end of unloading, a passage trend sequence is constructed for the entire path. Then, the passage status is judged and the passage level is evaluated by trend consistency and fluctuation characteristics. In the path combination stage, path optimization is completed based on the replacement of abnormal trend segments and the evaluation value of trend sequence combination. This invention enables dynamic classification, trend identification, and path selection optimization of the passage risk status of dipotassium hydrogen phosphate from loading to unloading under multi-station and multi-path conditions. Compared with the traditional management method that focuses solely on weighing records and batch registration, this invention provides more targeted judgment and processing logic in identifying load anomalies, passage bottlenecks, and path instability. At the path allocation decision level, it realizes the quantification of passage status trends and the optimization of path combination strategies, improving the intelligence, stability, and safety of material allocation management in complex scenarios. Attached Figure Description

[0036] Figure 1 This is a system module diagram of the present invention;

[0037] Figure 2 This is a system framework diagram of the present invention;

[0038] Figure 3 This is a schematic diagram of the off-center feature extraction module of the present invention;

[0039] Figure 4 This is a schematic diagram of the turning section risk identification module of the present invention;

[0040] Figure 5 This is a schematic diagram of the workstation access feature classification module of the present invention;

[0041] Figure 6 This is a schematic diagram of the path stability evaluation module of the present invention;

[0042] Figure 7 This is a schematic diagram of the path combination optimization module of the present invention. Detailed Implementation

[0043] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0044] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0045] Please see Figure 1 This invention provides a technical solution: a storage and distribution management system for dipotassium hydrogen phosphate, the system comprising:

[0046] The off-center loading feature extraction module monitors the mass flow rate of the conveyor belt output section of the silo during the loading process of dipotassium hydrogen phosphate, extracts the axial mass distribution and the offset of the center of gravity, and constructs a loading feature dataset.

[0047] The turning section risk identification module identifies the load response risk change trend of all turning sections in the path to the use area based on the loading feature dataset, and obtains the turning section risk identification result.

[0048] Based on the risk identification results of the turning section, the station access feature classification module analyzes the access response characteristics of the unloading station section corresponding to the last segment of each path stored to the use area, and obtains the unloading station access offset feature set;

[0049] The path stability assessment module constructs a path trend sequence of the complete path from the unloading station to the usage area based on the path offset feature set, and obtains the dipotassium hydrogen phosphate management results.

[0050] The loading feature dataset includes axial mass distribution, gravity center offset, and cross-sectional mass flow rate. The risk identification results of the turning section include the location of the turning section and the trend of load response risk changes. The unloading station passage offset feature set includes the unloading station section and passage response characteristics. The dipotassium hydrogen phosphate management results include the storage to use area path and passage trend sequence.

[0051] Please see Figure 2 and Figure 3 The off-center loading feature extraction module includes:

[0052] The cross-sectional quality acquisition submodule acquires mass flow density data collected from the left, middle and right positions of the silo conveyor belt outlet during the loading of dipotassium hydrogen phosphate. It calculates the difference between the mass flow density values ​​of the left and right sections, normalizes the difference into a width lateral mass offset value according to the width of the conveyor belt, and generates a lateral mass offset value sequence.

[0053] To acquire mass flow density data from three points (left, center, and right) at the outlet of the conveyor belt during the loading of dipotassium hydrogen phosphate, three weighing sensor positions (L, M, and R) are first set at the cross-section of the conveyor belt outlet. The mass flow of material passing through each second is collected at these positions. Let's assume the data for a certain second are 1.85 kg / s, 2.00 kg / s, and 2.15 kg / s, respectively. Then, the difference between the flow values ​​at points L and R is calculated, yielding 0.30 kg / s. To eliminate the influence of the conveyor belt width on the degree of offset, the difference is normalized. The conveyor belt width is set to 1.2 m, so the normalized lateral offset value is 0.30 ÷ 1.2 = 0.25 kg / (s·m), which is recorded as the lateral mass offset data point. Time series sampling is then performed to construct a complete data stream. Regarding the threshold setting for offset judgment, it is necessary to analyze the critical value of structural offset not interfering with path passage in historical loading tasks. Select the maximum offset value in 200 stable passage tasks, and statistically calculate the upper limit 95th percentile value as the offset safety threshold. In actual calculation, if the offset corresponding to this percentile value is ±0.10kg / (s·m), then this value is set as the control benchmark. If the current record is 0.25kg / (s·m), it exceeds the allowable boundary. The system marks this period as a lateral off-center loading state and generates a lateral mass offset value sequence.

[0054] The roller stress analysis submodule obtains the axial stress collected by stress sensors at the front and rear shafts of the conveying roller under the corresponding time period of the lateral mass offset value sequence, converts the stress value difference at both ends into the longitudinal gravity center offset distance, normalizes it according to the conveying length, and obtains the longitudinal gravity offset value sequence.

[0055] The axial stress collected by stress sensors at the front and rear axles of the conveyor rollers is obtained during the corresponding time period of the lateral mass offset value sequence. During the acquisition step, the stress measurements of the front and rear axles need to be read simultaneously, set as 186N and 172N, and the difference is calculated to be 14N. Combined with the roller structural characteristics, the center of gravity offset conversion is performed. The roller force transmission structural coefficient can be determined by the simulation model of structural dimensions and loading rotational inertia. Taking the force transmission coefficient of a conventional horizontal light-load roller as 0.025m / N, the offset is calculated as 14×0.025=0.35m. Then, it is normalized according to the total length of the conveyor belt of 10m, 0.35÷10=0.035, that is, the longitudinal offset ratio is 3.5%. The threshold reference method is to retrieve the maximum longitudinal offset ratio that has not triggered the turning section interference event in 100 consecutive runs. If the 95th percentile is 0.02, then 0.02 is set as the longitudinal offset judgment benchmark. The current normalized value is 0.035, which is already at the upper risk level. The system records the longitudinal gravity offset value sequence for this period based on this.

[0056] The off-center loading factor generation submodule calculates the ratio of each pair of values ​​based on the time-point data of the lateral mass offset value sequence and the longitudinal gravity offset value sequence, and constructs an offset factor vector. It selects a specified off-center loading factor based on the principal component contribution sorting to form the loading principal vector and obtains the loading feature dataset.

[0057] Based on the simultaneous time-point data of the lateral mass offset value sequence and the longitudinal gravity offset value sequence, the ratio of each pair of values ​​is calculated to form an offset factor vector. Normalized lateral and longitudinal offset values ​​are extracted at each time point. For example, if the lateral offset is 0.25 kg / (s·m) and the longitudinal offset is 0.035, then the offset factor is 0.25 ÷ 0.035 ≈ 7.14. This factor is dimensionless and represents the coupling degree of lateral and longitudinal offset intensity within the same time period. The sampling frequency is set to 1 Hz, generating one set of offset factors per second, for a total of 100 sets to form the offset factor vector. Principal component sorting of this vector requires normalization. The squared deviation of each factor from the mean of the entire sequence is calculated, and its contribution is statistically analyzed. Principal component truncation is performed according to the cumulative variance contribution rate, with a threshold of 80% cumulative contribution rate. Based on engineering experience, this truncation can retain the main offset structure information. For example, if the sum of the contribution rates of the first 6 factors after sorting is 0.81, then the first 6 factors are truncated to form the principal vector, and the system obtains the loaded feature dataset accordingly.

[0058] Please see Figure 2 and Figure 4 The turning section risk identification module includes:

[0059] The steering path extraction submodule obtains all steering section numbers stored in the usage area path, and collects the measured front and rear axle load change sequences of vehicles passing through each steering section in the scheduling batch, and constructs the steering section axle load change sequence in chronological order.

[0060] The system retrieves all turning segment numbers from the path stored in the usage area and collects the measured front and rear axle load change sequences for each turning segment during the time period when vehicles in the scheduling batch pass through each turning segment. This process requires extracting the path segment numbers marked with turning symbols from all path structure diagrams. Let's say the turning segment numbers in path A are T1, T2, T3, etc. The system extracts the time period from 13:45:00 to 13:45:10 when the vehicle passes through T1 based on the scheduling records, and collects the strain gauge load data of the vehicle's front and rear axles during this time period. The axle load data recorded by the sensor is sampled once per second, for a total of 10 sets of data. The front axle data is as follows: 3450N, 3480N, 3500N... The rear axle data is as follows: 3150N, 3180N, 3205N... After sampling, the two sets of data are merged in time axis order, and the values ​​of the two axes at each time point are recorded to form a structured load sequence, which serves as the axle load change input for section T1. Other paths are collected and constructed in the same way as the structure. The system finally generates the axle load change sequence for the turning section of all numbered paths.

[0061] The response difference fitting submodule calls the axle load change sequence of the steering section, matches the offset factor vector sequence in the loading feature dataset within the corresponding time period, calculates the change slope between two adjacent points in each sequence based on a fixed time window, constructs the axle load change slope sequence and the off-center load factor change slope sequence, calculates the difference between the two slopes at the same time point and summarizes them into a trend slope residual sequence.

[0062] The axle load variation sequence of the steering section is invoked, and the offset factor vector sequence in the loading feature dataset within the corresponding time period is matched. During the execution, the same time period must be used as the pairing reference. Let the time period of section T1 be 13:45:00 to 13:45:10. The offset factor sequence for this time period is retrieved in the loading feature dataset. Let the sampling results be: 7.14, 6.80, 7.01... corresponding to the second. These, along with the load values ​​collected in the axle load sequence, form two sequences of the same length. Then, each set is calculated in units of a fixed time window of 3 seconds. The slope change value between three consecutive points is calculated by dividing the difference between the last term and the first term by the interval length. For example, if the slope = (3500-3450) / 2 = 25 N / s, the corresponding offset factor slope is (7.01-7.14) / 2 = -0.065. A sequence of axial load change slope and an eccentric load factor change slope are generated sequentially. Then, the slope difference between the two sequences is calculated for the corresponding terms at the same time point, i.e., |25-(-0.065)| = 25.065, and recorded as the residual point value for that window. All residual points are then summarized to form a trend slope residual sequence.

[0063] The risk trend classification submodule constructs a feature vector based on the mean residual and fluctuation range of each segment in the trend slope residual sequence. This vector is then input into a support vector machine to determine the risk level and delineate the boundary, thereby obtaining the risk identification result of the turning segment.

[0064] The feature vector is constructed based on the mean and fluctuation range of the residuals in each turning segment of the trend slope residual sequence. This feature vector is then input into a support vector machine (SVM) for risk level determination and boundary delineation, yielding the risk identification results for the turning segments. During execution, firstly, based on the slope difference sequence generated by different turning segments under each path in the trend slope residual sequence, the mean and standard deviation of each residual sequence are extracted as input features to form a two-dimensional feature vector. For example, if the residuals of a turning segment under path A are {0.12-, 0.15-, 0.10-, 0.18-, 0.11}, then its mean is calculated to be μ = 0.132 and its standard deviation to be σ = 0.027, constructing a feature vector x = [0.132, 0.027]. This feature vector is then input into the SVM classification function for classification. The classification function f(x) has the following form:

[0065]

[0066] The function f(x) represents the classification decision function of the support vector machine for the input feature vector x. Its value reflects the relative position of the input sample on the classification plane. Specifically, when the value of f(x) is positive, it means that the sample falls on one side of the classification boundary, and when it is negative, it is on the other side. The further the value is from zero, the more discriminative the classification is. In risk level discrimination, multiple decision functions f(x) can be used to judge risk classification under one-to-many or many-to-many structures. By setting multiple classification boundary thresholds (such as f(x) = 0 between level 1 and 2, and f(x) = a between level 2 and 3), all samples are divided into intervals according to the value of f(x), thereby outputting risk levels 1, 2, 3, etc., to achieve a clear classification of the risk level of the turning segment corresponding to different residual feature vectors.

[0067] Where x is the residual feature vector of the current path segment to be judged, and α i Let y be the Lagrange multiplier for the i-th support vector in the training samples. i For its corresponding risk level label, x i For the feature vectors corresponding to the support vectors, the kernel function used is the radial basis function K(x). i ,x)=exp(-γ||x i -x|| 2 This is used to calculate the relationship between the current input sample x and the support vector x. iThe nonlinear similarity between them reflects their relative distance in the feature space. The kernel width parameter γ is set based on the reciprocal of the mean and standard deviation of the residual sequence. If the mean and standard deviation of the features in all samples are concentrated in the range of [0.05-, 0.25], then the average variance can be approximated as 0.001. The corresponding setting is γ = 1 / 0.001 = 1000. The bias term b is automatically solved by the KKT conditions during training and is set to -1.2.

[0068] If the current classification task uses three support vector samples:

[0069] Support vector sample 1: x1 = [0.140, 0.030], y1 = 2, α1 = 0.85;

[0070] Support vector sample 2: x2 = [0.128, 0.025], y2 = 1, α2 = 0.60;

[0071] Support vector sample 3: x3 = [0.135, 0.029], y3 = 3, α3 = 0.72.

[0072] Calculate the squared Euclidean distance between the input vector x = [0.132, 0.027] and each support vector:

[0073] ||x1-x|| 2 =(0.140-0.132) 2 +(0.030-0.027) 2 =0.000073;

[0074] ||x2-x|| 2 =(0.128-0.132) 2 +(0.025-0.027) 2 =0.00002;

[0075] ||x3-x|| 2 = (0.135-0.132) 2 +(0.029-0.027) 2 =0.000013.

[0076] Substituting into the kernel function respectively, we get:

[0077] K1=exp(-1000·0.000073)≈0.9296;

[0078] K2=exp(-1000·0.00002)≈0.9802;

[0079] K3=exp(-1000·0.000013)≈0.9871.

[0080] Substitute the classification function to calculate:

[0081] f(x)=sign(0.85·2·0.9296+0.60·1·0.9802+0.72·3·0.9871-1.2)=sign(3.096).

[0082] Since the result is greater than 0, the classification function output label is positive. According to the level label mapping rule set in the classification training, the classification result corresponds to risk level 3, indicating that the current turning segment belongs to the high-risk level. This level value will be directly output and written into the turning segment risk identification result vector under the corresponding turning segment of the entire path, such as [2,3,1,2,3], as input to the downstream module to participate in the path offset status classification and path trend combination processing. Each discrimination operation requires recalculating the kernel function value and executing the classification function, ensuring that the risk level result has traceability and a clear calculation path.

[0083] Please see Figure 2 and Figure 5 The workstation access feature classification module includes:

[0084] The unloading station positioning submodule extracts the end unloading station number of each path in the path to the use area, establishes the correlation between the station number and the risk residual curve, and obtains the unloading station segment matching set.

[0085] Extract the endpoint unloading station number of each path stored in the usage area path. Extract the endpoint identifier information of each path according to the path structure table in the current scheduling batch. Extract the number corresponding to the endpoint node from the path table as the unloading station number set. Then, match the unloading station number with the risk trend residual sequence output by the previous turning section risk identification module. For the unloading station number in each path, search for a segment with the same identifier in the residual sequence. If it exists, establish an index connection relationship for the risk residual curve corresponding to the number and record it in the corresponding mapping between unloading station number and residual sequence file. For example, if the endpoint of path A is station K3, check if K3 has residual data in the risk identification data. If it does, record its residual curve and mark path A connecting K3. Finally, after traversing all paths and completing the mapping, obtain the unloading station segment matching set.

[0086] The passage response difference calculation submodule calls the unloading station section matching set to obtain the residual trend sequence corresponding to the risk identification result of the turning section of the last segment of each station path, calls the offset factor vector in the dipotassium hydrogen phosphate loading feature data, extracts the local slope change trend in the same time period, calculates the curve offset direction consistency rate of the two trends under the same window, and generates the station passage trend consistency rate sequence.

[0087] The unloading station segment matching set is called to obtain the residual trend sequence corresponding to the risk identification result of the turning section of the last segment of each station path. Then, the offset factor vector sequence with time overlap with the above residual trend sequence is extracted from the dipotassium hydrogen phosphate loading feature dataset. A fixed-length sliding time window is used, with each calculation window lasting 5 seconds. Within each window, the values ​​of two adjacent time points in the two sets of sequences are extracted, and the slope is calculated separately. The slope is calculated by subtracting the previous value from the current value and then dividing by the time interval. Here, the time interval is fixed at 5 seconds. For example, if the residual trend sequence in the first window is 0.12 to 0.18, and the offset factor trend is 0.09 to 0.15... The residual slope is (0.18-0.12) / 5 = 0.012, and the offset factor slope is (0.15-0.09) / 5 = 0.012. Both slopes are marked as consistent if they are regular, and are set to 1 otherwise. The calculation results of each window are recorded to form a direction consistency mark sequence. In each path, the number of consistent windows is counted and divided by the total number of windows. For example, if a certain unloading station path contains 12 time windows, and 8 of them are consistent in direction, then its trend consistency rate is 8 / 12 = 0.667. This value is the station passage trend consistency rate. This process is performed on all station paths in sequence to finally form a station passage trend consistency rate sequence.

[0088] The workstation status classification submodule classifies the response status based on the trend consistency rate of each unloading workstation in the workstation traffic trend consistency rate sequence, sets the classification boundary according to the Bayesian discrimination rule, and outputs the unloading workstation traffic offset feature set.

[0089] The obtained workstation traffic trend consistency rate sequence is retrieved, with each trend consistency rate value denoted as u. This u represents the degree of consistency between the traffic trend at the end of the path and the loading offset trend within the same time window. Based on this consistency rate value u, the traffic response status of each unloading workstation is classified using Bayesian discrimination rules. Three status categories are defined: Category G... α Indicates a low response state, category G β Indicates the response state, category G γ This indicates a high response state, corresponding to different synchronization intensities between the workstation passage trend and the loading trend.

[0090] Calculate the posterior probability P(G) for each category. λ The Bayesian discriminant formula used is:

[0091]

[0092] Among them, P(G λ |u): Given a trend consistency rate value u, the workstation belongs to category G. λThe posterior probability; P(G) λ Category G λ The prior probability is usually obtained from sample statistics; ∑ θ : Represents all possible categories G θ (i.e., summation is performed on θ∈{α,β,γ}); P(u∣G θ )·P(G θ Category G θ The product of the conditional probability density function for a given sample value of u and the prior probability is used as part of the sum of the denominators; P(G θ The unloading station belongs to state category G. θ The prior probability represents the initial likelihood that a workstation belongs to a certain category without observing the current trend consistency rate value u; P(u|G λ ): In category G λ The conditional probability density of the trend consistency rate value u under given conditions is estimated by the normal distribution; u: Consistency rate of traffic flow at a certain unloading station (i.e., the degree of consistency between the trend direction of this station and the trend direction of the off-center loading factor, usually ranging from 0 to 1); μ λ Category G λ The mean of the sample trend consistency rate is calculated from the mean of the samples in that category; σ λ Category G λ The standard deviation of the sample trend consistency rate is obtained by taking the square root of the variance of the sample in that class; exp(): natural exponential function, representing the power of e; π: pi constant, approximately equal to 3.1416; The normalization factor of the normal distribution makes the area under the curve equal to 1.

[0093] Set sample statistics:

[0094] State category G a Prior probability P(G) a =0.3, mean μ a =0.55, standard deviation σ a =0.10;

[0095] State category G b Prior probability P(G) b =0.5, mean μ b =0.70, standard deviation σ b =0.08;

[0096] State category G c Prior probability P(G) c =0.2, mean μ c =0.85, standard deviation σ c =0.07.

[0097] Calculate the likelihood function value (normal distribution probability density) for each class:

[0098] Category G a :

[0099] Category G b :

[0100] Category G c :

[0101] Calculate the numerator (numerator of the posterior probability) for each class:

[0102] Category G a : 0.942·0.3=0.2826;

[0103] Category G b 4.827 * 0.5 = 2.4135;

[0104] Category G c : 1.001·0.2=0.2002.

[0105] Calculate the denominator (the sum of the numerators of each category):

[0106] ∑ η P(u∣G η )·P(G η = 0.2826 + 2.4135 + 0.2002 = 2.8963.

[0107] Calculate the posterior probability for each class:

[0108]

[0109] Classification Results: When classifying the consistency rate value u of the passage trend at the unloading station, the system sets the classification boundary based on the Bayesian discriminant rule, and calculates the posterior probability P(G) of each category. λ |u) performs calculations to determine the degree of matching between the current unloading station's traffic trend and various states. The classification boundary is set based on the principle of assigning the maximum posterior probability, i.e., if a certain category G λ Corresponding P(G) λ If the value of |u) is the largest, then the workstation is classified into that category; in this implementation scenario, there are three status categories: low response status G a Response state G b High response state G c, representing the degree of agreement between the offset trend of the unloading station and the loading characteristics during material transportation, from weak to strong. If the consistency rate of the passage trend of a certain station is u = 0.72, the posterior probabilities obtained after the above calculations are P(G a |u)=0.0976、P(G b |u)=0.8337、P(G c |u)=0.0691. According to the maximum a posteriori probability rule, the unloading station is determined to belong to the intermediate response state G. b The category indicates that the traffic trend and off-center load characteristics of this workstation are significantly correlated, but it does not exhibit extreme deviation, making it suitable for inclusion in the path stability assessment and analysis path.

[0110] Please see Figure 2 and Figure 6 The path stability assessment module includes:

[0111] The path structure reorganization submodule extracts all unloading station numbers in each path stored in the usage area based on the unloading station passage offset feature set, and arranges the corresponding station segment numbers in the path order to generate a path structure sequence set;

[0112] After receiving the passage offset feature data input by the system, the system first calls the conveyor path configuration library to extract the unloading station numbers on all paths from the storage area to the usage area. During this process, each path needs to be analyzed to confirm its physical conveying flow. Using the conveyor network structure diagram maintained by the system, the system reads the node numbers within each path and extracts their corresponding unloading station numbers. Then, the stations are sorted according to the direction from the start to the end of the path to complete the numbering sequence construction. For example, a path starting from the warehouse may pass through multiple unloading points. The system will arrange the numbers according to the order in which these points appear on the path. Next, the system needs to look up a table to compare the segment affiliation of these stations in the logistics area layout diagram, mapping each station to its corresponding functional segment number, and then forming a segment number sequence according to the path order. The system generates a path structure sequence set. During the construction process, there may be multiple overlapping parts of the path. Therefore, the system will mark each path with an independent path identifier to prevent confusion. If a path switches between three unloading points, but there are two path intersection nodes in the middle, the system will mark the path jump point to ensure that the path identifier is not covered. After the path structure is generated, an integrity check is required. The system will check whether the number of workstations in the path is within the expected range, whether the start and end points are in the correct functional areas, and whether the workstation order conforms to the path topology rules. In a typical application scenario, such as the sorting section of a packaging production line with three paths responsible for the delivery of raw materials, semi-finished products, and finished products, the system needs to organize all the unloading points involved in these three paths in sequence and build a complete path structure sequence set according to the path order.

[0113] The trend sequence construction submodule calls the path structure sequence set, matches the passage trend status of each workstation segment in the passage offset feature set of the unloading workstation, and combines them to form a path passage trend sequence set;

[0114] Using the path structure sequence set generated in the previous stage as input, the system reads the segment numbers of each path one by one and calls the traffic offset feature library to query the trend status of each segment within the statistical period. The traffic trend status is determined by the system based on the change in the frequency of passage at the workstation within a unit of time. For example, if the frequency of a certain unloading workstation increases significantly in the current period compared to the previous period, its trend status is determined to be rising. If the frequency does not change significantly, it is considered stable; if it decreases significantly, it is considered falling. To ensure consistency in the judgment of changes, the system sets a corresponding judgment threshold for each trend status. This threshold is determined by the average passage frequency and standard change range of historical data. For example, if the average passage frequency of a certain segment is 80 times per hour, and its current frequency reaches 100 times per hour... If the trend is 60 times, it is considered to be rising; if it drops to 60 times, it is considered to be falling; otherwise, it is considered stable. In this way, the system traverses all segments in the path structure sequence set, calls the passage offset feature matching process once for each segment number, records the corresponding trend status, and combines them into a path passage trend sequence according to the path order. In this process, the system also needs to determine whether there are abnormal trends. For example, if multiple consecutive segments on a path show drastic reverse changes in passage trends, the marking logic will be triggered to ensure that such paths are analyzed in the subsequent evaluation. For example, if a logistics path traverses four segments, the first three segments show an upward trend, and the last segment suddenly shows a downward trend, then the path is identified as a path with a sudden change risk. Finally, these trend status data are combined and packaged into a path trend sequence set.

[0115] The path level assessment submodule classifies each path according to the trend change direction, number of jumps and state fluctuation amplitude of each path in the path traffic trend sequence set, and outputs the dipotassium hydrogen phosphate management results.

[0116] After obtaining the path traffic trend sequence set, the system first decodes the trend sequence of each path. The system counts the number of trend change types. For example, if a path contains three states—rising, falling, and stable—it is identified as a complex trend path. If it contains only one trend type, it is marked as a single trend path. Next, the system counts the number of trend jumps, i.e., the number of times the trend state changes between adjacent segments of the path. If a path repeatedly changes from rising to falling or stable, or from falling to rising, the number of jumps increases. The system sets a reference range for the number of jumps as a criterion. For example, a path with more than five jumps is considered a high-frequency jump path, while a path with one to two jumps is considered a low-frequency jump path. To further measure path volatility, the system also evaluates the range of differences in traffic frequency between different segments within the path. The system extracts the maximum and minimum frequency of traffic in the starting and ending sections of a path, calculates the absolute difference between them, and compares it with a fluctuation threshold. If the difference exceeds the system's set threshold, such as a fluctuation exceeding 30 times per hour, the path is classified as a path with significant fluctuations; otherwise, it is classified as a path with moderate fluctuations. This fluctuation threshold is preset based on the average and variation standards of the entire path's traffic frequency. For example, if the average traffic frequency of a path is 90 times per hour, the system determines that the fluctuation threshold should be between 20 and 40 based on past data analysis. Based on this, the system sets classification standards for different paths. Finally, the system summarizes the path's trend status type, number of jumps, and fluctuation amplitude, and assigns path level labels according to rules. For example, Level 1 indicates small fluctuations and a simple trend, while Level 3 indicates frequent jumps and a complex trend, forming the final path level evaluation result and outputting the level information for each path.

[0117] Please see Figure 2 and Figure 7 It also includes a path combination optimization module, which performs cross-combination and recombination optimization on each traffic trend sequence in the dipotassium hydrogen phosphate management results to obtain the optimized management path;

[0118] The optimized management path includes cross-combined path segments and reorganized path structure;

[0119] The path combination optimization module includes:

[0120] The trend fluctuation screening submodule selects the path segment with the largest overall trend change based on the management results of dipotassium hydrogen phosphate, and records the corresponding storage tank start point and unloading station end point to generate a set of abnormal path change segments.

[0121] Based on the management results of dipotassium hydrogen phosphate, the system obtains the trend change sequence corresponding to each path. It then extracts the change amplitude of the trend state value of each segment in the path sequence, calculates the overall fluctuation degree of each path, and determines whether the total change in trend state in the path is the largest. By comparing the cumulative changes in trend state across all paths, the path segment with the largest total change is selected as the object to be processed. In practice, it is necessary to first extract the state change between adjacent segments in the path trend sequence. For example, if the path sequence is rising, falling, stable, rising again, then the changes from rising to falling and from falling to stable need to be calculated separately. The system determines the fluctuation levels of adjacent states and converts these levels into standard values, such as using fluctuation levels to represent specific frequency changes. Values ​​are assigned according to the system's preset level table. If the fluctuation level difference is significant, it is identified as a high-fluctuation segment. The values ​​of all fluctuation segments in the entire path are counted and summed. All paths are then sorted, and the path segment with the highest value is considered to be the path segment with the largest overall trend change. The system records the starting point number of the storage tank and the ending point number of the unloading station corresponding to the path segment as path identification information. At the same time, a record is generated, which includes the path ID, starting and ending positions, total change value, and the segment numbers involved.

[0122] The path segment replacement submodule calls the abnormal path segment set, filters other path segments with the same start and end points from the currently available paths, replaces the abnormal segments in the original path, and then recombines them to generate a path trend combination sequence set.

[0123] The system retrieves the set of abnormal path segments. First, based on the start and end information of each abnormal segment, it searches the currently available path database for other path segments with identical start and end points. The system performs matching through a path topology mapping table, requiring candidate path segments to be consistent with the original path in terms of logical connectivity and travel direction. After the search is completed, the path segments that meet the conditions are listed as candidate paths. Then, the segments marked as abnormal in the original path are replaced with the newly selected path segments. The replacement must maintain the continuity of the path structure and the correct numbering order. The replaced paths are then reassembled to generate a new path structure sequence. At the same time, the trend status value of the replaced segments is extracted from the travel trend feature database. The system replaces the original abnormal segment's trend status in the trend sequence and updates the path trend combination sequence. The updated sequence overwrites the original trend sequence record and adds new path segment number information to ensure that the combined path trend data can be accurately identified and subsequently evaluated. In practical application scenarios, for example, in a certain dipotassium hydrogen phosphate storage and transportation system, the trend of segment P3 with path segment number P3 fluctuates greatly, with the start and end points from S1 to W4. The system finds paths P5 and P8 with the same start and end points from the available paths, replaces P3 respectively to form a complete path structure, and then rereads the trend status combination corresponding to P5 and P8, merges it with the trend of the non-abnormal segment of the original path, and forms a new trend combination sequence set.

[0124] The combined scheme ranking submodule calculates the evaluation value of each combined path based on the mean square error and the maximum single-point trend jump value of each sequence in the path trend combined sequence set, and weights them according to the two indicators with a set ratio factor. It then outputs the path sequence with the smallest evaluation value to obtain the optimized management path.

[0125] After receiving the set of path trend combination sequences, two key indicators are calculated for each path. The first is the overall fluctuation of the trend state, calculated as its mean squared error, reflecting the overall stability of the path. The second is the maximum difference in trend values ​​between any two adjacent segments of the path, serving as a measure of the degree of trend jump. Regarding weighting, based on actual operational management needs, the system considers overall fluctuation to have a greater impact on logistics path planning; therefore, a weight of 0.6 is assigned to the trend mean squared error, and a weight of 0.4 is assigned to the single-point jump value. During the evaluation value calculation process… The two indicators are weighted and summed to form the final score of the path, as follows: Evaluation value = Mean squared error × 0.6 + Maximum jump value × 0.4. For example, if the mean squared error of a path is 28 and the maximum jump value is 15, then its evaluation value is calculated as follows: Evaluation value = 28 × 0.6 + 15 × 0.4 = 16.8 + 6.0 = 22.8. The system performs the same calculation on all combined paths accordingly, sorts them by evaluation value from smallest to largest, and the path with the lowest score is the path sequence with the most stable current trend and the smallest jump, which is then used as the final optimized path scheme for replacement and update.

[0126] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A storage and distribution management system for dipotassium hydrogen phosphate, characterized in that: The system includes: The off-center loading feature extraction module monitors the mass flow rate of the conveyor belt output section of the silo during the loading process of dipotassium hydrogen phosphate, extracts the axial mass distribution and the offset of the center of gravity, and constructs a loading feature dataset. Based on the loading feature dataset, the turning section risk identification module identifies the load response risk change trend of all turning sections in the path to the use area and obtains the turning section risk identification result. Based on the risk identification results of the turning section, the station access feature classification module analyzes the access response characteristics of the unloading station section corresponding to the last segment of each path stored to the use area, and obtains the unloading station access offset feature set; The path stability assessment module constructs a path trend sequence of the complete path to the use area based on the unloading station's path offset feature set, and obtains the dipotassium hydrogen phosphate management results. The turning section risk identification module includes: The steering path extraction submodule obtains all steering section numbers stored in the usage area path, and collects the measured front and rear axle load change sequences of vehicles passing through each steering section in the scheduling batch, and constructs the steering section axle load change sequence in chronological order. The response difference fitting submodule calls the axle load change sequence of the steering section, matches the offset factor vector sequence in the loading feature dataset within the corresponding time period, calculates the change slope between two adjacent points in each sequence based on a fixed time window, constructs the axle load change slope sequence and the off-center load factor change slope sequence, calculates the difference between the two slopes at the same time point and summarizes them into a trend slope residual sequence. The risk trend classification submodule constructs a feature vector based on the mean residual and fluctuation range of each segment in the trend slope residual sequence, and inputs it into the support vector machine to perform risk level discrimination and boundary division, thereby obtaining the risk identification result of the turning segment. The workstation access feature classification module includes: The unloading station positioning submodule extracts the end unloading station number of each path stored in the usage area path, extracts the end point identification information of each path according to the path structure table in the current scheduling batch, extracts the number corresponding to the end point node from the path table as the unloading station number set, matches the unloading station number with the risk trend residual sequence, and searches for whether there is a segment with the same identification number in the residual sequence for the unloading station number in each path. If it exists, an index connection relationship is established for the risk residual curve corresponding to the number to obtain the unloading station segment matching set. The passage response difference calculation submodule calls the unloading station section matching set to obtain the residual trend sequence corresponding to the end of each station path in the risk identification result of the turning section. It calls the offset factor vector in the dipotassium hydrogen phosphate loading feature data set to extract the local slope change trend in the same time period, calculates the curve offset direction consistency rate of the two trends in the same window, and generates the station passage trend consistency rate sequence. The workstation status classification submodule classifies the response status according to the trend consistency rate of each unloading workstation in the workstation traffic trend consistency rate sequence, sets the classification boundary according to the Bayesian discrimination rule, and outputs the unloading workstation traffic offset feature set.

2. The dipotassium hydrogen phosphate storage and distribution management system according to claim 1, characterized in that: The loading feature dataset includes axial mass distribution, gravity center offset, and cross-sectional mass flow rate. The turning section risk identification results include the turning section location and load response risk change trend. The unloading station passage offset feature set includes the unloading station section and passage response characteristics. The dipotassium hydrogen phosphate management results include the storage to use area path and passage trend sequence.

3. The dipotassium hydrogen phosphate storage and distribution management system according to claim 1, characterized in that: The off-center loading feature extraction module includes: The cross-sectional quality acquisition submodule acquires mass flow density data collected from the left, middle and right positions of the silo conveyor belt outlet during the loading of dipotassium hydrogen phosphate. It calculates the difference between the mass flow density values ​​of the left and right sections, normalizes the difference into a width lateral mass offset value according to the width of the conveyor belt, and generates a lateral mass offset value sequence. The roller stress analysis submodule obtains the axial stress collected by the stress sensors at the front and rear shafts of the conveying roller under the time period corresponding to the transverse mass offset value sequence, converts the stress value difference at both ends into the longitudinal gravity center offset distance, normalizes it according to the conveying length, and obtains the longitudinal gravity offset value sequence. The off-center loading factor generation submodule calculates the ratio of each pair of values ​​based on the time-point data of the lateral mass offset value sequence and the longitudinal gravity offset value sequence, and constructs an offset factor vector. Based on the principal component contribution sorting, it selects a specified off-center loading factor to construct the loading principal vector and obtains the loading feature dataset.

4. The dipotassium hydrogen phosphate storage and distribution management system according to claim 1, characterized in that: For risk level assessment and boundary delineation, the following formula is used: ; Calculate the support vector machine on the input feature vector Classification decision function All samples were sorted according to The numerical values ​​are used to define the boundaries, and the corresponding risk levels are output. in, This is the residual feature vector of the current path segment to be judged. For the training sample, the first Lagrange multipliers of the support vectors, Its corresponding risk level label, The feature vectors corresponding to the support vectors. This is the current input sample. With support vectors Nonlinear similarity between them It is a bias term.

5. The dipotassium hydrogen phosphate storage and distribution management system according to claim 1, characterized in that: To classify response states by setting classification boundaries, the formula is as follows: ; Calculate the consistency rate value under a given trend. In this case, the workstation belongs to the category posterior probability The classification boundary is set based on the principle of attribution of the maximum posterior probability, to determine the degree of matching between the current unloading station traffic trend and each state. in, It is a category The prior probability, It represents all possible categories. Summation, In the category The trend consistency rate value appears under the condition The conditional probability density, The unloading station belongs to the status category. The prior probability represents the probability of consistency with the current trend without observing the current trend value. In this case, the initial probability that the workstation belongs to a certain category is... Indicate category The conditional probability density of the sample value u.

6. The dipotassium hydrogen phosphate storage and distribution management system according to claim 1, characterized in that: The path stability assessment module includes: The path structure reorganization submodule extracts all unloading station numbers in each path stored in the usage area based on the unloading station passage offset feature set, and arranges the corresponding station segment numbers in the path order to generate a path structure sequence set; The trend sequence construction submodule calls the path structure sequence set, matches the passage trend status of each workstation segment in the passage offset feature set of the unloading workstation, and combines them to form a path passage trend sequence set; The path level assessment submodule performs graded processing based on the trend change direction, number of jumps, and state fluctuation amplitude of each path in the path traffic trend sequence set, and outputs the dipotassium hydrogen phosphate management results.

7. The dipotassium hydrogen phosphate storage and distribution management system according to claim 1, characterized in that: It also includes a path combination optimization module, which performs cross-combination and recombination optimization on each traffic trend sequence in the dipotassium hydrogen phosphate management results to obtain an optimized management path; The optimized management path includes cross-combined path segments and reorganized path structure.

8. The dipotassium hydrogen phosphate storage and distribution management system according to claim 7, characterized in that: The path combination optimization module includes: The trend fluctuation screening submodule selects the path segment with the largest overall trend change based on the management results of dipotassium hydrogen phosphate, and records the corresponding storage tank start point and unloading station end point to generate a set of abnormal path change segments. The path segment replacement submodule calls the path change abnormal segment set, filters other path segments with the same start and end points from the currently available paths, replaces the abnormal segments in the original path, and then recombines them to generate a path trend combination sequence set. The combined scheme ranking submodule calculates the evaluation value of each combined path based on the mean squared error and the maximum single-point trend jump value of each sequence in the path trend combined sequence set, and weights them according to the two indicators with a set proportional factor. It then outputs the path sequence with the smallest evaluation value to obtain the optimized management path.

Images