Onion powder drying process multi-source data fusion intelligent regulation and control method based on Internet of Things
By using an IoT-based multi-source data fusion intelligent control method, the problem of insufficient data fusion during the onion powder drying process was solved, achieving precise control and improved energy efficiency in the onion powder drying process.
Patent Information
- Application Number
- CN202610056306.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-01-16
- Publication Date
- 2026-02-17
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
Existing IoT-based onion powder drying control methods lack a standardized fusion mechanism for multi-source heterogeneous sensor data, making it difficult to adapt to the differences in sulfide volatility characteristics among different onion varieties. They also ignore the nonlinear interaction between slice thickness and hot air velocity, leading to increased energy consumption, decreased product rehydration, and localized over-drying or clumping. Furthermore, the control system cannot respond in real time to changes in the moisture content of the raw materials.
By using an IoT-based multi-source data fusion intelligent control method, multi-dimensional parameters are acquired using heterogeneous sensors, a decision tree model is established, the interaction influence coefficients of the process flow are analyzed, and a dynamic optimal parameter decision model is generated to achieve real-time control.
It enables precise control of the drying process, improves product quality, suppresses environmental disturbances, enhances energy efficiency, and reduces production risks.
Smart Images
Figure CN121541479A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of intelligent control technology, specifically to an intelligent control method for the multi-source data fusion of the onion powder drying process based on the Internet of Things. Background Technology
[0002] Existing IoT-based onion powder drying control methods lack a standardized fusion mechanism for multi-source heterogeneous sensor data (such as temperature, humidity, near-infrared spectroscopy, and weight changes), leading to lag and error accumulation in raw material condition assessment. Process parameter decisions often rely on static thresholds or single machine learning models (such as BP neural networks), making it difficult to adapt to the differences in sulfide volatility characteristics between different onion varieties (such as purple and yellow). The coupled effects of various drying stages (pretreatment-drying-pulverization) are only analyzed through linear regression, ignoring the nonlinear interaction between slice thickness and hot air velocity, resulting in increased energy consumption and decreased product rehydration properties. Control systems often use fixed timing commands, failing to respond in real-time to dynamic changes in raw material moisture content, and are prone to localized over-drying or clumping when environmental temperature and humidity fluctuate. This makes it difficult to balance drying efficiency and product quality, hindering the industrial production of high-value-added onion powder. Summary of the Invention
[0003] To address the aforementioned technical issues, this paper presents an intelligent control method for the onion powder drying process based on Internet of Things (IoT) multi-source data fusion. This technical solution resolves the problems mentioned above.
[0004] To achieve the above objectives, the technical solution adopted by the present invention is as follows: An IoT-based intelligent control method for the onion powder drying process, incorporating multi-source data fusion, includes: S1. Based on heterogeneous sensors, acquire multi-dimensional parameters of historical target onion raw material types, and divide the multi-dimensional parameters of real-time target onion raw material types according to the process flow preference for making onion powder products, to obtain the real-time onion raw material state vector of the process flow of the onion powder product type to be made. S2. Obtain standardized process flow data for the onion powder product type to be produced, establish decision trees for each onion powder product type, and use the real-time onion raw material state vector of the process flow of the onion powder product type to be produced as input to generate the ideal parameter range of the process flow of the onion powder product type to be produced. S3. Based on the process flow of the onion powder product type to be produced, each process flow is taken as a node, and the ideal parameter range of the process flow of the onion powder product type to be produced is taken as the corresponding process flow node attribute. The interaction influence coefficient between the process flows of the onion powder product type to be produced is analyzed. S4. Based on the interaction coefficient between the process flow of the onion powder product type to be produced and the ideal parameter range of the process flow of the onion powder product type to be produced, establish the optimal parameter range decision model for the onion powder product type and generate a real-time process flow control instruction sequence for the onion powder product type to be produced.
[0005] Preferably, step S1 specifically includes: Based on multi-source heterogeneous sensors, multi-dimensional parameters of historical target onion raw material types are used for data alignment and outlier processing. Obtain multi-dimensional parameters of onion raw materials for several historical batches of onion powder products, and establish a multi-dimensional parameter matrix of onion raw materials for several historical batches of onion powder products. Using linear algebra, a linear mapping is performed on the multi-dimensional parameter matrix of onion raw materials for several historical batches of onion powder products to obtain the multi-dimensional parameter vector matrix of onion raw materials for several historical batches of onion powder products. Based on the multi-dimensional parameter vector matrix of onion raw materials used in several historical batches of onion powder products, the distribution of multi-dimensional parameter vector values of onion raw materials observed under the quality of several historical batches of onion powder products is verified, and the quality prior of onion powder products under different multi-dimensional parameter vectors of onion raw materials is determined. Based on the quality prior of different onion raw material multidimensional parameter vectors under various historical batches of onion powder production, the frequency of occurrence of each onion raw material multidimensional parameter vector value is counted, and the contribution value of each onion raw material multidimensional parameter vector value to the quality prior is calculated. Based on the probability density function, the posterior probability distribution of the quality of the onion powder product type produced under different onion raw material multidimensional parameter vector values is calculated by using the contribution value of each onion raw material multidimensional parameter vector value to the quality prior. Based on the posterior probability distribution of the quality of the onion powder product type produced under different onion raw material multi-dimensional parameter vector values, the optimal onion raw material multi-dimensional parameter vector value for each onion powder product type is selected and denoted as the process flow preference onion raw material parameter range vector for producing onion powder product type.
[0006] Preferably, step S1 further includes: Acquire multi-dimensional parameters of the target onion raw material type in real time for preprocessing; Using linear algebra, a linear mapping is performed on the multi-dimensional parameter matrix of onion raw materials for real-time onion powder product type to obtain the multi-dimensional parameter vector matrix of onion raw materials for real-time onion powder product type. Based on the process flow preference onion raw material parameter range vector for producing onion powder products, an SVM support vector machine is trained. The process flow preference onion raw material parameter range vector is used as the feature input, and the quality index of producing onion powder products is used as the output to obtain the optimal process flow preference onion raw material parameter range hyperplane boundary for each type of onion powder product. Substitute the multi-dimensional parameter vector matrix of onion raw materials for the real-time onion powder product type into the hyperplane boundary of the optimal process flow preference onion raw material parameter range for the corresponding real-time onion powder product type to generate the real-time onion raw material state vector of the process flow for the onion powder product type to be produced.
[0007] Preferably, step S2 specifically includes: Obtain standardized process parameters for the type of onion powder product to be produced; Based on DT regression decision tree, the real-time onion raw material state vector of the process flow of the onion powder product type to be produced is used as the root node, the standardized parameters of the process flow of the onion powder product type to be produced are used as the branch nodes, and the process flow quality indicators of each onion powder product type to be produced are used as the leaf nodes. The decision tree for each onion powder product type to be produced is established according to the maximum information gain of the onion raw material state vector to the standardized parameters of the process flow of the branch nodes. Based on the decision tree of each type of onion powder product to be produced, the real-time onion raw material state vector of the process flow of each type of onion powder product to be produced is used as input, and the executable parameter range of the real-time onion raw material state vector of the process flow of each type of onion powder product to be produced is used as output.
[0008] Preferably, step S2 further includes: Based on the contribution value of each onion raw material multi-dimensional parameter vector value to the quality prior, the executable parameter range of the real-time onion raw material state vector of the process flow for each type of onion powder product to be produced is marked, and the executable parameter array of the optimal quality index of the real-time onion raw material state vector of the process flow for each type of onion powder product to be produced is established. Using the non-dominated sorting of NSGA-II, and in accordance with maximizing the process flow quality margin, minimizing the total process flow energy consumption, and minimizing the total process flow time for each type of onion powder product to be produced, the process flow quality indicators corresponding to the executable parameter array of the real-time onion raw material state vector of the process flow for each type of onion powder product to be produced are recursively verified and screened, and the ideal parameter range of the real-time onion raw material state vector of the process flow for each type of onion powder product to be produced is determined.
[0009] Preferably, step S3 specifically includes: Obtain the process flow execution task sequence of the onion powder product type to be produced. Use the process flow execution task process as the node, the ideal parameter range of the real-time onion raw material state vector of the process flow of the onion powder product type to be produced as the node attribute, and the direction of the process flow execution task process as the edge to construct the directed graph network of the process flow of the onion powder product type to be produced. Based on the directed graph network of the process flow of the onion powder product type to be produced, each adjacent node is selected according to the edge pointing relationship of the directed graph network of the process flow, and the process flow adjacency matrix of the onion powder product type to be produced is established. Based on the adjacency matrix of the process flow of the onion powder product type to be produced, the eigenvector centrality of each independent node under the interaction influence of adjacent nodes is calculated and normalized. The edge weights of the directed graph network of the process flow of the onion powder product type to be produced are then assigned, resulting in a directed weighted graph network of the process flow of the onion powder product type to be produced.
[0010] Preferably, step S3 further includes: Based on the directed weighted graph network of the process flow for the onion powder product type to be produced, each process flow task-covered node is selected as the independent variable to exert influence, and the process flow quality index of the onion powder product type to be produced is used as the dependent variable to be affected. By using multiple offline regression, the degree of influence of the independent variable to exert influence on the dependent variable is calculated, and the interaction coefficients between each node in the directed weighted graph network of the process flow for the onion powder product type to be produced are determined.
[0011] Preferably, step S4 specifically includes: Obtain the parameters of the process flow for the onion powder product type to be produced in real time; Based on the ideal parameter range of the process flow for the type of onion powder product to be produced, determine the upper and lower limits of the adjustable parameters of the process flow for the type of onion powder product to be produced. Based on linear programming, the objective function is to keep the parameters of the process flow of the onion powder product type to be produced within the adjustable upper and lower limits of the process flow of the onion powder product type. The interaction coefficients between the nodes in the directed weighted graph network of the process flow of the onion powder product type are used as the constraint variables of the corresponding process flow task. The optimal parameter range decision model of the onion powder product type is constructed. Using the interior point method, the objective function and constraint decision variables of the optimal parameter interval decision model for the onion powder product type are solved to generate the adjustable parameter decision variable vector of the process flow for the onion powder product type to be produced. This vector is then substituted into the linprog function to generate the real-time process flow control instruction sequence for the onion powder product type to be produced.
[0012] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention proposes an IoT-based intelligent control scheme for the onion powder drying process, incorporating multi-source data fusion. It utilizes a heterogeneous sensor network to collect multi-dimensional parameters of the onion raw material (such as moisture, temperature, and composition) in real time. This data is combined with historical data to construct a raw material state vector. A decision tree model is then used to match standardized process parameter ranges for different product types. Furthermore, the interaction coefficients are analyzed through node-based process flow analysis, ultimately establishing a dynamic optimal parameter decision model and generating a precise control command sequence. The beneficial effects include: achieving precise control of the drying process driven by data, improving product quality; and the IoT real-time feedback mechanism effectively suppresses environmental disturbances, improving overall energy efficiency compared to traditional methods. Attached Figure Description
[0013] Figure 1 This is a flowchart of an intelligent control method for multi-source data fusion in the onion powder drying process based on the Internet of Things. Detailed Implementation
[0014] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.
[0015] Reference Figure 1 As shown, the intelligent control method for multi-source data fusion in the onion powder drying process based on the Internet of Things includes: S1. Based on heterogeneous sensors, acquire multi-dimensional parameters of historical target onion raw material types, and divide the multi-dimensional parameters of real-time target onion raw material types according to the process flow preference for making onion powder products, to obtain the real-time onion raw material state vector of the process flow of the onion powder product type to be made. Step S1 specifically includes: Based on multi-source heterogeneous sensors, multi-dimensional parameters of historical target onion raw material types are used for data alignment and outlier processing. Obtain multi-dimensional parameters of onion raw materials for several historical batches of onion powder products, and establish a multi-dimensional parameter matrix of onion raw materials for several historical batches of onion powder products. Using linear algebra, a linear mapping is performed on the multi-dimensional parameter matrix of onion raw materials for several historical batches of onion powder products to obtain the multi-dimensional parameter vector matrix of onion raw materials for several historical batches of onion powder products. Based on the multi-dimensional parameter vector matrix of onion raw materials used in several historical batches of onion powder products, the distribution of multi-dimensional parameter vector values of onion raw materials observed under the quality of several historical batches of onion powder products is verified, and the quality prior of onion powder products under different multi-dimensional parameter vectors of onion raw materials is determined. Based on the quality prior of different onion raw material multidimensional parameter vectors under various historical batches of onion powder production, the frequency of occurrence of each onion raw material multidimensional parameter vector value is counted, and the contribution value of each onion raw material multidimensional parameter vector value to the quality prior is calculated. Based on the probability density function, the posterior probability distribution of the quality of the onion powder product type produced under different onion raw material multidimensional parameter vector values is calculated by using the contribution value of each onion raw material multidimensional parameter vector value to the quality prior. Based on the posterior probability distribution of the quality of the onion powder product type produced under different onion raw material multi-dimensional parameter vector values, the optimal onion raw material multi-dimensional parameter vector value for each type of onion powder product is selected and denoted as the process flow preference onion raw material parameter range vector for the type of onion powder product. Step S1 also includes: Acquire multi-dimensional parameters of the target onion raw material type in real time for preprocessing; Using linear algebra, a linear mapping is performed on the multi-dimensional parameter matrix of onion raw materials for real-time onion powder product type to obtain the multi-dimensional parameter vector matrix of onion raw materials for real-time onion powder product type. Based on the process flow preference onion raw material parameter range vector for producing onion powder products, an SVM support vector machine is trained. The process flow preference onion raw material parameter range vector is used as the feature input, and the quality index of producing onion powder products is used as the output to obtain the optimal process flow preference onion raw material parameter range hyperplane boundary for each type of onion powder product. Substitute the multi-dimensional parameter vector matrix of onion raw materials for the real-time onion powder product type into the hyperplane boundary of the optimal process flow preference onion raw material parameter range for the corresponding real-time onion powder product type to generate the real-time onion raw material state vector of the process flow for the onion powder product type to be produced. When using it, please refer to the steps outlined above: As a further development, heterogeneous sensor information is fused and standardized through multi-source data fusion. Key parameters are extracted by using linear algebraic transformation for feature dimensionality reduction. Based on Bayesian probabilistic inference, the optimal parameter range for raw materials aimed at achieving high-quality results is mined from historical data. Finally, a robust classification boundary is constructed using SVM (Support Vector Machine). Real-time raw material data is mapped into quantifiable, standardized state vectors that can directly guide subsequent process decisions. This realizes the transformation of process knowledge from experience-dependent to data-driven automated mining and provides a forward-looking decision-making basis. It can accurately predict the state of raw materials before production, significantly improving production stability.
[0016] Furthermore, an implementation example of step S1 is provided: Scenario: An onion powder factory produces premium cold-dried fine powder. ) and ordinary hot air powder ( Two products.
[0017] Step S1 implementation process: 1. Data preparation: Collect production data from the past year Data from 1,000 batches of high-quality raw materials, including: moisture content (M), sugar content (S), and color saturation (C).
[0018] After preprocessing, the three-dimensional features [M, S, C] are mapped to two principal components, forming a two-dimensional vector matrix [PC1, PC2].
[0019] 2. Excavation Optimal range: Calculate the posterior probability P([PC1, PC2] | Q= high quality).
[0020] By plotting the probability density diagram, it was found that the probability is highest when PC1 is in the interval [0.5, 1.2] and PC2 is in the interval [-0.3, 0.4].
[0021] These two-dimensional vector points constitute The process flow prefers a range of onion raw material parameters.
[0022] 3. Train the SVM classifier: Mark the points within the above range as high quality. Raw materials (label 1).
[0023] Other production However, the quality is average, and the production... The raw material data is marked as Other (label 0).
[0024] Using these two sets of data, an SVM model (using the RBF kernel) was trained. This model learned to draw a complex closed curve (decision boundary) in the two-dimensional space [PC1, PC2], and then... Circle out the raw materials.
[0025] 4. Real-time raw material status determination: A new batch of raw materials arrived, and the sensor measured the following data: M=66%, S=21Brix, C=0.8.
[0026] Preprocessing and mapping: The data is transformed using linear algebra to obtain the real-time vector [PC1_real=0.8, PC2_real=0.1].
[0027] SVM classification: Substitute this point into the trained... The judgment is made in the SVM model.
[0028] Generate state vector: The model determines that this point is located inside the decision boundary and belongs to the high-quality Type_A raw material.
[0029] The system generates real-time state vectors, for example: [Product Type: Status code: 1, Confidence level: 0.95, Distance from center: 0.05.
[0030] This vector means that the raw material is very suitable for production. The product has a 95% probability of being a high-quality product.
[0031] S2. Obtain standardized process flow data for the onion powder product type to be produced, establish decision trees for each onion powder product type, and use the real-time onion raw material state vector of the process flow of the onion powder product type to be produced as input to generate the ideal parameter range of the process flow of the onion powder product type to be produced. Step S2 specifically includes: Obtain standardized process parameters for the type of onion powder product to be produced; Based on DT regression decision tree, the real-time onion raw material state vector of the process flow of the onion powder product type to be produced is used as the root node, the standardized parameters of the process flow of the onion powder product type to be produced are used as the branch nodes, and the process flow quality indicators of each onion powder product type to be produced are used as the leaf nodes. The decision tree for each onion powder product type to be produced is established according to the maximum information gain of the onion raw material state vector to the standardized parameters of the process flow of the branch nodes. Based on the decision tree of each type of onion powder product to be produced, the real-time onion raw material state vector of the process flow of each type of onion powder product to be produced is used as the input, and the executable parameter range of the real-time onion raw material state vector of the process flow of each type of onion powder product to be produced is used as the output. Step S2 also includes: Based on the contribution value of each onion raw material multi-dimensional parameter vector value to the quality prior, the executable parameter range of the real-time onion raw material state vector of the process flow for each type of onion powder product to be produced is marked, and the executable parameter array of the optimal quality index of the real-time onion raw material state vector of the process flow for each type of onion powder product to be produced is established. Using the non-dominated sorting of NSGA-II, and in accordance with maximizing the process flow quality margin, minimizing the total process flow energy consumption, and minimizing the total process flow time for each type of onion powder product to be produced, the process flow quality indicators corresponding to the executable parameter array of the real-time onion raw material state vector of the process flow for each type of onion powder product to be produced are recursively verified and screened, and the ideal parameter range of the real-time onion raw material state vector of the process flow for each type of onion powder product to be produced is determined.
[0032] When using it, please refer to the steps outlined above: As a further development, a decision tree is established for each type of onion powder product to be produced, based on regression decision trees. Real-time raw material status is used as input, and the optimal process parameter range for multiple objectives is used as output. Based on the principle of maximizing information gain, a mapping relationship is established from raw material characteristics to historically successful process parameters and quality indicators, outputting an initial parameter range based on statistics. Then, a multi-objective optimization algorithm (NSGA-II) is introduced to recursively filter parameters within the range through non-dominated sorting, balancing multiple competing objectives such as maximizing quality margin and minimizing energy consumption and time. Finally, an ideal parameter range that achieves Pareto optimality across multiple key indicators is found. This achieves a leap from simply ensuring product quality to comprehensively optimizing production efficiency. It can automatically generate an optimal process parameter range that combines high quality, low energy consumption, and short cycle time, not only improving the accuracy of decision-making but also allowing for flexible selection of the best strategy according to actual production needs. Furthermore, recursive verification ensures the robustness of the output parameters, significantly reducing production risks and improving overall economic benefits.
[0033] S3. Based on the process flow of the onion powder product type to be produced, each process flow is taken as a node, and the ideal parameter range of the process flow of the onion powder product type to be produced is taken as the corresponding process flow node attribute. The interaction influence coefficient between the process flows of the onion powder product type to be produced is analyzed. Step S3 specifically includes: Obtain the process flow execution task sequence of the onion powder product type to be produced. Use the process flow execution task process as the node, the ideal parameter range of the real-time onion raw material state vector of the process flow of the onion powder product type to be produced as the node attribute, and the direction of the process flow execution task process as the edge to construct the directed graph network of the process flow of the onion powder product type to be produced. Based on the directed graph network of the process flow of the onion powder product type to be produced, each adjacent node is selected according to the edge pointing relationship of the directed graph network of the process flow, and the process flow adjacency matrix of the onion powder product type to be produced is established. Based on the adjacency matrix of the process flow of the onion powder product type to be produced, the eigenvector centrality of each independent node under the interaction influence of the adjacent nodes is calculated and normalized. The edge weights of the directed graph network of the process flow of the onion powder product type to be produced are assigned, and the directed weighted graph network of the process flow of the onion powder product type to be produced is obtained. Step S3 also includes: Based on the directed weighted graph network of the process flow for the onion powder product type to be produced, each process flow task-covered node is selected as the independent variable to exert influence, and the process flow quality index of the onion powder product type to be produced is used as the dependent variable to be affected. By using multiple offline regression, the degree of influence of the independent variable to exert influence on the dependent variable is calculated, and the interaction coefficients between each node in the directed weighted graph network of the process flow for the onion powder product type to be produced are determined.
[0034] When using it, please refer to the steps outlined above: As a further development, the process flow is abstracted into a weighted directed network. The network topology is analyzed using the eigenvector centrality algorithm in the graph, key process nodes are intelligently identified, and their initial influence weights are quantified. Then, using a multiple linear regression model driven by historical data, the quantitative influence coefficients of key node process parameters on the final quality indicators are accurately calculated. By combining directed graph networks with multiple regression, the indirect and cross-influence coupling relationships between process links are systematically revealed, achieving accurate quantification of the impact effect. At the same time, key bottleneck links are automatically identified based on eigenvector centrality, and with the model's high interpretability and predictive ability, the quality control level and decision-making efficiency are significantly improved.
[0035] S4. Based on the interaction coefficient between the process flow of the onion powder product type to be produced and the ideal parameter range of the process flow of the onion powder product type to be produced, establish the optimal parameter range decision model of the onion powder product type and generate the real-time process flow control instruction sequence of the onion powder product type to be produced. Step S4 specifically includes: Obtain the parameters of the process flow for the onion powder product type to be produced in real time; Based on the ideal parameter range of the process flow for the type of onion powder product to be produced, determine the upper and lower limits of the adjustable parameters of the process flow for the type of onion powder product to be produced. Based on linear programming, the objective function is to keep the parameters of the process flow of the onion powder product type to be produced within the adjustable upper and lower limits of the process flow of the onion powder product type. The interaction coefficients between the nodes in the directed weighted graph network of the process flow of the onion powder product type are used as the constraint variables of the corresponding process flow task. The optimal parameter range decision model of the onion powder product type is constructed. Using the interior point method, the objective function and constraint decision variables of the optimal parameter interval decision model for the onion powder product type are solved to generate the adjustable parameter decision variable vector of the process flow for the onion powder product type to be produced. This vector is then substituted into the linprog function to generate the real-time process flow control instruction sequence for the onion powder product type to be produced.
[0036] When using it, please refer to the steps outlined above: As a further development, compared to traditional PID controllers that isolate and control a single parameter, this solution constructs an optimal parameter range decision model for the onion powder product type. The resulting control command is the optimal solution through coordinated adjustment of all parameters, avoiding data oscillation problems. Furthermore, the model uses real-time data as input, and each solution is a re-optimization of the current operating conditions. This enables the system to adaptively cope with external disturbances such as fluctuations in incoming materials and changes in equipment status, demonstrating strong robustness. Secondly, by anticipating the interactions between parameters through constraints, the adjustment decisions are smooth and gradual (moving the range within the constraints rather than drastically changing the setpoint), reducing fluctuations in the production process and achieving adaptive decision-making from monitoring-alarm-manual intervention to analysis-decision-automatic execution.
[0037] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention. The scope of protection claimed by the appended claims and their equivalents is defined.
Claims
1. An IoT-based multi-source data fusion intelligent regulation method for onion powder drying process, characterized in that, The method comprises the following steps: S1, based on heterogeneous sensors, obtaining the multidimensional parameters of the historical target onion raw material type, and dividing the multidimensional parameters of the real-time target onion raw material type according to the process flow preference of the onion powder product type to obtain the process flow real-time onion raw material state vector of the onion powder product type to be produced; S2, obtaining the process flow standardized data of the onion powder product type to be produced, establishing the decision tree of each onion powder product type, taking the process flow real-time onion raw material state vector of the onion powder product type to be produced as the input, and generating the ideal parameter interval of the process flow of the onion powder product type to be produced; S3, based on the process flow of the onion powder product type to be produced, taking each process flow as a node, taking the ideal parameter interval of the process flow of the onion powder product type to be produced as the corresponding process flow node attribute, and analyzing the interaction influence coefficient between the process flows of the onion powder product type to be produced; S4, based on the interaction influence coefficient between the process flows of the onion powder product type to be produced and the ideal parameter interval of the process flow of the onion powder product type to be produced, establishing an optimal parameter interval decision model of the onion powder product type, and generating a real-time process flow control instruction sequence of the onion powder product type to be produced. 2.The IoT-based intelligent regulation method for multi-source data fusion in an onion powder drying process according to claim 1, characterized in that, Step S1 specifically comprises: Based on multiple source heterogeneous sensors, the multidimensional parameters of the historical target onion raw material type are taken for data alignment and abnormal value processing; Obtaining the multidimensional parameters of the onion raw material of the historical batches of the onion powder product type, and establishing the multidimensional parameter matrix of the onion raw material of the historical batches of the onion powder product type; Using linear algebra, linear mapping is performed on the multidimensional parameter matrix of the onion raw material of the historical batches of the onion powder product type to obtain the multidimensional parameter vector matrix of the onion raw material of the historical batches of the onion powder product type; According to the multidimensional parameter vector matrix of the onion raw material of the historical batches of the onion powder product type, the onion raw material multidimensional parameter vector value distribution observed under the quality of the historical batches of the onion powder product type is verified, and the quality prior of the historical batches of the onion powder product type under the given different onion raw material multidimensional parameter vectors is determined; Based on the quality prior of the historical batches of the onion powder product type under the given different onion raw material multidimensional parameter vectors as a condition, the frequency of occurrence of each onion raw material multidimensional parameter vector value is counted, and the contribution value of each onion raw material multidimensional parameter vector value to the quality prior is calculated; Based on the probability density function, the contribution value of each onion raw material multidimensional parameter vector value to the quality prior is calculated, and the posterior probability distribution of the quality of the onion powder product type produced under different onion raw material multidimensional parameter vectors is calculated; According to the posterior probability distribution of the quality of the onion powder product type produced under different onion raw material multidimensional parameter vectors, the optimal onion raw material multidimensional parameter vector value of each onion powder product type is screened out, which is recorded as the process flow preferred onion raw material parameter range vector of the onion powder product type. 3.The IoT-based intelligent regulation and control method for multi-source data fusion of onion powder drying process according to claim 2, characterized in that, Step S1 further comprises: Obtaining the multidimensional parameters of the real-time target onion raw material type for preprocessing; Utilize linear algebra, and linear mapping is performed on the multi-dimensional parameter matrix of the onion raw material of the real-time onion powder product type, to obtain a multi-dimensional parameter vector matrix of the onion raw material of the real-time onion powder product type; Based on the process-preferred onion raw material parameter range vector of the onion powder product type, a SVM support vector machine is trained, the process-preferred onion raw material parameter range vector is taken as a characteristic input, and a quality index of the onion powder product type is taken as an output, to obtain an optimal process-preferred onion raw material parameter range hyperplane boundary of each onion powder product type; The multi-dimensional parameter vector matrix of the real-time onion raw material of the onion powder product type is substituted into the optimal process-preferred onion raw material parameter range hyperplane boundary of the corresponding real-time onion powder product type, to generate a process real-time onion raw material state vector of the to-be-made onion powder product type. 4.The IoT-based intelligent regulation and control method for multi-source data fusion of an onion powder drying process according to claim 3, characterized in that, Step S2 specifically comprises: Obtaining a process standardization parameter of the to-be-made onion powder product type; Based on a DT regression decision tree, the process real-time onion raw material state vector of the to-be-made onion powder product type is taken as a root node, the process standardization parameter of the to-be-made onion powder product type is taken as a branch node, and a process quality index of each to-be-made onion powder product type is taken as a leaf node, a decision tree of each to-be-made onion powder product type is established according to a maximum information gain of the onion raw material state vector to the branch node of the process standardization parameter as a division decision; Based on the decision tree of each to-be-made onion powder product type, the process real-time onion raw material state vector of each to-be-made onion powder product type is taken as an input, and an executable parameter interval of the process real-time onion raw material state vector of each to-be-made onion powder product type is taken as an output. 5.The IoT-based intelligent regulation and control method for multi-source data fusion of an onion powder drying process according to claim 4, characterized in that, Step S2 further comprises: According to a posteriori contribution value of each onion raw material multi-dimensional parameter vector value to a quality priori, the executable parameter interval of the process real-time onion raw material state vector of each to-be-made onion powder product type is marked, and an executable parameter array of an optimal quality index of the process real-time onion raw material state vector of each to-be-made onion powder product type is established; According to the maximum process quality margin, the minimum process total energy consumption and the minimum process total time length of each to-be-made onion powder product type, the corresponding process quality index of the process real-time onion raw material state vector executable parameter array of each to-be-made onion powder product type is recursively verified and screened by using the non-dominated sorting of the NSGA-II, to determine an ideal parameter interval of the process real-time onion raw material state vector of the to-be-made onion powder product type. 6.The IoT-based multi-source data fusion intelligent regulation and control method for onion powder drying process according to claim 5, characterized in that, Step S3 specifically comprises: Obtaining a process execution task progress sequence of the to-be-made onion powder product type, taking a process execution task progress as a node, taking the ideal parameter interval of the process real-time onion raw material state vector of the to-be-made onion powder product type as a node attribute, and taking a direction of the process execution task progress as an edge, to construct a process directed graph network of the to-be-made onion powder product type. Based on the process flow directed graph network of the to-be-made onion powder product type, according to the edge direction relationship of the process flow directed graph network, adjacent nodes are screened out, and a process flow adjacency matrix of the to-be-made onion powder product type is established; According to the process flow adjacency matrix of the to-be-made onion powder product type, the feature vector centrality of each independent node under the interaction of adjacent nodes is normalized, the edge weight of the process flow directed graph network of the to-be-made onion powder product type is given, and a process flow directed weighted graph network of the to-be-made onion powder product type is obtained. 7.The IoT-based intelligent regulation and control method for multi-source data fusion of an onion powder drying process according to claim 6, characterized in that, Step S3 further includes: Based on the process flow directed weighted graph network of the to-be-made onion powder product type, the process task covered nodes of each process flow process are screened out as the influencing independent variables, the process quality indicators of the to-be-made onion powder product type are taken as the influenced dependent variables, the influence degree of the influencing independent variables on the influenced dependent variables is calculated according to the multiple offline regression, and the interaction coefficients between the nodes in the process flow directed weighted graph network of the to-be-made onion powder product type are determined. 8.The IoT-based multi-source data fusion intelligent regulation and control method for an onion powder drying process according to claim 7, characterized in that, Step S4 specifically includes: Obtaining the parameters of the real-time process flow of the to-be-made onion powder product type; Based on the ideal parameter interval of the process flow of the to-be-made onion powder product type, the upper and lower limit values of the adjustable parameters of the process flow of the to-be-made onion powder product type are determined; Based on linear programming, taking the process flow of the to-be-made onion powder product type as the target function to keep the parameters of the real-time process flow of the to-be-made onion powder product type in the adjustable parameter upper and lower limit value interval, and taking the interaction coefficients between the nodes in the process flow directed weighted graph network of the to-be-made onion powder product type as the constraint variables of the corresponding process flow process, an optimal parameter interval decision model of the onion powder product type is constructed; According to the interior point method, the linear relationship between the objective function and the constraint condition decision variable of the optimal parameter interval decision model of the onion powder product type is solved to generate the adjustable parameter decision variable vector of the process flow of the to-be-made onion powder product type, which is substituted into the linprog function to generate a real-time process flow control instruction sequence of the to-be-made onion powder product type.