Dairy product production line control system based on internet of things

Abstract

The application discloses a dairy product production line control system based on Internet of Things and particularly relates to the technical field of dairy processing, which comprises the following steps: based on the real-time collection of the density, viscosity, fat content and temperature of incoming materials, a multi-parameter feedforward model is used to generate a reference filling duration, and a dynamic filling command is output through the weighing feedback PID compensation of bottle position delay matching; a valve body monitoring and micro-cleaning module evaluates the valve body fouling health index through machine learning of vibration and pressure signals, triggers online micro-cleaning on demand, and verifies the cleanliness through the closed-loop detection of conductivity and turbidity; a single package full-parameter traceability module binds each packaging unit with the filling process data, generates a distributed digital identity, stores the data fingerprint in a chain for evidence, and realizes single package level filling parameter visual traceability through a packaging two-dimensional code. The application improves the filling precision consistency and the comprehensive efficiency of the equipment, and refines the quality traceability granularity to each minimum packaging unit.

Description

Technical Field

[0001] This invention relates to the field of dairy processing technology, and more specifically, to a dairy production line control system based on the Internet of Things. Background Technology

[0002] In the filling stage of dairy product production lines, filling accuracy directly determines the finished product's net content qualification rate, making it a core indicator of quality control. Currently, conventional filling control systems have incorporated some automated instruments, but their control strategies still primarily rely on fixed parameters. During production, after upstream sterilization, homogenization, and temporary storage, the temperature, viscosity, and fat content of the milk entering the filling machine dynamically change due to fluctuations in operating conditions. Existing filling machines typically operate based solely on preset single filling time or volume parameters, failing to adaptively adjust to changes in the material properties, leading to frequent deviations in actual filling volume from the target value. On-site operations can only rely on manual periodic sampling and post-filling rejection, resulting in waste of packaging materials and supplies, and making it difficult to ensure the consistency of net content across the entire batch.

[0003] On the other hand, as a key component in the filling process, the filling valve's effective flow diameter changes due to the gradual deposition of scale on the valve core and seat during long-term continuous operation. This scaling phenomenon slowly affects filling speed and accuracy. Current valve cleaning strategies mostly rely on timed full-line shutdowns for CIP cleaning, lacking online assessment methods for the scaling status of individual valves and on-demand micro-cleaning capabilities. Full-line shutdown cleaning not only leads to unnecessary production capacity losses, but the cleaning interval is entirely dependent on experience. Cleaning too early wastes chemicals and production time, while cleaning too late results in valve blockage and batch-specific filling deviations. Therefore, a predictive maintenance solution capable of online sensing of valve status and timely intervention is urgently needed.

[0004] Furthermore, in terms of dairy product quality and safety traceability, the industry currently generally uses the method of printing production batch codes on packaging. When individual quality complaints arise, this traceability method can only pinpoint the production date and shift information at most, and cannot restore the complete processing parameters of each smallest packaging unit at the moment of filling, such as the actual filling time, valve number, and real-time material status. This crude batch traceability makes accurate root cause analysis difficult, preventing companies from quickly isolating the source of the problem and optimizing processes, and also failing to provide consumers with reliable and transparent quality proof of each package's parameters. Therefore, there is an urgent need for a technical means to achieve accurate traceability of all parameters at the single-package level. Summary of the Invention

[0005] To overcome the above-mentioned deficiencies of the prior art, embodiments of the present invention provide an Internet of Things-based dairy production line control system.

[0006] To achieve the above objectives, the present invention provides the following technical solution: The IoT-based dairy production line control system includes the following modules: The filling adaptive control module is used to collect the density, viscosity, fat content and temperature of the incoming filling material in real time. Based on the material characteristic feedforward model, it generates the benchmark filling valve opening time and combines it with the bottle position delay matching bottle weighing feedback compensation amount to calculate the filling command time and control the opening of the filling valve. The valve body monitoring and micro-cleaning module is used to collect vibration and dynamic pressure signals of the filling valve, evaluate the scaling health index of the valve body through a machine learning model, and trigger online micro-cleaning for the valve when the index is lower than a preset threshold. The cleanliness is verified by closed-loop detection of conductivity and turbidity. The single-package full-parameter traceability module is used to accurately bind each smallest packaging unit with the filling process data, generate distributed digital identity and digital fingerprint, store the fingerprint on the blockchain, and assign a QR code containing the digital identity to the packaging surface, supporting the visual traceability of single-package-level filling parameters.

[0007] Specifically, the filling adaptive control module includes: A Coriolis mass flow meter, an online viscosity sensor, an infrared fat content analyzer, and a temperature probe are sequentially installed on the main pipe of the annular distributor between the outlet of the filling buffer tank and the inlet of the filling valve. The signals from each sensor are connected to the edge IoT gateway via IO-Link or 4-20mA. The edge computing controller reads the real-time values ​​of the aforementioned sensors every 200 milliseconds to obtain the instantaneous density, dynamic viscosity, fat content, and temperature of the incoming material. These values ​​are then substituted into a multi-parameter filling time feedforward model to calculate the baseline filling valve opening time. The feedforward model consists of a theoretical baseline time multiplied by a correction function based on viscosity, temperature, and fat content.

[0008] Specifically, the correction function is pre-established in the following manner: Under laboratory conditions, for the same type of dairy product, with a fixed filling pressure difference, a full factorial experiment was conducted by changing the material temperature, viscosity, and fat content, and the valve opening time required to actually fill the target volume under each combination of conditions was recorded. The correction factor for each combination is calculated using the theoretical baseline duration as the denominator. Then, the correction function is obtained by fitting a multivariate quadratic polynomial regression with viscosity, temperature, and fat content as independent variables and the correction factor as the dependent variable.

[0009] Specifically, the edge computing controller is also used to combine the opening duration of the reference filling valve with the bottle weighing feedback; The edge controller maintains a first-in-first-out shift buffer queue of length N to compensate for the delay of N bottle positions between the filling station and the bottle weighing station. This ensures that the current weighing result is precisely bound to the filling command duration of the corresponding packaging unit N cycles ago. The deviation between the weighing result and the standard weight target value calculated from the target filling volume is used as input. The corresponding historical filling command duration is referenced, and an incremental PID algorithm with integral separation is used to generate the feedback compensation duration. Finally, the filling command duration is output.

[0010] Specifically, the valve body monitoring and micro-cleaning module includes: A triaxial MEMS accelerometer installed on the outer wall of the valve core actuation guide sleeve of each filling valve, and a miniature dynamic pressure sensor installed in the valve inlet chamber, the signal is sent to the edge computing controller at a sampling rate of 1kHz after low-frequency filtering. The edge computing controller captures the time-domain signal window of valve core impact and liquid pressure change during the opening-closing cycle of each filling valve. It uses short-time Fourier transform to extract the energy distribution and peak frequency of the preset frequency band, forming a multi-dimensional feature vector. This vector is then input into a pre-trained lightweight gradient boosting tree model, and outputs a valve body scaling health index with a value ranging from 0 to 1. The lower the value, the more severe the scaling.

[0011] Specifically, the training labels for the lightweight gradient boosting tree model are constructed as follows: The first filling cycle after the production line has undergone a complete standard CIP cleaning and the valve body has been confirmed to be clean is defined as a healthy state, and the label value is assigned to 1.0. The last filling cycle of the valve before it was forced to stop for maintenance due to continuous exceeding the filling volume tolerance was defined as a severe scaling state, and the label value was assigned to 0.0. The intermediate state label was generated by the equipment engineer through a linear mapping of the percentage of the scaling thickness to the effective cross-sectional area of ​​the valve seat flow channel recorded by the disassembly and inspection endoscope, based on the time interval between two adjacent full-line cleanings. This resulted in a continuous label value between 0 and 1 for supervised training.

[0012] Specifically, when the health index of any filling valve drops to a preset warning threshold, the edge controller generates an online micro-cleaning command: close the feed valve in front of the valve, open the bypass CIP micro-pipeline, inject heated dilute alkali solution and flush the valve core and valve seat in a pulsating flow, then discharge sterile pure water before sterile air purging, and use the micro conductivity sensor and turbidity sensor installed at the bypass return end to collect the conductivity sequence and turbidity sequence of the flushing solution, and calculate the conductivity stability index and turbidity peak residual; When both are below their respective preset thresholds, the micro-cleaning is deemed effective and purging continues; otherwise, additional pulsating flushing cycles are added until the target is met or an alarm is triggered.

[0013] Specifically, the single-package full-parameter traceability module includes: A high-speed industrial camera and an ultraviolet laser marking machine are deployed on the conveyor belt after the bottle exit star wheel of the filling machine. An edge traceability agent accurately binds the current packaging unit with the filling process data messages recorded by the filling adaptive control module based on the phase and timestamp of the rotary encoder of the filling machine. This generates a unique event data packet containing the packaging serial number, filling time, filling valve number, filling command duration, incoming material temperature, viscosity, fat content, density, and post-filling weight. A hash algorithm is used to calculate a digital fingerprint to generate a distributed digital identity for the packaging that conforms to the W3C DID specification. The digital fingerprint and digital identity are then packaged into a transaction and submitted to the smart contract of the consortium blockchain through the edge blockchain client for notarization. At the same time, the laser marking machine engraves a QR code containing digital identity information on the surface of the packaging.

[0014] Specifically, after consumers or quality inspectors scan the QR code, the traceability code parser obtains the evidence record from the consortium blockchain and pulls the corresponding complete filling and processing parameter details from the cloud manufacturing execution system. The processing curve and data of a single package from the moment of filling are displayed in a visual chart on the front-end page.

[0015] The technical effects and advantages of this invention are as follows: This invention effectively overcomes three major shortcomings of existing technologies: the significant impact of material characteristic fluctuations on filling accuracy, the reliance on timed shutdowns for cleaning to maintain filling valve scaling, and the excessively coarse particle size in traceability. Specifically, the adaptive filling control module utilizes a multi-parameter feedforward model and a weighing feedback composite control system with precise bottle position delay matching. This allows the filling time to automatically adjust to real-time changes in incoming material density, viscosity, fat content, and temperature, significantly reducing net content deviations caused by material characteristic fluctuations and minimizing packaging material waste and manual sampling frequency. The valve body monitoring and micro-cleaning module, based on machine learning diagnosis of vibration and pressure signals, enables online real-time assessment of the scaling status of individual filling valves and on-demand online micro-cleaning. Closed-loop verification using conductivity and turbidity ensures cleaning cleanliness, completely changing the traditional extensive mode of full-line shutdown cleaning and significantly improving overall equipment efficiency and the ability to maintain filling consistency.

[0016] Meanwhile, the single-package full-parameter traceability module precisely binds each smallest packaging unit to its complete processing data at the moment of filling, and generates an immutable single-package traceability chain through distributed digital identity and blockchain notarization technology. This refines the granularity of quality traceability from the batch level to each individual product. Combined with the digital identity QR code on the packaging, consumers and quality inspectors can instantly obtain the filling parameter details and processing curve of the single package, providing reliable data support for accurate root cause analysis and continuous process improvement, and significantly enhancing consumers' transparent trust in product quality. Attached Figure Description

[0017] Figure 1This is a system block diagram of the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] like Figure 1 As shown, the control system modules for a dairy production line based on the Internet of Things are as follows: The filling adaptive control module collects the incoming material density, viscosity, fat content and temperature in real time through online sensors, and uses the data to generate a feedforward baseline filling time by substituting it into a pre-fitted correction function. At the same time, it uses a shift buffer queue to accurately match the bottle weighing feedback, and generates a compensation time through an incremental PID algorithm. The two are combined to output the filling command, dynamically adapting to fluctuations in material characteristics.

[0020] The specific steps are as follows: Step S11: On the main pipe of the annular distributor between the outlet of the filling buffer tank and the inlet of the filling valve, install a Coriolis mass flow meter, an online viscosity sensor, an infrared fat content analyzer and a Pt100 temperature probe in sequence. The signals of each sensor are connected to the edge IoT gateway through IO-Link or 4-20mA. Step S12: The edge computing controller reads the real-time values ​​of the above sensors every 200 milliseconds to obtain the instantaneous density of the incoming material. Dynamic viscosity Fat content and temperature And input it into a multi-parameter filling time feedforward model that has been established in advance through regression experiments. The model expression is: ,in The reference filling valve opening time, For the target filling volume, The valve body flow coefficient, For the equivalent flow-throttling area, The filling pressure difference is calculated in real time by the differential pressure between the pressure transmitter installed in the gas phase space at the top of the filling buffer tank and the back pressure sensor in the sealed chamber of the filling machine. The update cycle is synchronized with the sensor reading cycle at 200 milliseconds. Based on correction functions for viscosity, temperature, and fat content, output the baseline filling valve opening time for the current material state. Correction function The method of obtaining the data is as follows: Under laboratory conditions, for the same type of dairy product, under different filling pressures... On a fixed filling platform, the material temperature is changed by a constant temperature water bath and by adjusting different homogenization pressures. (Range 4℃-25℃), viscosity (Range 1.5-8 mPa·s) and fat content (Range 0.1%-4.0%), conduct a full factorial experiment and record the actual filling volume of the target volume under each combination of conditions. Required valve opening time Based on theoretical benchmark duration Use the denominator to calculate the correction factor for each combination. , and then with For independent variable, As the dependent variable, a multivariate quadratic polynomial regression was used. The coefficients were obtained by fitting. to And use the fitted polynomial as the correction function. Online invocation within an embedded edge computing controller. As the theoretical baseline duration, For the target filling volume, The valve body flow coefficient, This is the correction factor for the i-th group of experiments. This refers to the actual valve opening time. to is the regression coefficient.

[0021] Step S13, This is combined with the net content deviation reported by the weighing sensor at the bottle outlet of the previous canned product. Since there are N bottle positions (N is an integer, depending on the number of position intervals in the filling machine's rotation direction and the equivalent number of bottles in the conveyor belt transition section) between the filling station and the bottle outlet weighing station, the current weighing sensor outputs the actual filled weight. This actually corresponds to the packaging unit that was filled N filling cycles ago, not the immediately preceding canned product. Therefore, the edge controller maintains a first-in-first-out shift buffer queue Q of length N. Each filling cycle will use the output from step S12 of that cycle... Values ​​and generated in step S13 The value is enqueued as a set of control parameter items, and historical control parameter items from N cycles ago are retrieved from the head of the queue. Let the current cycle number be k, then the current output of the load cell is the actual weight of the fill. The corresponding historical filling command duration is The index of the incoming material parameters used is The sensor dataset at any given time. Feedback compensation duration. The generation formula is revised as follows: ;in, The feedback compensation duration for the k-th period. Number the current filling cycle. The target value for standard weight. This represents the actual weight of the container output by the weighing sensor in the k-th cycle. The filling command duration for the (kN)th cycle. This is the bottle position delay between the filling station and the weighing station. This represents an incremental PID algorithm that takes the filling weight deviation as input and performs integral separation constraint with reference to the duration of historical filling orders; The weighing sensor will measure the actual weight of the filling. The data is transmitted to the edge controller, which calculates the volume deviation. The feedback compensation time is generated through an incremental PID algorithm that performs dead-zone processing and integral separation. The final output is the duration of this filling command. ;in For the duration of the filling command, The reference filling valve opening time, To provide feedback on compensation duration, The nominal density of this dairy product at a standard temperature of 4°C is provided by the formulation parameter library of the manufacturing execution system. If the nominal density is not defined for the current product, the average density measured continuously for 60 seconds at a constant temperature of 4°C by a Coriolis mass flow meter before the first filling of the shift will be used. ; For volume deviation, This is the actual filling weight. This is the target value for standard weight.

[0022] Step S14: The edge controller transmits data via the EtherCAT bus. The real-time control word is written into the PLC of the filling machine, and the PLC drives the solenoid valve corresponding to the filling valve to press... The timer is activated, a single filling is completed, and the usage of this batch is recorded. The instantaneous values ​​of each sensor are used to form a set of filling process data messages and sent to the traceability module. By quickly compensating for filling volume deviations caused by fluctuations in material characteristics through feedforward, and then eliminating the remaining errors through feedback, the filling volume qualification rate is significantly improved.

[0023] The valve body monitoring and micro-cleaning module collects the working signals of each filling valve through vibration and pressure sensors. After extracting features through short-time Fourier transform, the signals are input into a lightweight gradient boosting tree model to evaluate the valve body scaling health index in real time. When the index drops to the threshold, online micro-cleaning is triggered, and the cleaning effect is verified in a closed loop using conductivity and turbidity sensors.

[0024] The specific steps are as follows: Step S21: Attach a triaxial MEMS accelerometer to the outer wall of the valve core action guide sleeve of each filling valve, and install a miniature dynamic pressure sensor in the inlet chamber of the valve. The accelerometer and pressure sensor signals are sent to the edge computing controller at a sampling rate of 1kHz after low-frequency filtering. Step S22: The edge computing controller extracts a time-domain signal window containing valve core impact and liquid pressure changes for each filling valve during each opening-closing cycle. It uses short-time Fourier transform to extract the energy distribution and peak frequency in the 0-500Hz frequency band, constructing a multi-dimensional feature vector. This vector is then input into a lightweight gradient boosting tree model pre-trained based on historical cleaning cycle data. This model outputs the current valve body scaling health index H, where H ranges from 0 to 1, with lower values ​​indicating more severe scaling. The training labels for the lightweight gradient boosting tree model are constructed as follows: after the production line performs a complete standard CIP full-line cleaning and confirms the valve body is clean, the first filling valve... The filling cycle is defined as a healthy state, with a label value of 1.0. The last filling cycle of the valve before it is forced to stop for maintenance due to continuous filling volume exceeding the tolerance range is defined as a severe scaling state, with a label value of 0.0. Intermediate state labels are obtained as follows: between two adjacent full-line cleanings, the operating cycle is divided into several intervals by dividing the time equally. The equipment engineer performs a linear mapping based on the percentage of the scaling thickness to the effective cross-sectional area of ​​the valve seat flow channel recorded by the disassembly and inspection endoscope. A scaling thickness percentage of 0% is mapped to 1.0, and a percentage of more than 30% is mapped to 0.0. This generates continuous label values ​​between 0 and 1, which are used for supervised training of the model.

[0025] Step S23: When the health index H of any filling valve drops to the preset warning threshold When the value is 0.35 (e.g.), the edge controller immediately generates an online micro-cleaning command for the valve. The micro-cleaning command includes closing the upstream feed valve of the valve, opening the bypass CIP micro-line connected to the valve, injecting 60°C dilute alkali solution and rinsing the valve core and valve seat in a pulsating flow for 4 seconds, and then purging with sterile air. A miniature conductivity sensor and a turbidity sensor are installed at the return end of the bypass CIP micro-pipeline. After the pulsating flow flushing ends and before the sterile air purging begins, the edge controller closes the outlet valve of the pulsation generator and opens the bypass pure water valve, injecting sterile pure water at a constant pressure of 0.3 MPa. After flowing through the valve core and valve seat, the water is discharged from the return end. The injection time is... = 2 seconds; during this period, the conductivity sensor and turbidity sensor acquire the conductivity sequence of the reflux liquid at a sampling rate of 200 Hz. and turbidity sequence The edge controller calculates the cleanliness assessment criteria: Conductivity stability index: ;in The conductivity stability index, Duration of pure water injection For time variables, For the time series of reflux liquid conductivity, The reference conductivity value; Turbidity peak residual: ;in The residual value of the turbidity peak. This is a time series of turbidity in the reflux liquid. The baseline conductivity value for sterile pure water at the same temperature is pre-calibrated and stored in the edge controller parameter table. When simultaneously meeting... and hour( This is the conductivity stability threshold, typically 5 μS / cm·s; (Turbidity threshold, typical value 10 NTU) is used to determine that micro-cleaning is effective. The edge controller opens the sterile air valve to purge and drain the cleaned air, and sets the cleaning flag to successful. Otherwise, if the residue exceeds the standard, an additional pulsed flushing cycle will be automatically added, and the pure water test will be performed again. If the standard is still not met after two consecutive additional cycles, an alarm will be triggered and the production line operator will be notified. Simultaneously, the valve will be marked as awaiting offline maintenance. After cleaning is completed, the closed-loop verification will be performed... and The values ​​are uploaded to the cloud platform's device management module along with the cleaning event for archiving and for cleaning parameter optimization and analysis.

[0026] The dilute alkaline solution is a sodium hydroxide solution with a mass concentration of 1.5%-2.5%, injected at a temperature range of 55℃-70℃. The pulsating flow is generated by an electromagnetic diaphragm pulsation generator with a pulsation frequency of 2-6 Hz, a peak pulsation pressure of 0.3-0.6 MPa, and a flushing duration of 3-8 seconds. The specific values ​​of the above parameters are linearly adjusted according to the scaling health index H: when H is in the range of 0.25-0.35, a higher concentration (2.5%), a higher temperature (68-70℃), a longer flushing time (6-8 seconds), and a higher pulsation frequency (5-6 Hz) are used; when H is in the range of 0.35 to 0.50, a lower concentration (1.5%-2.0%), a lower temperature (55-60℃), a shorter flushing time (3-5 seconds), and a lower pulsation frequency (2-4 Hz) are used to reduce chemical and energy costs.

[0027] Step S24: The filling machine speed is not stopped throughout the micro-cleaning process. Instead, the cam or bottle-missing mechanism automatically skips the bottle position corresponding to the valve to be cleaned, while the remaining valves continue normal filling. After cleaning, the feed valve is reopened, and its health index is temporarily reset to 1.0. Simultaneously, the edge controller packages and uploads information such as the micro-cleaning event's occurrence time, valve number, and pre-cleaning health index to the cloud platform for archiving by the equipment management module. After the temporary reset, the valve's scaling health index H is not directly used as 1.0 in subsequent operations. Instead, vibration and pressure signals are re-acquired starting from the first filling cycle after micro-cleaning, and a lightweight gradient boosting tree model outputs new values ​​cycle by cycle, while a transition smoothing coefficient is set. ,make ,in A health index for smoothed display / use. This is the transition smoothing coefficient. The model outputs a health index; after smoothing for 10 consecutive cycles, it is completely switched back to the original output value to avoid falsely low H values ​​caused by initial data noise. Step S22 is then repeated for continuous monitoring.

[0028] It enables early detection and online micro-cleaning of scale buildup in filling valves, significantly reducing filling volume deviations and downtime for cleaning caused by valve blockage.

[0029] The single-package full-parameter traceability module accurately binds each packaging unit to the filling process data through an edge traceability agent, generates a distributed digital identity that conforms to the W3C DID specification, submits the data fingerprint to the consortium blockchain for notarization, and assigns a code to the packaging surface; consumers can scan the code to obtain the complete filling and processing parameter details of the single package from the blockchain;

[0030] The specific steps are as follows: Step S31: Deploy a high-speed industrial camera and an ultraviolet laser marking machine on the conveyor belt after the bottle exit star wheel of the filling machine. Whenever a filled and sealed dairy product package passes by the trigger photoelectric sensor, the edge tracing agent accurately binds the current package with the filling process data message recorded in Module 1 based on the phase and timestamp of the filling machine's rotary encoder, generating a unique event data packet for the package, which includes the package serial number UID, filling time, filling valve number, and actual... Incoming material temperature, viscosity, fat content, density, and weight after filling; Step S32: The edge traceability agent uses a hash algorithm to calculate the digital fingerprint of the above event data packet, and calls the distributed identity generation service of the package to generate the DID Document of the smallest package of the dairy product according to the W3C DID specification. The method of the DID is "dairy", the method-specific identifier is based on the hash concatenation of UID and production line code, and the public key and authentication information are stored in the DID Document. Step S33: The edge traceability agent sends the plaintext and digital fingerprint of the event data packet to the cloud manufacturing execution system simultaneously via the MQTT protocol. At the same time, it packages the digital fingerprint, DID, timestamp, and production line ID into a lightweight transaction and submits it to the smart contract of the consortium blockchain through the edge blockchain client. After the smart contract verifies the signature, it writes this tamper-proof record into the block. Step S34: The laser marking machine engraves a QR code containing DID information on the surface of the packaging. After the consumer or quality inspector scans the code, the traceability code parser will obtain the transaction record from the blockchain and pull the corresponding complete filling and processing parameter details from the cloud MES. The front-end page displays the detailed processing curve and data of the bottle of dairy product from the moment of filling in a visual chart, truly realizing transparent traceability at the single package level.

[0031] By leveraging IoT event-driven processes and blockchain-based evidence storage, traceability is refined from batch to the smallest packaging unit, providing solid technical support for accurate quality attribution and consumer trust.

[0032] The above formulas are all dimensionless calculations. Dimensionless calculations can be performed using various methods such as standardization, which will not be elaborated here. The formulas are derived from software simulations based on a large amount of collected data, and the preset parameters in the formulas can be set by those skilled in the art according to the actual situation.

[0033] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, ATA hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. The semiconductor medium can be a solid-state ATA hard disk.

[0034] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0035] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0036] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0037] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0038] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0039] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable ATA hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0040] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A dairy production line control system based on the Internet of Things, characterized in that, Includes the following modules: The filling adaptive control module is used to collect the density, viscosity, fat content and temperature of the incoming filling material in real time. Based on the material characteristic feedforward model, it generates the benchmark filling valve opening time and combines it with the bottle position delay matching bottle weighing feedback compensation amount to calculate the filling command time and control the opening of the filling valve. The valve body monitoring and micro-cleaning module is used to collect vibration and dynamic pressure signals of the filling valve, evaluate the scaling health index of the valve body through a machine learning model, and trigger online micro-cleaning for the valve when the index is lower than a preset threshold. The cleanliness is verified by closed-loop detection of conductivity and turbidity. The single-package full-parameter traceability module is used to accurately bind each smallest packaging unit with the filling process data, generate distributed digital identity and digital fingerprint, store the fingerprint on the blockchain, and assign a QR code containing the digital identity to the packaging surface, supporting the visual traceability of single-package-level filling parameters.

2. The IoT-based dairy production line control system according to claim 1, characterized in that, The filling adaptive control module includes: A Coriolis mass flow meter, an online viscosity sensor, an infrared fat content analyzer, and a temperature probe are sequentially installed on the main pipe of the annular distributor between the outlet of the filling buffer tank and the inlet of the filling valve. The signals from each sensor are connected to the edge IoT gateway via IO-Link or 4-20mA. The edge computing controller reads the real-time values ​​of the aforementioned sensors every 200 milliseconds to obtain the instantaneous density, dynamic viscosity, fat content, and temperature of the incoming material. These values ​​are then substituted into a multi-parameter filling time feedforward model to calculate the baseline filling valve opening time. The feedforward model consists of a theoretical baseline time multiplied by a correction function based on viscosity, temperature, and fat content.

3. The IoT-based dairy production line control system according to claim 2, characterized in that, The correction function is pre-established in the following manner: Under laboratory conditions, for the same type of dairy product, with a fixed filling pressure difference, a full factorial experiment was conducted by changing the material temperature, viscosity, and fat content, and the valve opening time required to actually fill the target volume under each combination of conditions was recorded. The correction factor for each combination is calculated using the theoretical baseline duration as the denominator. Then, the correction function is obtained by fitting a multivariate quadratic polynomial regression with viscosity, temperature, and fat content as independent variables and the correction factor as the dependent variable.

4. The IoT-based dairy production line control system according to claim 2, characterized in that, The edge computing controller is also used to combine the opening time of the reference filling valve with the bottle weighing feedback; The edge controller maintains a first-in-first-out shift buffer queue of length N to compensate for the delay of N bottle positions between the filling station and the bottle weighing station. This ensures that the current weighing result is precisely bound to the filling command duration of the corresponding packaging unit N cycles ago. The deviation between the weighing result and the standard weight target value calculated from the target filling volume is used as input. The corresponding historical filling command duration is referenced, and an incremental PID algorithm with integral separation is used to generate the feedback compensation duration. Finally, the filling command duration is output.

5. The IoT-based dairy production line control system according to claim 1, characterized in that, The valve body monitoring and micro-cleaning module includes: A triaxial MEMS accelerometer installed on the outer wall of the valve core actuation guide sleeve of each filling valve, and a miniature dynamic pressure sensor installed in the valve inlet chamber, the signal is sent to the edge computing controller at a sampling rate of 1kHz after low-frequency filtering. The edge computing controller captures the time-domain signal window of valve core impact and liquid pressure change during the opening-closing cycle of each filling valve. It uses short-time Fourier transform to extract the energy distribution and peak frequency of the preset frequency band, forming a multi-dimensional feature vector. This vector is then input into a pre-trained lightweight gradient boosting tree model, and outputs a valve body scaling health index with a value ranging from 0 to 1. The lower the value, the more severe the scaling.

6. The IoT-based dairy production line control system according to claim 5, characterized in that, The training labels for the lightweight gradient boosting tree model are constructed as follows: The first filling cycle after the production line has undergone a complete standard CIP cleaning and the valve body has been confirmed to be clean is defined as a healthy state, and the label value is assigned to 1.

0. The last filling cycle of the valve before it was forced to stop for maintenance due to continuous exceeding the filling volume tolerance was defined as a severe scaling state, and the label value was assigned to 0.

0. The intermediate state label was generated by the equipment engineer through a linear mapping of the percentage of the scaling thickness to the effective cross-sectional area of ​​the valve seat flow channel recorded by the disassembly and inspection endoscope, based on the time interval between two adjacent full-line cleanings. This resulted in a continuous label value between 0 and 1 for supervised training.

7. The IoT-based dairy production line control system according to claim 5, characterized in that, When the health index of any filling valve drops to a preset warning threshold, the edge controller generates an online micro-cleaning command: close the feed valve in front of the valve, open the bypass CIP micro-pipeline, inject heated dilute alkali solution and flush the valve core and valve seat in a pulsating flow, then discharge sterile pure water before sterile air purging, and use the micro conductivity sensor and turbidity sensor installed at the bypass return end to collect the conductivity sequence and turbidity sequence of the flushing solution, and calculate the conductivity stability index and turbidity peak residual; When both are below their respective preset thresholds, the micro-cleaning is deemed effective and purging continues; otherwise, additional pulsating flushing cycles are added until the target is met or an alarm is triggered.

8. The IoT-based dairy production line control system according to claim 1, characterized in that, The single-package full-parameter traceability module includes: A high-speed industrial camera and an ultraviolet laser marking machine are deployed on the conveyor belt after the bottle exit star wheel of the filling machine. An edge traceability agent accurately binds the current packaging unit with the filling process data messages recorded by the filling adaptive control module based on the phase and timestamp of the rotary encoder of the filling machine. This generates a unique event data packet containing the packaging serial number, filling time, filling valve number, filling command duration, incoming material temperature, viscosity, fat content, density, and post-filling weight. A hash algorithm is used to calculate a digital fingerprint to generate a distributed digital identity for the packaging that conforms to the W3C DID specification. The digital fingerprint and digital identity are then packaged into a transaction and submitted to the smart contract of the consortium blockchain through the edge blockchain client for notarization. At the same time, the laser marking machine engraves a QR code containing digital identity information on the surface of the packaging.

9. The IoT-based dairy production line control system according to claim 8, characterized in that, After consumers or quality inspectors scan the QR code, the traceability code parser obtains the evidence record from the consortium blockchain and pulls the corresponding complete filling and processing parameter details from the cloud manufacturing execution system. The processing curve and data of a single package from the moment of filling are displayed in a visual chart on the front-end page.

Images