Real-time fruit juice production line quality monitoring system and method based on Internet of Things

Abstract

The invention relates to the technical field of flexible production, in particular to a juice production line quality real-time monitoring system and method based on the Internet of Things, and the system comprises a data acquisition module, a data center control module, a multi-level control module and a production line flexible control module. The method comprises the following steps: acquiring real-time parameter data in a production process based on various parameter sensors distributed at key stations of a production line, and preprocessing the real-time parameter data through an edge computing node; analyzing the real-time parameter data to obtain abnormal data and a production parameter control strategy; analyzing the abnormal data of the fruit juice through an analytic hierarchy process control model, controlling key parameters of a production line in combination with a production parameter control strategy, and generating a global control strategy; executing automatic parameter adjustment according to the production parameter control strategy; the automatic parameter adjustment automatically controls production process parameters according to characteristics of types; and after the switching is completed, verifying the quality and carrying out parameter optimization control.

Description

Technical Field

[0001] The present invention relates to the field of flexible production technology, and in particular to a real-time monitoring system and method for the quality of a juice production line based on the Internet of Things. Background Art

[0002] Although industrial Internet of Things technology has been widely used in real-time monitoring of the manufacturing industry, in the specific field of juice production, the existing production control and flexible switching mechanisms still have many bottlenecks.

[0003] At the production control level, traditional production lines mostly use centralized control systems based on SCADA / PLC. However, existing control systems often lack a deep understanding of these material characteristics and the ability to analyze the dynamic correlations between parameters, making it difficult to achieve automated and refined parameter adjustment, resulting in large fluctuations in product quality. Secondly, the data generated by existing monitoring systems often suffer from problems such as inconsistent interface standards and data silos, making it difficult to conduct effective real-time analysis and utilization. Although some systems have introduced data collection, data analysis is mostly conducted offline or relies on manual judgment. There is a lack of online predictive models and intelligent feedback mechanisms based on machine learning, making it impossible to quickly respond to production anomalies and achieve closed-loop quality control.

[0004] In terms of flexible control of production lines, frequent switching of multiple types of juice is the norm. However, existing technologies are inefficient in dealing with such switching: First, when switching between different juices, residues in pipelines and equipment are inevitable, and they need to rely on modules such as CIP or HPP for cleaning and sterilization. This not only increases energy consumption such as water and electricity, but also significantly prolongs production preparation time and downtime, reducing equipment utilization. Second, although the flexibility of equipment combinations has been improved, the underlying control command issuance and data collection still rely on traditional SCADA / PLC systems with insufficient real-time performance and low intelligence. This makes it difficult to achieve rapid and automatic process parameter adjustment and optimization when switching between different juice recipes on the production line, and it is impossible to effectively shorten the switching cycle and ensure the stability of product quality after switching. Summary of the Invention

[0005] The present invention aims to provide a real-time quality monitoring system and method for a juice production line based on the Internet of Things, comprising a data acquisition module, a data center control module, a multi-level control module, and a production line flexible control module. Based on a variety of parameter sensors distributed at key workstations on the juice production line, real-time parameter data from the juice production process is collected and pre-processed through edge computing nodes; the real-time parameter data is analyzed to obtain abnormal juice data and a production parameter control strategy; the data center control module includes an abnormality characteristic database, a deep learning prediction unit, and a parameter optimization control unit; the abnormal juice data is analyzed using a hierarchical analysis control model, and the key parameters of the production line are controlled in combination with the production parameter control strategy to generate a global control strategy; when the juice type is switched, the parameters are automatically adjusted according to the production parameter control strategy; the automatic parameter adjustment automatically controls the production process parameters according to the characteristics of the juice type; after the switch is completed, the juice quality is verified and parameter optimization control is performed.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A real-time monitoring system for juice production line quality based on the Internet of Things, including:

[0008] The data acquisition module collects real-time parameter data during the juice production process based on multiple parameter sensors distributed at key workstations of the juice production line, and pre-processes it through edge computing nodes;

[0009] A data center control module is used to analyze the real-time parameter data to obtain juice abnormality data and production parameter control strategies; the data center control module includes an abnormality characteristic database, a deep learning prediction unit, and a parameter optimization control unit;

[0010] The multi-level control module analyzes abnormal juice data through the hierarchical analysis control model, controls key parameters of the production line in combination with the production parameter control strategy, and generates a global control strategy;

[0011] The flexible control module of the production line is used to perform automatic parameter adjustment according to the production parameter control strategy when the juice type is switched; the automatic parameter adjustment automatically controls the production process parameters according to the characteristics of the juice type; after the switch is completed, the juice quality is verified and the parameter optimization control is performed.

[0012] Preferably, the real-time parameter data includes temperature, humidity, pressure, flow rate, pH value and sugar content; the edge computing node preprocesses the collected real-time parameter data, including data cleaning, outlier removal and data standardization.

[0013] Preferably, the abnormal characteristics database stores abnormal data, abnormal types and processing methods in various historical juice production processes;

[0014] The deep learning prediction unit is used to predict abnormal juice data based on the real-time parameter data and the abnormal characteristics database;

[0015] The parameter optimization control unit is used to generate a production parameter control strategy based on the predicted juice abnormality data and the information in the abnormality characteristic database.

[0016] Preferably, the hierarchical analysis control model includes a data analysis layer, an anomaly analysis layer, a control strategy generation layer and a structure optimization layer;

[0017] The data analysis layer uses edge computing nodes to perform spatiotemporal alignment and feature fusion on real-time parameter data, mapping the physical state space of the production line in real time and constructing a multi-dimensional parameter tensor;

[0018] The anomaly analysis layer combines the juice anomaly data to perform an anomaly pattern analysis on the multi-dimensional parameter tensor, generating three types of fault sources: equipment-level anomaly, process-level anomaly, and raw material-level anomaly;

[0019] The control strategy generation layer analyzes the fault source to obtain the parameter adjustment boundary conditions, combines the production parameter control strategy to control the key parameters of the production line, and generates a global control strategy;

[0020] The structural optimization layer dynamically iterates the abnormal characteristic database and the deep learning prediction unit based on the fault source and the global control strategy.

[0021] Preferably, the specific process of performing automatic parameter adjustment according to the production parameter control strategy is:

[0022] Based on the material property database of juice types, a dynamic correlation matrix between production process parameters and juice physical and chemical indicators is established, and the optimal parameter reference range of the target juice is determined through gradient similarity matching;

[0023] A multi-objective collaborative optimization mechanism is used to construct a dynamic adjustment priority queue within the optimal parameter reference range. The sterilization temperature and pH value constitute a constraint pair, and parameter linkage adjustment is achieved through a fuzzy PID controller. The sugar content parameter is controlled by feedforward compensation based on online near-infrared spectroscopy feedback.

[0024] The real-time adjustment parameters are used to predict the adjustment quality fluctuation curve through physical-chemical coupling simulation to generate a parameter confidence evaluation matrix;

[0025] When the parameter confidence evaluation matrix exceeds the threshold, the back propagation optimization algorithm is triggered, and the parameter adjustment step size is dynamically corrected according to the second-order derivative characteristics of the quality fluctuation curve. The optimization path characteristics are encoded as knowledge graph fragments and injected into the abnormal characteristics database to form a self-evolution mechanism for parameter adjustment rules.

[0026] Preferably, after the switching is completed, the specific process of verifying the juice quality and performing parameter optimization control includes:

[0027] After the juice type switch is completed, collect the newly produced juice samples; conduct quality testing on the juice samples, including sensory evaluation, physical and chemical analysis and microbiological indicators;

[0028] Compare the test results with the standard quality requirements to determine whether the juice quality is qualified;

[0029] If the juice quality is qualified, the current production parameter settings are recorded;

[0030] If it fails to meet the standards, the production parameters will be adjusted based on the test results and the information in the abnormal characteristics database, and the quality test will be carried out again until the juice quality meets the standards.

[0031] A method for real-time monitoring of juice production line quality based on the Internet of Things, comprising:

[0032] Based on multiple parameter sensors distributed at key workstations of the juice production line, real-time parameter data of the juice production process is collected and pre-processed through edge computing nodes;

[0033] The real-time parameter data is analyzed to obtain juice abnormality data and production parameter control strategies; the data center control module includes an abnormality characteristic database, a deep learning prediction unit, and a parameter optimization control unit;

[0034] The abnormal data of juice was analyzed by using the hierarchical analysis control model, and the key parameters of the production line were controlled by combining the production parameter control strategy to generate a global control strategy;

[0035] When the juice type is switched, the parameters are automatically adjusted according to the production parameter control strategy; the parameter automatic adjustment automatically controls the production process parameters according to the characteristics of the juice type; after the switch is completed, the juice quality is verified and the parameter optimization control is performed.

[0036] Compared with the prior art, the present invention has the following beneficial effects:

[0037] 1. This invention deploys multiple parameter sensors at key workstations in the juice production line, combined with the preprocessing capabilities of edge computing nodes, to clean, denoise, and standardize real-time parameters. The data center control module integrates an abnormality characteristics database, a deep learning prediction unit, and a parameter optimization control unit. It performs spatiotemporal alignment and feature fusion of collected data within milliseconds, constructing a multidimensional parameter tensor. Using a hierarchical analysis control model, it automatically analyzes three types of fault sources: equipment-level, process-level, and raw material-level. This significantly reduces defective product rates and rework costs, improving juice quality consistency and the overall stability of the production line.

[0038] 2. This invention proposes a flexible production line control module and a multi-objective collaborative optimization mechanism. Utilizing a dynamic correlation matrix between production process parameters and physical and chemical indicators, constructed based on a material properties database, the optimal parameter reference range for the target juice is automatically determined through gradient similarity matching. During the switching process, real-time parameter adjustments are simulated using physical-chemical coupling to generate a quality fluctuation prediction curve and construct a parameter confidence assessment matrix. When the assessment result exceeds a preset threshold, a backpropagation optimization algorithm is automatically triggered to apply a second-order derivative feature correction to the adjustment step size, shortening the switching cycle and significantly improving the efficiency and quality stability of multi-variety production.

[0039] 3. The present invention combines deep learning prediction with knowledge graphs to construct a self-evolutionary closed-loop mechanism for parameter optimization control. The deep learning prediction unit can predict possible quality fluctuations in advance based on real-time parameters and a historical abnormal characteristics database; the parameter optimization control unit generates a multi-level control strategy, and encodes the optimization path features into knowledge graph fragments, which are injected into the abnormal characteristics database to achieve dynamic accumulation of process adjustment rules. As the system runs, the database and prediction model are continuously improved under the continuous iteration of the structural optimization layer, the knowledge graph becomes increasingly rich, and the identification and response to new abnormal patterns become more accurate. This self-learning mechanism greatly reduces the cost of later maintenance and model retraining, giving the system good scalability and robustness. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 This is a structural diagram of a real-time monitoring system for juice production line quality based on the Internet of Things provided by the present invention;

[0041] Figure 2 A flow chart of a method for real-time monitoring of juice production line quality based on the Internet of Things provided by the present invention;

[0042] Figure 3 A schematic diagram of the hierarchical analysis control model structure provided by an embodiment of the present invention;

[0043] Figure 4 A schematic diagram of the automatic parameter adjustment structure provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0044] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0045] The present invention provides a method for real-time monitoring of the quality of a juice production line based on the Internet of Things, which is applied to a real-time monitoring system for the quality of a juice production line based on the Internet of Things. For a specific system structure diagram and method flow chart, please refer to Figure 1 and Figure 2 ;

[0046] Example 1

[0047] As an embodiment of the present invention, refer to Figure 2 The first step is applied to the data acquisition module of a real-time quality monitoring system for a juice production line based on the Internet of Things. The data acquisition module is based on multiple parameter sensors distributed at key positions of the juice production line to collect real-time parameter data during the juice production process and pre-process it through edge computing nodes.

[0048] Furthermore, the real-time parameter data includes temperature, humidity, pressure, flow rate, pH value and sugar content; the edge computing node preprocesses the collected real-time parameter data, including data cleaning, outlier removal and data standardization.

[0049] In this example, by deploying multiple parameter sensors at key workstations in the juice production line and integrating them with the preprocessing capabilities of edge computing nodes, efficient data collection and processing of real-time parameters such as temperature, humidity, pressure, flow rate, pH, and sugar content are achieved. Edge computing nodes cleanse, remove outliers, and normalize data locally, significantly improving data quality and system response speed. Compared to the traditional method of uploading all data to the cloud for processing, edge computing processes data close to the data source, reducing latency and improving the system's real-time performance and stability.

[0050] As an embodiment of the present invention, refer to Figure 2 The second step is applied to a data center control module of a real-time monitoring system for the quality of a juice production line based on the Internet of Things. The data center control module is used to analyze the real-time parameter data to obtain juice abnormality data and production parameter control strategies; the data center control module includes an abnormal characteristic database, a deep learning prediction unit and a parameter optimization control unit.

[0051] Furthermore, the abnormal characteristics database stores abnormal data, abnormal types and treatment methods in various historical juice production processes;

[0052] The deep learning prediction unit is used to predict abnormal juice data based on the real-time parameter data and the abnormal characteristics database;

[0053] The parameter optimization control unit is used to generate a production parameter control strategy based on the predicted juice abnormality data and the information in the abnormality characteristic database.

[0054] In this embodiment, by integrating the abnormal characteristics database, the deep learning prediction unit and the parameter optimization control unit, the intelligentization of abnormality detection and parameter optimization in the juice production process is realized. The abnormal characteristics database stores abnormal data, abnormality types and processing methods in the historical production process, providing rich training data for the deep learning prediction unit. The deep learning prediction unit can predict possible abnormal situations in advance based on real-time parameter data and the abnormal characteristics database, thereby realizing real-time monitoring and early warning of the production process. The parameter optimization control unit generates corresponding production parameter control strategies based on the predicted abnormal data and the information in the database to ensure the stability of the production process and the consistency of product quality. Compared with the traditional abnormality handling method that relies on manual experience or static rules, the present invention improves the accuracy and response speed of abnormality detection through the combination of deep learning and historical data, reduces the need for human intervention, and improves production efficiency and product quality.

[0055] As an embodiment of the present invention, refer to Figure 2 The third step is applied to a multi-level control module of a real-time quality monitoring system for a juice production line based on the Internet of Things. The multi-level control module analyzes the abnormal data of the juice through the hierarchical analysis control model, controls the key parameters of the production line in combination with the production parameter control strategy, and generates a global control strategy.

[0056] Furthermore, the hierarchical analysis control model includes a data analysis layer, an abnormality analysis layer, a control strategy generation layer and a structure optimization layer. Figure 3 ;

[0057] The data analysis layer uses edge computing nodes to perform spatiotemporal alignment and feature fusion on real-time parameter data, mapping the physical state space of the production line in real time and constructing a multi-dimensional parameter tensor;

[0058] The anomaly analysis layer combines the juice anomaly data to perform an anomaly pattern analysis on the multi-dimensional parameter tensor, generating three types of fault sources: equipment-level anomaly, process-level anomaly, and raw material-level anomaly;

[0059] The control strategy generation layer analyzes the fault source to obtain parameter adjustment boundary conditions, combines the production parameter control strategy to control the key parameters of the production line, and generates a global control strategy;

[0060] The structural optimization layer dynamically iterates the abnormal characteristic database and the deep learning prediction unit based on the fault source and the global control strategy.

[0061] In this embodiment, the multi-level control module proposed by the present invention utilizes a hierarchical analytical control model to enable in-depth analysis of abnormal data and precise control of key parameters during the juice production process, thereby generating a global control strategy and improving the intelligence level of the production line and the stability of product quality. The model comprises a data analysis layer, an anomaly resolution layer, a control strategy generation layer, and a structure optimization layer. These layers work together to create a dynamic, adaptive control system. Compared to traditional static control systems, the multi-level control module of the present invention achieves real-time monitoring and optimized control of the production process through dynamic modeling and intelligent decision-making, significantly improving production efficiency and product quality.

[0062] As an embodiment of the present invention, refer to Figure 2 The fourth step is applied to a production line flexible control module of a real-time quality monitoring system for a juice production line based on the Internet of Things. The production line flexible control module is used to automatically adjust parameters according to the production parameter control strategy when the juice type is switched; the automatic parameter adjustment automatically controls the production process parameters according to the characteristics of the juice type; after the switch is completed, the juice quality is verified and the parameter optimization control is performed.

[0063] Furthermore, the specific process of performing automatic parameter adjustment according to the production parameter control strategy is as follows:

[0064] Based on the material property database of juice types, a dynamic correlation matrix between production process parameters and juice physical and chemical indicators is established, and the optimal parameter reference range of the target juice is determined through gradient similarity matching;

[0065] A multi-objective collaborative optimization mechanism is used to construct a dynamic adjustment priority queue within the optimal parameter reference range. The sterilization temperature and pH value constitute a constraint pair, and parameter linkage adjustment is achieved through a fuzzy PID controller. The sugar content parameter is controlled by feedforward compensation based on online near-infrared spectroscopy feedback.

[0066] The real-time adjustment parameters are used to predict the adjustment quality fluctuation curve through physical-chemical coupling simulation to generate a parameter confidence evaluation matrix;

[0067] When the parameter confidence evaluation matrix exceeds the threshold, the back propagation optimization algorithm is triggered, and the parameter adjustment step size is dynamically corrected according to the second-order derivative characteristics of the quality fluctuation curve. The optimization path characteristics are encoded as knowledge graph fragments and injected into the abnormal characteristics database to form a self-evolution mechanism for parameter adjustment rules.

[0068] In this embodiment, the flexible production line control module provided by the present invention enables automatic parameter adjustment based on a material properties database during juice type switching, significantly improving the production line's ability to quickly adapt to a variety of juice products and reducing quality fluctuations and downtime caused by switching. By constructing a dynamic correlation matrix between production process parameters and juice physical and chemical indicators, combined with a gradient similarity matching method, the optimal parameter baseline range for the target juice can be accurately determined, providing an efficient reference for subsequent control. The use of a multi-objective collaborative optimization mechanism and a fuzzy PID controller effectively enhances parameter coordination and quality stability during the production process. Furthermore, through physical-chemical coupled simulation prediction, the quality fluctuation curve is adjusted in real time, enabling dynamic evaluation and feedback of control effects. The optimized path is encoded as a knowledge graph fragment and injected into the abnormal properties database, forming a continuously evolving control strategy library, thereby continuously enhancing the system's intelligent control level and adaptive capabilities, comprehensively improving the flexible control performance and product quality consistency of the juice production line.

[0069] Furthermore, after the switch is completed, the specific process of verifying the juice quality and performing parameter optimization control includes:

[0070] After the juice type switch is completed, collect the newly produced juice samples; conduct quality testing on the juice samples, including sensory evaluation, physical and chemical analysis and microbiological indicators;

[0071] Compare the test results with the standard quality requirements to determine whether the juice quality is qualified;

[0072] If the juice quality is qualified, the current production parameter settings are recorded;

[0073] If it fails to meet the standards, the production parameters will be adjusted based on the test results and the information in the abnormal characteristics database, and the quality test will be carried out again until the juice quality meets the standards.

[0074] In this embodiment, by carrying out systematic quality verification and parameter optimization control after the juice type is switched, the consistency of product quality and the stability of the production process are effectively guaranteed. By comprehensively testing the taste, color, nutritional components and microbial indicators of newly produced juice samples, and comparing the test results with the preset standards, rapid judgment and accurate feedback on product quality can be achieved. If the test results meet the quality requirements, the system immediately records the current production parameter settings as a parameter benchmark for subsequent production of the same product, enhancing the traceability of parameter management and the reusability of the production process. If the quality is unqualified, the system relies on the historical abnormality information and corresponding control experience in the abnormal characteristics database to intelligently adjust the relevant process parameters and repeat the test process until the product meets the quality requirements. This closed-loop optimization mechanism effectively prevents defective products from entering the market and improves the product qualification rate and quality consistency in the juice production process.

[0075] This invention provides a real-time quality monitoring system for a juice production line based on the Internet of Things. By constructing a data acquisition module, a data center control module, a multi-level control module, and a flexible production line control module, the system achieves intelligent monitoring and efficient regulation of the entire juice production process. The system utilizes edge computing to improve data processing efficiency, integrates deep learning with an anomaly characteristics database to achieve precise anomaly prediction and parameter optimization control, and incorporates a multi-level analysis model to enhance the hierarchical analysis of fault sources and the generation of global strategies. Furthermore, a flexible control mechanism improves adaptability and quality stability when switching between multiple juice types, significantly enhancing production efficiency, product qualification rate, and the system's intelligence level.

[0076] Example 2

[0077] As an embodiment of the present invention, refer to Figure 2 The first step is applied to the data acquisition module of a real-time quality monitoring system for a juice production line based on the Internet of Things. The data acquisition module is based on multiple parameter sensors distributed at key positions of the juice production line to collect real-time parameter data during the juice production process and pre-process it through edge computing nodes.

[0078] Furthermore, the real-time parameter data includes temperature, humidity, pressure, flow rate, pH value and sugar content (unit: °Brix). These parameters are collected in real time by sensors placed at key production line locations (such as raw material mixing tanks, sterilization equipment inlets and outlets, and filling machines).

[0079] Preprocessing: The edge computing node performs the following processing on the collected data:

[0080] Data cleaning removes invalid data due to sensor failure or external interference; outlier elimination uses statistical methods to remove outliers; and data normalization normalizes parameters of different dimensions to the [0, 1] range for ease of subsequent analysis. The use of edge computing nodes significantly reduces data transmission bandwidth requirements and cloud computing pressure, enabling millisecond-level local data preprocessing and rapid response. This is crucial for critical processes such as juice sterilization, which requires instant feedback control, and addresses the high latency and poor real-time performance issues of traditional cloud-based architectures.

[0081] As an embodiment of the present invention, refer to Figure 2 The second step is applied to a data center control module of a real-time monitoring system for the quality of a juice production line based on the Internet of Things. The data center control module is used to analyze the real-time parameter data to obtain juice abnormality data and production parameter control strategies; the data center control module includes an abnormal characteristic database, a deep learning prediction unit and a parameter optimization control unit.

[0082] Furthermore, the abnormal characteristic database is stored in a relational database to store abnormal data, abnormal types and processing methods in various historical juice production processes, as shown in Table 1.

[0083] The deep learning prediction unit is used to use a long short-term memory network to learn and recognize patterns based on the real-time parameter data, and is trained in combination with historical data in the abnormal characteristics database to predict abnormal juice data. Through the LSTM model, it can not only detect current anomalies, but also predict potential quality risks in advance. This is especially important for the production of juice that is prone to browning and oxidation. It can achieve preventive control rather than a delayed passive response, which is significantly different from traditional threshold-based alarm systems.

[0084] Table 1 Example of abnormal characteristics database

[0085]

[0086] The parameter optimization control unit is used to generate a production parameter control strategy based on the predicted juice abnormality data and information in the abnormal characteristics database when the deep learning unit predicts an abnormality risk or the real-time data shows a deviation.

[0087] As an embodiment of the present invention, refer to Figure 2 The third step is applied to a multi-level control module of a real-time quality monitoring system for a juice production line based on the Internet of Things. The multi-level control module analyzes the abnormal data of the juice through the hierarchical analysis control model, controls the key parameters of the production line in combination with the production parameter control strategy, and generates a global control strategy.

[0088] Furthermore, the hierarchical analysis control model includes a data analysis layer, an anomaly analysis layer, a control strategy generation layer and a structure optimization layer;

[0089] The data analysis layer uses edge computing nodes to perform spatiotemporal alignment and feature fusion of real-time parameter data from different sensors and different workstations (fusing multiple related parameters into comprehensive features that can better reflect the system status); principal component analysis is used to extract key comprehensive indicators, such as fusing the sterilizer inlet temperature, outlet temperature, and steam pressure into a composite feature that characterizes the sterilization intensity.

[0090] The production line's physical state space is mapped in real time, constructing a multidimensional parameter tensor (a 6×N-dimensional matrix, where 6 represents the parameter type and N represents the time step). The data analysis layer provides differentiated data cleaning rules for temperature, pH, and sugar content. Temperature data is processed using a sliding window Kalman filter, pH data uses residual correction based on a process mechanism model, and sugar content data undergoes spectral noise removal.

[0091] The anomaly analysis layer combines the juice anomaly data with the isolation forest algorithm to analyze the anomaly pattern of the multidimensional parameter tensor, generating three types of fault sources: equipment-level anomalies, process-level anomalies, and raw material-level anomalies. The fault source data is referenced in Table 2. The fault source classification table is generated using the isolation forest algorithm to generate three types of fault sources: equipment-level, process-level, and raw material-level, demonstrating the accuracy of fault diagnosis.

[0092] The control strategy generation layer analyzes the fault source to obtain parameter adjustment boundary conditions, combines the production parameter control strategy to control the key parameters of the production line, and generates a global control strategy;

[0093] The structural optimization layer dynamically iterates the abnormal characteristic database and the deep learning prediction unit based on the fault source and the global control strategy.

[0094] Table 2 Fault source classification table

[0095]

[0096] Starting from the overall data, this method combines machine learning (isolation forest) and expert knowledge to perform layered and classified fault diagnosis. Compared with traditional single-parameter threshold alarms or simple rule matching, it significantly improves the accuracy and speed of fault location, effectively distinguishing between equipment problems, process setting issues, and raw material fluctuations. At the same time, through continuous learning from actual operations, the model and knowledge base are continuously optimized, enabling the system to cope with new operating conditions and unknown abnormal patterns, maintaining long-term efficiency and robustness.

[0097] As an embodiment of the present invention, refer to Figure 2 The fourth step is applied to a production line flexible control module of a real-time quality monitoring system for a juice production line based on the Internet of Things. The production line flexible control module is used to automatically adjust parameters according to the production parameter control strategy when the juice type is switched; the automatic parameter adjustment automatically controls the production process parameters according to the characteristics of the juice type; after the switch is completed, the juice quality is verified and the parameter optimization control is performed.

[0098] Furthermore, the specific process of automatically adjusting parameters according to the production parameter control strategy is referred to Figure 4 The specific process is:

[0099] Based on a database of juice material properties, a dynamic correlation matrix is established between production process parameters and juice physical and chemical indicators. Gradient similarity matching is then used to determine the optimal parameter baseline range for the target juice. Furthermore, based on a database of juice material properties (which stores the physical and chemical properties of different juices, such as viscosity, acidity, sugar content, and heat sensitivity), a dynamic correlation matrix is established between production process parameters (sterilization temperature, time, and filling pressure) and juice physical and chemical indicators (vitamin C retention, color, and flavor score). A gradient similarity matching algorithm is used to match the target juice's characteristics with records in the database, identifying the most similar historical production case or theoretical model, thereby determining the optimal parameter baseline range for the new juice. Gradient similarity matching may involve calculating the cosine similarity of the gradient vectors between the target juice's characteristic vector and those in the database. For example, for high-acid juices, a reverse parameter derivation model is used, first locking in the target vitamin C retention value and then inferring the sterilization temperature range. For high-sugar juices, an asymmetric adjustment window is established, implementing an exponential sensitivity gradient adjustment within ±5% of the sugar content baseline value.

[0100] The multi-objective collaborative optimization mechanism NSGA-II was employed to simultaneously optimize multiple objectives, such as maximizing vitamin C retention, minimizing the browning index, minimizing energy consumption, and minimizing adjustment time. Parameters were within a defined baseline range; sterilization temperature and pH constituted a strong constraint pair (for example, increasing temperature increases the minimum allowable pH to ensure sterilization effectiveness; the specific relationship was determined by fitting experimental data); and equipment capacity constraints (such as the maximum flow rate of a pump) were also present. Within the Pareto optimal solution set, the final parameter combination was selected based on pre-set priorities (for example, quality over energy consumption). For the strong constraint pair of sterilization temperature and pH, a fuzzy PID controller was used for coordinated regulation. The sugar content parameter was compared with the set target value based on real-time feedback from an online near-infrared spectrometer. A feedforward controller (based on a simple proportional model) then made compensatory adjustments downstream (such as during ingredient mixing) or upstream (for example, during pulp selection) to account for fluctuations in raw material sugar content. Equipment state coupling factors were introduced. For example, vibration sensor data from the filling equipment was monitored in real time to calculate the equipment wear coefficient. When the wear coefficient exceeds a preset threshold (such as 0.7), a constraint term that minimizes equipment stress (estimated through simulation) is automatically added to the objective function of NSGA-II, and the Lagrange multiplier method is used to reconstruct the parameter feasible domain to ensure that process adjustments do not accelerate equipment aging.

[0101] The real-time adjustment parameters are used to predict the adjustment quality fluctuation curve through physical-chemical coupling simulation to generate a parameter confidence evaluation matrix;

[0102] When the parameter confidence assessment matrix exceeds the threshold, the back propagation optimization algorithm is triggered. The parameter adjustment step size is dynamically corrected according to the second-order derivative characteristics of the quality fluctuation curve (the bending direction and degree of the curve can indicate the acceleration or deceleration of the adjustment trend). The optimization path characteristics (including initial state, adjustment sequence, final parameters, effect evaluation, etc.) are encoded as knowledge graph fragments and injected into the abnormal characteristics database to form a self-evolution mechanism for parameter adjustment rules.

[0103] Before performing parameter adjustments, the parameters to be adjusted in real time are input into a simplified physical-chemical coupling simulation model to simulate and predict the dynamic fluctuation curves of key quality indicators (such as outlet temperature, vitamin content, and color) during the adjustment process. A parameter confidence assessment matrix is generated based on the degree of deviation between the simulated predicted quality fluctuation curve and the target range. When the confidence assessment matrix shows that a parameter adjustment may cause the quality to exceed the acceptable range (for example, the confidence level is lower than the threshold of 0.9), an optimization algorithm based on the backpropagation idea is triggered (not training a neural network, but borrowing the idea of adjusting weights by backpropagation of errors). Specifically, the second-order derivative of the predicted quality fluctuation curve at the key point is calculated (characterizing the curvature and trend change of the curve). If the absolute value of the second-order derivative is too large (indicating drastic fluctuations or rapid trend changes), the adjustment step size of the parameter is dynamically reduced. Otherwise, the step size can be appropriately increased to achieve a smoother and safer parameter transition.

[0104] Each successful parameter optimization path (including the initial state, target juice properties, matched benchmark interval, NSGA-II optimization process, actual adjustment sequence, final parameters, and comparison of simulation predictions with actual quality verification results) is encoded as a structured knowledge graph fragment. This is then injected into the knowledge base associated with the abnormal properties database, forming a self-evolving mechanism for parameter adjustment rules. When similar switches are subsequently performed, the system can prioritize and reuse successful experiences from the knowledge graph.

[0105] Furthermore, after the switch is completed, the specific process of verifying the juice quality and performing parameter optimization control includes:

[0106] After the juice type is switched, collect samples of the newly produced juice; conduct quality inspections on the juice samples, including sensory evaluation (color), physical and chemical analysis (nutritional components such as vitamin content and sugar-acid ratio), and microbiological indicators (such as the total number of colonies per unit area); refer to Table 3 for specific data examples. The quality inspection and parameter adjustment record sheet records the closed-loop process of juice quality verification and parameter optimization.

[0107] Table 3 Quality inspection and parameter adjustment record

[0108] Test items Test results Qualified Parameter Adjustment Sensory evaluation Normal color yes none Physical and chemical analysis Insufficient sugar content no Increase the syrup ratio Microbiological analysis Total colony count exceeds the standard no Extend sterilization time

[0109] Compare the test results with the standard quality requirements to determine whether the juice quality is qualified;

[0110] If the juice quality is qualified, the current production parameter settings are recorded;

[0111] If it fails to meet the standards, the production parameters will be adjusted based on the test results and the information in the abnormal characteristics database, and the quality test will be carried out again until the juice quality meets the standards.

[0112] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A real-time monitoring system for juice production line quality based on the Internet of Things, characterized by: include: The data acquisition module collects real-time parameter data during the juice production process based on multiple parameter sensors distributed at key workstations of the juice production line, and pre-processes it through edge computing nodes; A data center control module is used to analyze the real-time parameter data to obtain juice abnormality data and production parameter control strategies; the data center control module includes an abnormality characteristic database, a deep learning prediction unit, and a parameter optimization control unit; The multi-level control module analyzes abnormal juice data through the hierarchical analysis control model, controls key parameters of the production line in combination with the production parameter control strategy, and generates a global control strategy; The flexible control module of the production line is used to perform automatic parameter adjustment according to the production parameter control strategy when the juice type is switched; the automatic parameter adjustment automatically controls the production process parameters according to the characteristics of the juice type; after the switch is completed, the juice quality is verified and the parameter optimization control is performed.

2. The real-time quality monitoring system for a juice production line based on the Internet of Things according to claim 1, characterized in that: The real-time parameter data includes temperature, humidity, pressure, flow rate, pH value and sugar content; the edge computing node preprocesses the collected real-time parameter data, including data cleaning, outlier removal and data standardization.

3. The real-time quality monitoring system for a juice production line based on the Internet of Things according to claim 1 is characterized by: The abnormal characteristics database stores abnormal data, abnormal types and processing methods in various historical juice production processes; The deep learning prediction unit is used to predict abnormal juice data based on the real-time parameter data and the abnormal characteristics database; The parameter optimization control unit is used to generate a production parameter control strategy based on the predicted juice abnormality data and the information in the abnormality characteristic database.

4. The real-time quality monitoring system for a juice production line based on the Internet of Things according to claim 1 is characterized by: The hierarchical analysis control model includes a data analysis layer, an abnormality analysis layer, a control strategy generation layer and a structure optimization layer; The data analysis layer uses edge computing nodes to perform spatiotemporal alignment and feature fusion on real-time parameter data, mapping the physical state space of the production line in real time and constructing a multi-dimensional parameter tensor; The anomaly analysis layer combines the juice anomaly data to perform an anomaly pattern analysis on the multi-dimensional parameter tensor, generating three types of fault sources: equipment-level anomaly, process-level anomaly, and raw material-level anomaly; The control strategy generation layer analyzes the fault source to obtain parameter adjustment boundary conditions, combines the production parameter control strategy to control the key parameters of the production line, and generates a global control strategy; The structural optimization layer dynamically iterates the abnormal characteristic database and the deep learning prediction unit based on the fault source and the global control strategy.

5. The real-time quality monitoring system for a juice production line based on the Internet of Things according to claim 1 is characterized by: The specific process of performing automatic parameter adjustment according to the production parameter control strategy is as follows: Based on the material property database of juice types, a dynamic correlation matrix between production process parameters and juice physical and chemical indicators is established, and the optimal parameter reference range of the target juice is determined through gradient similarity matching; A multi-objective collaborative optimization mechanism is used to construct a dynamic adjustment priority queue within the optimal parameter reference range. The sterilization temperature and pH value constitute a constraint pair, and parameter linkage adjustment is achieved through a fuzzy PID controller. The sugar content parameter is controlled by feedforward compensation based on online near-infrared spectroscopy feedback. The real-time adjustment parameters are used to predict the adjustment quality fluctuation curve through physical-chemical coupling simulation to generate a parameter confidence evaluation matrix; When the parameter confidence evaluation matrix exceeds the threshold, the back propagation optimization algorithm is triggered, and the parameter adjustment step size is dynamically corrected according to the second-order derivative characteristics of the quality fluctuation curve. The optimization path characteristics are encoded as knowledge graph fragments and injected into the abnormal characteristics database to form a self-evolution mechanism for parameter adjustment rules.

6. The real-time quality monitoring system for a juice production line based on the Internet of Things according to claim 1, characterized in that: After the switch is completed, the specific process of verifying the juice quality and performing parameter optimization control includes: After the juice type switch is completed, collect the newly produced juice samples; conduct quality testing on the juice samples, including sensory evaluation, physical and chemical analysis and microbiological indicators; Compare the test results with the standard quality requirements to determine whether the juice quality is qualified; If the juice quality is qualified, the current production parameter settings are recorded; If it fails to meet the standards, the production parameters will be adjusted based on the test results and the information in the abnormal characteristics database, and the quality test will be carried out again until the juice quality meets the standards.

7. A method for real-time monitoring of juice production line quality based on the Internet of Things, characterized in that: include: Based on multiple parameter sensors distributed at key workstations of the juice production line, real-time parameter data of the juice production process is collected and pre-processed through edge computing nodes; The real-time parameter data is analyzed to obtain juice abnormality data and production parameter control strategies; the data center control module includes an abnormality characteristic database, a deep learning prediction unit, and a parameter optimization control unit; The abnormal data of juice was analyzed by using the hierarchical analysis control model, and the key parameters of the production line were controlled by combining the production parameter control strategy to generate a global control strategy; When the juice type is switched, the parameters are automatically adjusted according to the production parameter control strategy; the parameter automatic adjustment automatically controls the production process parameters according to the characteristics of the juice type; after the switch is completed, the juice quality is verified and the parameter optimization control is performed.

Images