An intelligent gas generating agent production management system

Abstract

The present invention relates to the technical field of production management, specifically to an intelligent gas generating agent production management system, the system includes: a raw material monitoring module collects operating temperature from a chemical reactor and a mixer, monitors pressure and output rate, records equipment response time, measures equipment operation frequency, analyzes equipment operation characteristics, and screens key performance indicators. In the present invention, by real-time monitoring of key parameters of chemical reactions and using data analysis technology, the key performance indicators in the production process are effectively mastered, thereby achieving accurate prediction of production demand and potential failures. In addition, by constructing a dynamic graph of dependencies between devices and using graph convolutional network learning technology, the understanding of device interactions is further enhanced, and the efficiency of fault diagnosis and risk prediction is optimized. This strategy significantly reduces downtime caused by failures, improves equipment operation efficiency, ensures consistency of product quality, saves costs for enterprises and enhances market competitiveness.

Description

Technical Field

[0001] The present invention relates to the technical field of production management, and in particular to an intelligent gas generating agent production management system. Background Art

[0002] The field of production management technology covers methods and practices that use computer systems, software applications, and other information technologies to optimize production and manufacturing processes. The purpose of this field is to improve production efficiency, reduce costs, and enhance product quality and consistency through automation and intelligent means. Production management systems can coordinate and control every aspect from raw material procurement, process flow, equipment scheduling to final product inspection and inventory management. Modern production management systems also integrate data analysis and real-time monitoring capabilities, allowing companies to respond to problems on the production line in real time, optimize resource allocation, and improve the transparency and traceability of overall operations.

[0003] Among them, the intelligent gas generating agent production management system uses intelligent technology to manage and optimize the production process of gas generating agents. The main uses include automated control of production processes, ensuring the quality standards of gas generating agents, and predicting and adjusting production needs through precise data analysis. By integrating sensors, control units, and user interfaces, real-time monitoring of production equipment and precise control of production parameters (such as temperature, pressure, chemical ratio, etc.) can be achieved to optimize output and reduce waste. The application of the system helps production companies improve operational efficiency, reduce production costs, and enhance market competitiveness.

[0004] Existing technologies are inadequate in responding quickly to subtle changes and complex equipment interactions in the production process, and often fail to effectively identify or predict potential fault propagation paths. The lack of efficient dynamic analysis and prediction capabilities leads to delays in fault handling, which can quickly affect the operating efficiency and production costs of the entire production line. For example, without real-time equipment behavior monitoring and in-depth dependency analysis, once an equipment failure occurs, it spreads without timely intervention, leading to reduced production efficiency and increased operating costs. In addition, traditional methods are often limited in equipment performance optimization and data-driven decision support, which limits their ability to improve operational efficiency and reduce resource waste. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose an intelligent gas generating agent production management system.

[0006] In order to achieve the above object, the present invention adopts the following technical solution: an intelligent gas generating agent production management system comprises:

[0007] The raw material monitoring module collects operating temperature, monitors pressure and output rate, records equipment response time, measures equipment operation frequency, analyzes equipment operation characteristics, and screens key performance indicators from chemical reactors and mixers;

[0008] The device interaction graph module uses the key performance indicators to construct a dynamic interaction graph between devices, takes the devices as nodes of the graph, defines the dependency relationships as edges, performs graph structure optimization processing, and obtains an optimized dynamic graph;

[0009] The dependency analysis module imports the optimized dynamic graph, learns the interaction between nodes through the graph convolution network, identifies the dependency relationship between nodes, obtains the device dependency matrix, predicts the fault diffusion path according to the device dependency matrix, and generates an abnormal prediction path;

[0010] The abnormality identification module uses the abnormal prediction path to monitor the device behavior, identify the operation mode that deviates from the normal, locate the starting point of the fault, analyze the risk nodes associated with the fault, and generate risk node analysis results.

[0011] As a further solution of the present invention, the step of obtaining the key performance indicators is specifically as follows:

[0012] The operating temperature T and monitoring pressure P are collected every 10 minutes from the chemical reactor and the mixer, using the formula:

[0013]

[0014] Calculate the average operating temperature and average pressure per hour to obtain the average operating data D per hour avg ;

[0015] Among them, T i , P i Represent the temperature and pressure measured every 10 minutes, n is the number of measurements per hour; record the equipment output rate R and equipment response time τ per hour, using the formula:

[0016]

[0017] Get the average output rate and average response time per hour, and generate the equipment response data R per hour avg and τ avg ;

[0018] Among them, R i , τ i Respectively represent the output rate and response time per hour, n is the number of measurements; the operating temperature, monitoring pressure, output rate and equipment response time are analyzed using the formula:

[0019]

[0020] Calculate and generate key performance indicators (KPIs);

[0021] Among them, T avg , P avg , R avg , τ avg They are the average operating temperature, average pressure, average output rate and average response time respectively.

[0022] As a further solution of the present invention, the step of obtaining the optimized dynamic graph is specifically as follows:

[0023] Based on the key performance indicators (KPIs) of each device i , assigning a unique node identifier ID to each device i And calculate the initial weight of the node using the formula:

[0024]

[0025] Generate the normalized weight of the device node;

[0026] Where KPI represents the key performance indicator of the i-th device, n is the total number of devices, α is the normalization parameter, and W i is the normalized weight of the device node;

[0027] Using the normalized weight W i , define the dependency between devices. If device A affects the performance of device B, then an edge E is established between A and B. AB , using the formula:

[0028]

[0029] Determine the edge weights and generate a dependency graph between devices;

[0030] Among them, W AB is the weight of the edge, β adjusts the sensitivity of the dependency;

[0031] According to the dependency graph between the devices, the graph structure is optimized using the formula:

[0032]

[0033] Adjust the graph structure to generate an optimized dynamic interaction graph;

[0034] Among them, W total is the overall weight of the graph, E is the set of edges in the graph, and |W in the formula i -W j | represents the difference in node weights.

[0035] As a further solution of the present invention, the step of obtaining the device dependency matrix is ​​specifically as follows:

[0036] Import the optimized dynamic graph, process it through the graph convolutional network, identify the node relationship between devices, and use the formula:

[0037]

[0038] Calculate the local structural sensitivity of each node i and generate the structural sensitivity result;

[0039] Among them, S i represents the structural sensitivity of node i, N(i) is the set of nodes connected to node i, and d ij is the distance between node i and node j, W ij For edge E ij The weight of , λ is the attenuation parameter;

[0040] Based on the structural sensitivity results, the device dependency matrix M is constructed using the formula:

[0041]

[0042] Calculate the dependency relationship for each pair of devices i and j and generate a device dependency matrix;

[0043] Among them, M ij is the element in the dependency matrix, ∑ k∈N(i) S k It is the sum of the structural sensitivities of the nodes directly connected to point i.

[0044] As a further solution of the present invention, the step of acquiring the abnormal prediction path is specifically as follows:

[0045] Starting from the device dependency matrix, use the formula:

[0046]

[0047] Calculate the initial influence of each device and generate an initial influence vector;

[0048] Among them, I i represents the initial influence of node i, α, β, and γ are the linear adjustment coefficient of distance, the base of exponential decay, and the rate of exponential decay, respectively.

[0049] The initial influence vector is used to simulate the propagation process of the fault between devices. The dynamic system model is used for simulation, and the formula is adopted:

[0050] F t =M×F t；1 +δ·(1-F t；1 )

[0051] Track the fault propagation path between differentiated devices and generate fault propagation vectors;

[0052] Among them, F t 、F t；1 They represent the device response states at consecutive time points, δ is the system recovery rate, and M is the device dependency matrix;

[0053] The fault propagation vector is analyzed over a long period of time to identify the key fault propagation paths and key impact nodes, using the formula:

[0054]

[0055] Evaluate the maximum impact of each device during the entire simulation period and generate abnormal prediction paths;

[0056] Wherein, P represents the maximum impact of multiple devices during the simulation period.

[0057] As a further solution of the present invention, the step of obtaining the risk node analysis result is specifically:

[0058] The abnormal prediction path is used to monitor the device behavior, using the formula:

[0059]

[0060] Calculate the deviation index D of device i i , generating a deviation index for equipment that deviates from normal operation;

[0061] Among them, D i represents the deviation index of device i, μ j represents the average operating level of the equipment, α adjusts the sensitivity of the deviation index, P ij is the dependence of device i on device j, X j is the current operation data of device j;

[0062] Using the deviation index, the fault starting point and the associated risk node are determined using the formula:

[0063]

[0064] Generate a risk level for each device;

[0065] Among them, R i is the risk level of device i, and λ controls the sensitivity of risk rating;

[0066] Analyze the risk levels and identify high-risk nodes using the formula:

[0067] F i =sgn(Ri -θ)·log(1+R i )

[0068] When the risk level exceeds θ, device i is classified as a high-risk node and the risk node analysis results are generated;

[0069] Among them, F i is the risk node, θ is the risk threshold, and sgn is the sign function.

[0070] Compared with the prior art, the advantages and positive effects of the present invention are:

[0071] In the present invention, by real-time monitoring of key parameters of chemical reactions and using data analysis technology, key performance indicators in the production process are effectively mastered, thereby achieving accurate prediction of production demand and potential failures. In addition, by constructing a dynamic graph of dependencies between devices and using graph convolutional network learning technology, the understanding of device interactions is further enhanced, and the efficiency of fault diagnosis and risk prediction is optimized. This strategy significantly reduces downtime caused by failures, improves equipment operating efficiency, ensures consistency in product quality, saves costs for enterprises and enhances market competitiveness. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] Figure 1 is a system flow chart of the present invention;

[0073] Figure 2 A flowchart of the steps for obtaining the key performance indicators of the present invention;

[0074] Figure 3 This is a flow chart of the steps for obtaining the optimized dynamic graph of the present invention;

[0075] Figure 4 A flow chart of the steps for obtaining the device dependency matrix of the present invention;

[0076] Figure 5 A flowchart of the steps for obtaining an abnormal prediction path of the present invention;

[0077] Figure 6 This is a flow chart of the steps for obtaining the risk node analysis results of the present invention. DETAILED DESCRIPTION

[0078] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0079] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, in the description of the present invention, "multiple" means two or more, unless otherwise clearly and specifically defined.

[0080] Embodiment 1

[0081] See also Figure 1 , an intelligent gas generating agent production management system includes:

[0082] The raw material monitoring module collects operating temperature, monitors pressure and output rate, records equipment response time, measures equipment operation frequency, analyzes equipment operation characteristics, and screens key performance indicators from chemical reactors and mixers;

[0083] The device interaction graph module uses key performance indicators to build a dynamic interaction graph between devices, taking the devices as nodes of the graph, defining the dependencies as edges, and performing graph structure optimization to obtain an optimized dynamic graph.

[0084] The dependency analysis module imports the optimized dynamic graph, learns the interaction between nodes through the graph convolution network, identifies the dependency relationship between nodes, obtains the device dependency matrix, predicts the fault diffusion path based on the device dependency matrix, and generates the abnormal prediction path;

[0085] The anomaly identification module uses the anomaly prediction path to monitor device behavior, identify operating modes that deviate from the norm, locate the starting point of the fault, analyze the risk nodes associated with the fault, and generate risk node analysis results.

[0086] Key performance indicators include equipment response time, operating frequency, and operating temperature. The optimized dynamic graph includes equipment nodes, dependency edges, and graph structure optimization details. The equipment dependency matrix includes node relationship strength, fault impact assessment, and fault propagation probability. The risk node analysis results include risk rating, key fault nodes, and fault impact range.

[0087] See also Figure 2 , the specific steps for obtaining key performance indicators are:

[0088] The operating temperature T and monitoring pressure P are collected every 10 minutes from the chemical reactor and the mixer, using the formula:

[0089]

[0090] Calculate the average operating temperature and average pressure per hour to obtain the average operating data D per hour avg ;

[0091] Among them, T i , P i Represent the temperature and pressure measured every 10 minutes, n is the number of measurements per hour; record the equipment output rate R and equipment response time τ per hour, using the formula:

[0092]

[0093] Get the average output rate and average response time per hour, and generate the equipment response data R per hour avg and τ avg ;

[0094] Among them, R i , τ i Represent the output rate and response time per hour respectively, and n is the number of measurements;

[0095] Analyze operating temperature, monitoring pressure, output rate and equipment response time using the formula:

[0096]

[0097] Calculate and generate key performance indicators (KPIs);

[0098] Among them, T avg , P avg , R avg , τ avg They are the average operating temperature, average pressure, average output rate and average response time respectively.

[0099] formula:

[0100]

[0101] Parameter acquisition and calculation

[0102] Average operating temperature T avg :

[0103] Collect data: Assume that the temperature of the reactor is recorded every 10 minutes for one hour, for a total of 6 data points.

[0104] Data points: T = {150, 155, 152, 158, 151, 153} (unit: degrees Celsius)

[0105] Calculate the average:

[0106] Average monitoring pressure P avg :

[0107] Collect data: Again, record the pressure every 10 minutes.

[0108] Data points: P = {10, 12, 11, 13, 12, 11} (unit: bar, abbreviated as bar)

[0109] Calculate the average:

[0110] Average output rate R avg :

[0111] Data collected: Output rate data recorded hourly.

[0112] Data points: R = {50, 48, 49, 52, 51, 50} (unit: kilograms per hour)

[0113] Calculate the average:

[0114] Average response time τ avg :

[0115] Data collected: The time it takes for the device to respond after each change.

[0116] Data point: τ = {2, 2, 3, 2, 4, 3} (unit: seconds)

[0117] Calculate the average:

[0118] Calculation process and examples

[0119] Now use the above data to calculate the KPI:

[0120] Temperature contribution: 0.3×(153.17) 2 =0.3×23464.51=7039.35

[0121] Pressure contribution:

[0122] Output Contribution:

[0123] Response time contribution: 0.2×2.67=0.534

[0124] Integrate the above results into KPI calculation:

[0125]

[0126] Result interpretation:

[0127] The calculated KPI value is 1918.76, which comprehensively reflects the overall performance of the equipment under given conditions, including temperature stability, pressure suitability, output efficiency and response time. A high KPI value indicates that the equipment performs well in these aspects, while a low value is expected to indicate performance problems.

[0128] See also Figure 3 , the steps to obtain the optimized dynamic graph are as follows:

[0129] Key performance indicators (KPIs) for each device i , assigning a unique node identifier ID to each device i And calculate the initial weight of the node using the formula:

[0130]

[0131] Generate the normalized weight of the device node;

[0132] Where KPI represents the key performance indicator of the i-th device, n is the total number of devices, α is the normalization parameter, and W i is the normalized weight of the device node;

[0133] Using the normalized weight W i , define the dependency between devices. If device A affects the performance of device B, then an edge E is established between A and B. AB , using the formula:

[0134]

[0135] Determine the edge weights and generate a dependency graph between devices;

[0136] Among them, W AB is the weight of the edge, β adjusts the sensitivity of the dependency;

[0137] According to the dependency graph between devices, the graph structure is optimized using the formula:

[0138]

[0139] Adjust the graph structure to generate an optimized dynamic interaction graph;

[0140] Among them, W total is the overall weight of the graph, E is the set of edges in the graph, and |W in the formula i -W j | represents the difference in node weights.

[0141] Step 1: Node weight calculation

[0142] formula:

[0143]

[0144] Parameter explanation:

[0145] KPI i : Key performance indicators of the ith device.

[0146] α: Regularization parameter used to adjust the impact of the KPI value.

[0147] W i : The weight of the device node.

[0148] n: total number of devices.

[0149] Derivation process:

[0150] Assume that there are three devices with KPI values ​​of 30, 40, and 50 respectively. Set α = 10. Calculate the weight of each device as follows:

[0151] Calculate the index-adjusted value for each KPI:

[0152]

[0153] Calculate the denominator of the weight (the sum of all adjustments):

[0154]

[0155] Calculate the weight W of each device i :

[0156]

[0157] Step 2: Calculate the edge weights of the inter-device dependency graph

[0158] formula:

[0159]

[0160] Parameter explanation:

[0161] W A , W B : The weight of the two connected nodes.

[0162] β: Parameter that adjusts the sensitivity of the dependency.

[0163] W AB : The weight of the edge.

[0164] Derivation process:

[0165] Assume β = 1.5. Consider devices A and B, with weights of 0.09 and 0.24 respectively. The weight of edge AB is calculated as follows:

[0166]

[0167] Step 3: Calculation of overall dependency weight of dynamic interaction graph

[0168] formula:

[0169]

[0170] Parameter explanation:

[0171] W total : The overall weight of the graph.

[0172] E: The set of all edges in the graph.

[0173] W i , W j : The weights of nodes i and j.

[0174] Derivation process:

[0175] Consider a simple graph with only edge AB. Using the previous weight calculation results:

[0176]

[0177] This value W total It represents the overall risk measure of the dependencies in the graph. The smaller it is, the more stable the system is.

[0178] See also Figure 4 , the specific steps for obtaining the device dependency matrix are:

[0179] Import the optimized dynamic graph and process it through the graph convolutional network to identify the node relationship between devices using the formula:

[0180]

[0181] Calculate the local structural sensitivity of each node i and generate the structural sensitivity result;

[0182] Among them, S i represents the structural sensitivity of node i, N(i) is the set of nodes connected to node i, and d ij is the distance between node i and node j, W ij For edge E ij The weight of , λ is the attenuation parameter;

[0183] Based on the structural sensitivity results, the device dependency matrix M is constructed using the formula:

[0184]

[0185] Calculate the dependency relationship for each pair of devices i and j and generate a device dependency matrix;

[0186] Among them, M ij is the element in the dependency matrix, ∑ k∈N(i) S k It is the sum of the structural sensitivities of the nodes directly connected to point i.

[0187] formula:

[0188]

[0189] Parameter explanation:

[0190] S i : Structural sensitivity of node i.

[0191] N(i): The set of nodes directly connected to node i.

[0192] d ij : The distance between nodes i and j.

[0193] W ij : The edge weight between nodes i and j.

[0194] λ: Regularization parameter used to adjust the sensitivity ratio.

[0195] Calculation process and examples:

[0196] Suppose there is a small network where node i is connected to two other nodes j 1 and j 2 Directly connected. λ=1.

[0197] Calculate i to j 1 Contributions:

[0198]

[0199] Calculate i to j 2 Contributions:

[0200]

[0201] Add up all contributions to get S i :

[0202] S i =0.03385+0.03984=0.07369

[0203] Result description:

[0204] This S i = 0.07369 indicates that the structural sensitivity of node i in this network is determined by its distance to adjacent nodes and edge weights.

[0205] Step 2: Construction of device dependency matrix M

[0206] formula:

[0207]

[0208] Parameter explanation:

[0209] M ij : The dependence strength of device i on device j.

[0210] S i , S j : Structural sensitivity of devices i and j.

[0211] N(i): The set of devices directly connected to device i, used for normalization.

[0212] Calculation process and examples:

[0213] Continue using the above S i and assuming that S j = 0.045 and the same network:

[0214] Accumulate all S of i k (include and ):

[0215]

[0216] Calculate M ij :

[0217]

[0218] Result description:

[0219] Here M ij =0.045 indicates the degree of dependence of device i on device j, indicating that j plays a relatively large role in the propagation of fault i.

[0220] See also Figure 5 ,The specific steps for obtaining the abnormal prediction path are:

[0221] Starting with the device dependency matrix, use the formula:

[0222]

[0223] Calculate the initial influence of each device and generate an initial influence vector;

[0224] Among them, I i represents the initial influence of node i, α, β, and γ are the linear adjustment coefficient of distance, the base of exponential decay, and the rate of exponential decay, respectively.

[0225] The initial influence vector is used to simulate the propagation process of faults between devices. The dynamic system model is used for simulation, and the formula is adopted:

[0226] F t =M×F t；1 +δ·(1-F t；1 )

[0227] Track the fault propagation path between differentiated devices and generate fault propagation vectors;

[0228] Among them, F t 、F t；1 They represent the device response states at consecutive time points, δ is the system recovery rate;

[0229] Perform long-term analysis on the fault propagation vector to identify the key fault propagation path and key impact nodes using the formula:

[0230]

[0231] Evaluate the maximum impact of each device during the entire simulation period and generate abnormal prediction paths;

[0232] Wherein, P represents the maximum impact of multiple devices during the simulation period.

[0233] Detailed explanation and example of step 1 formula

[0234] formula:

[0235]

[0236] Parameter explanation:

[0237] I i : The initial influence of node i, indicating its influence in the network.

[0238] M ij : The dependency weight of device i on device j in the device dependency matrix.

[0239] d ij : The distance or connection strength between devices i and j.

[0240] α, β, γ: adjustment parameters used to adjust the impact of distance on dependency weights.

[0241] Assumed values:

[0242] Total number of devices n = 3

[0243]

[0244]

[0245] α=0.5, β=0.3, γ=0.7

[0246] Calculation process:

[0247] For node 1, I 1 The calculation is as follows:

[0248]

[0249] Result interpretation:

[0250] I 1 =0.366 means that device 1 has a higher influence in the network and can affect the status of other devices to a greater extent.

[0251] Detailed explanation and example of step 2 formula

[0252] formula:

[0253] F t =M×F t；1 +δ·(1-F t；1 )

[0254] Parameter explanation:

[0255] F t : Fault state vector at time t.

[0256] M: Device dependency matrix.

[0257] δ: System recovery rate, which indicates the system's ability to recover from failures.

[0258] Assumed values:

[0259] Initial state F 0 =[1,0,0] T (Device 1 is initially in a faulty state)

[0260] δ=0.1

[0261] Calculation process:

[0262]

[0263] Result interpretation:

[0264] F 1 The figure shows the fault status of each device after a fault propagation cycle. Device 3 is most affected, with a value of 0.4, indicating that the fault has a greater probability of spreading in device 3.

[0265] Detailed explanation and example of the formula for step 3

[0266] formula:

[0267]

[0268] Parameter explanation:

[0269] P: The final abnormal prediction path, indicating the maximum impact of each device in all simulation cycles.

[0270] Calculation process:

[0271] Assumption F t The values ​​after 3 cycles are as follows:

[0272]

[0273] Result interpretation:

[0274] P shows the maximum impact of each device during the entire fault simulation. Device 3 has the highest maximum impact, reaching 0.524, indicating that device 3 is the most vulnerable node when a fault occurs.

[0275] See also Figure 6 , the specific steps for obtaining the risk node analysis results are:

[0276] Use the abnormal prediction path to monitor device behavior, using the formula:

[0277]

[0278] Calculate the deviation index D of device i i , generating a deviation index for equipment that deviates from normal operation;

[0279] Among them, D i represents the deviation index of device i, μ j represents the average operating level of the equipment, and α adjusts the sensitivity of the deviation index;

[0280] Using the deviation index, determine the fault starting point and associated risk nodes using the formula:

[0281]

[0282] Generate a risk level for each device;

[0283] Among them, R i is the risk level of device i, and λ controls the sensitivity of risk rating;

[0284] Analyze the risk level and identify high-risk nodes using the formula:

[0285] F i =sgn(R i -θ)·log(1+R i )

[0286] When the risk level exceeds θ, device i is classified as a high-risk node and the risk node analysis result is generated; where F i is the risk node, θ is the risk threshold, and sgn is the sign function.

[0287] Detailed explanation and derivation of the formula in step 1

[0288] formula:

[0289]

[0290] Parameter Description:

[0291] P ij : The dependence of device i on device j.

[0292] X j : Current operation data of device j.

[0293] μ j : Average operation data of device j.

[0294] α: Nonlinear influence coefficient, used to adjust the sensitivity of deviation from the exponent.

[0295] Calculation example

[0296] Assume there are three devices:

[0297] P = [0.50.30.2]

[0298] X = [100150120] (current operation data of the device)

[0299] μ=[80140100](average operation data of the device)

[0300] α=2

[0301] Calculation process:

[0302]

[0303] D 1 =0.5×1.5625+0.3×1.1476+0.2×1.44

[0304] D 1 =0.78125+0.34428+0.288

[0305] D 1 =1.41353

[0306] explain:

[0307] D 1The value of is 1.41353, which indicates the total deviation of the operating data of device 1 from its average value. 1 The value indicates that there is a greater risk of abnormal operation or failure of the device 1.

[0308] Detailed explanation and derivation of the formula in step 2

[0309] formula:

[0310]

[0311] Parameter Description:

[0312] D i : Deviation index of device i calculated from step 1.

[0313] λ: risk amplification factor.

[0314] Calculation example

[0315] Continuing with the results from step 1, assume that:

[0316] D = [1.41353 1.0 0.8]

[0317] λ=0.1

[0318] Calculation process:

[0319]

[0320] R 1 =0.44026×0.131476

[0321] R 1 =0.05787

[0322] explain:

[0323] R 1 The value of is 0.05787, which reflects the relative risk level of device 1. A smaller value indicates that its risk is relatively low but non-zero, indicating that device 1 is in a low risk state in the current fault prediction model.

[0324] Detailed explanation and derivation of the formula in step 3

[0325] formula:

[0326] F i =sgn(R i -θ)·log(1+R i )

[0327] Parameter Description:

[0328] R i : Risk level of device i calculated from step 2.

[0329] θ: Risk threshold, used to determine which devices have a risk level high enough to be considered a critical risk node.

[0330] sgn: symbolic function, used to determine R i Whether it exceeds θ.

[0331] Calculation example

[0332] Continuing with the results of step 2, assume that:

[0333] R = [0.05787 0.100 0.150]

[0334] θ=0.1

[0335] Calculation process:

[0336] For device 1:

[0337] F 1 =sgn(0.05787-0.1)·log(1+0.05787)

[0338] F 1 = -1·log(1.05787)

[0339] F 1 =-0.056

[0340] For device 2:

[0341] F 2 =sgn(0.100-0.1)·log(1+0.100)

[0342] F 2 =0·log(1.1)

[0343] F 2 =0

[0344] For device 3:

[0345] F 3 =sgn(0.150-0.1)·log(1+0.150)

[0346] F 3 =1·log(1.150)

[0347] F 3 =0.13976

[0348] explain:

[0349] F 1 The value is -0.056, which means that the risk level of device 1 does not exceed the threshold, indicating that its risk is low.

[0350] F 2 The value of is 0, because the risk level of device 2 is just above the threshold and is not considered a high-risk node.

[0351] F 3 The value is 0.13976, indicating that the risk level of device 3 exceeds the threshold and is a high-risk node.

[0352] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. An intelligent gas generating agent production management system, characterized in that: The system comprises: The raw material monitoring module collects operating temperature, monitors pressure and output rate, records equipment response time, measures equipment operation frequency, analyzes equipment operation characteristics, and screens key performance indicators from chemical reactors and mixers; The device interaction graph module uses the key performance indicators to construct a dynamic interaction graph between devices, takes the devices as nodes of the graph, defines the dependency relationships as edges, performs graph structure optimization processing, and obtains an optimized dynamic graph; The dependency analysis module imports the optimized dynamic graph, learns the interaction between nodes through the graph convolution network, identifies the dependency relationship between nodes, obtains the device dependency matrix, predicts the fault diffusion path according to the device dependency matrix, and generates an abnormal prediction path; The abnormality identification module uses the abnormal prediction path to monitor the device behavior, identify the operation mode that deviates from the normal, locate the starting point of the fault, analyze the risk nodes associated with the fault, and generate the risk node analysis results; The steps for obtaining the abnormal prediction path are specifically as follows: Starting from the device dependency matrix, use the formula: Calculate the initial influence of each device and generate an initial influence vector; Among them, I i represents the initial influence of node i, α, β, and γ are the linear adjustment coefficient of distance, the base of exponential decay, and the rate of exponential decay, respectively. ij is the element in the dependency matrix, which is the dependency relationship between each pair of devices i and j, d ij is the distance between node i and node j; The initial influence vector is used to simulate the propagation process of the fault between devices. The dynamic system model is used for simulation, and the formula is adopted: F t =M×F t-1 +δ·(1-F t-1 ) Track the fault propagation path between differentiated devices and generate fault propagation vectors; Among them, F t 、F t-1 They represent the device response states at consecutive time points, δ is the system recovery rate, and M is the device dependency matrix; The fault propagation vector is analyzed over a long period of time to identify the key fault propagation paths and key impact nodes, using the formula: Evaluate the maximum impact of each device during the entire simulation period and generate abnormal prediction paths; Among them, P represents the maximum impact of multiple devices during the simulation period; The steps for obtaining the risk node analysis results are specifically as follows: The abnormal prediction path is used to monitor the device behavior, using the formula: Calculate the deviation index D of device i i , generating a deviation index for equipment that deviates from normal operation; Among them, D i represents the deviation index of device i, μ j represents the average operating level of the equipment, α adjusts the sensitivity of the deviation index, P ij is the dependence of device i on device j, X j is the current operation data of device j; Using the deviation index, the fault starting point and the associated risk node are determined using the formula: Generate a risk level for each device; Among them, R i is the risk level of device i, and λ controls the sensitivity of risk rating; Analyze the risk levels and identify high-risk nodes using the formula: F i =sgn(R i -θ)·log(1+R i ) When the risk level exceeds θ, device i is classified as a high-risk node and the risk node analysis results are generated; Among them, F i is the risk node, θ is the risk threshold, and sgn is the sign function.

2. The intelligent gas generating agent production management system according to claim 1 is characterized in that: The steps for obtaining the key performance indicators are specifically as follows: The operating temperature T and monitoring pressure P are collected every 10 minutes from the chemical reactor and the mixer, using the formula: Calculate the average operating temperature and average pressure per hour to obtain the average operating data D per hour avg ; Among them, T i , P i Represent the temperature and pressure measured every 10 minutes, n is the number of measurements per hour; record the equipment output rate R and equipment response time τ per hour, using the formula: Get the average output rate and average response time per hour, and generate the equipment response data R per hour avg and τ avg ; Among them, R i , τ i Represent the output rate and response time per hour respectively, and n is the number of measurements; Analyzing the operating temperature, monitoring pressure, production rate and equipment response time, the formula is used: Calculate and generate key performance indicators (KPIs); Among them, T avg , P avg , R avg , τ avg They are the average operating temperature, average pressure, average output rate and average response time respectively.

3. The intelligent gas generating agent production management system according to claim 2 is characterized in that: The steps for obtaining the optimized dynamic graph are specifically as follows: Based on the key performance indicators (KPIs) of each device i , assigning a unique node identifier ID to each device i And calculate the initial weight of the node using the formula: Generate the normalized weight of the device node; Where KPI represents the key performance indicator of the i-th device, n is the total number of devices, α is the normalization parameter, and W i is the normalized weight of the device node; Using the normalized weight W i , define the dependency between devices. If device A affects the performance of device B, then an edge E is established between A and B. AB , using the formula: Determine the edge weights and generate a dependency graph between devices; Among them, W AB is the weight of the edge, β adjusts the sensitivity of the dependency; According to the dependency graph between the devices, the graph structure is optimized using the formula: Adjust the graph structure to generate an optimized dynamic interaction graph; Among them, W total is the overall weight of the graph, E is the set of edges in the graph, and |W in the formula i -W j | represents the difference in node weights.

4. The intelligent gas generating agent production management system according to claim 3 is characterized in that: The steps for obtaining the device dependency matrix are specifically as follows: Import the optimized dynamic graph, process it through the graph convolutional network, identify the node relationship between devices, and use the formula: Calculate the local structural sensitivity of each node i and generate the structural sensitivity result; Among them, S i represents the structural sensitivity of node i, N(i) is the set of nodes connected to node i, and d ij is the distance between node i and node j, W ij For edge E ij The weight of , λ is the attenuation parameter; Based on the structural sensitivity results, the device dependency matrix M is constructed using the formula: Calculate the dependency relationship for each pair of devices i and j and generate a device dependency matrix; Among them, M ij is the element in the dependency matrix, ∑ k∈N(i) S k It is the sum of the structural sensitivities of the nodes directly connected to point i.

Images