Brittleness baking control process based on three-section variable-temperature hot air

Abstract

According to the brittleness baking control process based on the three-section variable-temperature hot air, the hot air flow field and material moisture migration are accurately regulated and controlled, and integrated control over brittleness structure shaping and energy-saving baking of products such as pet chews and jerky is achieved. A hardware architecture surrounds a'sensing-decision-execution 'closed-loop design, each part is tightly coupled with a central controller through an industrial bus, a dual-objective optimization system of brittleness formation and energy consumption control is constructed according to the high-protein and multi-fiber material characteristics of pet food, and the energy consumption of the pet food is controlled on the premise of ensuring that the brittleness index C of the pet food is greater than or equal to 65. Through a heat recovery unit, intermittent countercurrent air supply and LSTM feedforward control, the comprehensive energy consumption is reduced compared with the traditional process, and the unit brittleness energy consumption is reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of energy-saving baking technology, in particular to a three-stage variable-temperature hot air brittleness baking energy-saving oven and a three-stage variable-temperature hot air brittleness baking control process. BACKGROUND

[0002] Traditional baking equipment generally adopts a single-stage constant-temperature hot air process, which is to continuously bake after one-time preheating to a fixed temperature (usually 100-120℃). There are three major energy-saving defects:

[0003] 1) Low heat recovery efficiency, the heat exchange efficiency of traditional ovens is less than 40%, the waste heat utilization rate is only 25%, and a large amount of high-temperature exhaust gas is directly discharged;

[0004] 2) Poor temperature control accuracy, temperature difference fluctuation up to ±5℃, resulting in excessive heating energy consumption redundancy (single batch baking energy consumption up to 80kWh / kg);

[0005] 3) Insufficient intelligence level, only 30% of the equipment is equipped with basic energy consumption monitoring, relying on manual experience adjustment, no dynamic power control, and high unit product energy consumption ratio. SUMMARY

[0006] In order to solve the technical problems existing in the prior art, the present application provides the following technical solutions:

[0007] On the one hand, a three-stage variable-temperature hot air brittleness baking energy-saving oven is provided, comprising:

[0008] (1) Thermal and circulation unit: containing heating array arranged in subareas (preheating area and dehydration area), main fan and frequency converter, the heating array is connected with PLC digital output module through AC contactor, and power adjustment is realized through PLC controller to achieve 0-10V continuous control;

[0009] (2) Sensing detection network: composed of thermocouple, humidity sensor and weight detection unit, the thermocouple is connected with PLC through PLC controller, and the humidity sensor is connected with edge computing gateway through I2C protocol;

[0010] (3) Actuator group: including electric damper (0-90° opening), dehumidification electromagnetic valve and plate heat exchanger, all of which are realized through PLC controller to achieve closed-loop control;

[0011] (4) Central control platform: using PLC as the lower computer, communicating with frequency converter, HMI and edge computing gateway through PROFIBUS / PROFINET bus.

[0012] Further, the interactive control mode of the PLC, edge computing gateway and HMI includes:

[0013] (1) The PLC outputs a 0-10V signal to adjust the heating wire power through the PLC controller;

[0014] (2) The edge computing gateway receives 18 sensor data every 500ms, calculates and optimizes the set value based on the brittle formation quantization model, and feeds it back to the PLC;

[0015] (3) The HMI enables parameter setting (temperature 65-85℃, wind speed 0.5-1.2m / s), recipe calling and real-time monitoring functions.

[0016] Furthermore, the temperature sensor is connected to the thermocouple via a compensating wire, the humidity sensor is connected to the edge computing gateway via an I2C bus, and all data is transmitted via PROFINET industrial Ethernet with a communication rate of ≥100Mbps.

[0017] The main fan and the frequency converter communicate via PROFIBUS DP bus, and the speed closed-loop control accuracy is ±0.05m / s;

[0018] The actuator adopts a hybrid control of analog (0-10V) and digital (24V DC) signals, and the damper opening and wind speed are coordinated to form a spiral hot air flow field.

[0019] On the other hand, a brittle baking control process based on the above-mentioned energy-saving oven is provided, including the following steps:

[0020] (1) Preheating and shaping stage: control the temperature at 65±2℃ and the time at 8-10min. The PLC executes the adaptive fuzzy PID control law through the PLC controller to adjust the power of the heating wire. The main fan maintains a wind speed of 0.5m / s to form a surface protein film of 0.15±0.02mm.

[0021] (2) Dehydration and embrittlement stage: Heat to 85±1℃ at a rate of 1.0℃ / min, maintain for 15-18min, control humidity at 30±2%RH, execute embrittlement quantification model C, and when C is greater than the threshold, reduce heating power by 5% and increase target humidity to 32%RH;

[0022] (3) Stable cooling stage: control the temperature at 45±2℃ for 6-8 min, reduce the wind speed in three stages, execute the system-level energy-saving optimization objective function, recover the heat of the exhaust gas through the plate heat exchanger, and determine that baking is complete when the humidity fluctuation is ≤±2%RH, the weight change rate |dW / dt|<0.05g / min and the brittleness index C=68-72.

[0023] Furthermore, the expression for the adaptive fuzzy PID control law is:

[0024] ,in:

[0025] (1) u(k): The controller output value at time k (corresponding to the heating power percentage or actuator control signal);

[0026] (2) e(k): System deviation at time k, T_set is the set temperature, and T_actual is the actual temperature measured by the OMEGA thermocouple;

[0027] (3) EC(k): Error change rate at time k, EC(k) = e(k) - e(k-1) (℃ / sampling period), sampling period is 200ms;

[0028] (4) Kp(k), Ki(k), Kd(k): dynamic PID parameters, which are the proportional, integral, and derivative coefficients at time k, respectively;

[0029] (5) k: Discrete time index, representing the kth sampling time, synchronized with the PLC scanning cycle.

[0030] Furthermore, the expression for the brittleness formation quantification model is:

[0031] ,in:

[0032] (1) C: Brittleness index, target range 65-75, characterizing the fracture strength of pet food;

[0033] (2) k1, k2, k3: weighting coefficients, satisfying Where k1=0.35 (heating rate weight), k2=0.42 (moisture gradient weight), and k3=0.23 (humidity control accuracy weight);

[0034] (3) ΔT / Δt: Heating rate during the dehydration and embrittlement stage (°C / min), preferably 1.0°C / min;

[0035] (4) ΔM / Δx: Moisture gradient inside pet food (% / mm), The calculations show that Hcenter is the central humidity and Hsurface is the surface humidity.

[0036] (5) : Humidity control accuracy integral term ((%RH)²·s), H_set is the set humidity, H_actual is the actual humidity measured by the SENSIRION SHT85 sensor, and t_total is the total time of the dehydration stage (s).

[0037] (6) τ: Integral time variable (s), representing the cumulative effect of humidity deviation over time.

[0038] Furthermore, the expression for the system-level energy-saving optimization objective function is as follows:

[0039] ,in:

[0040] (1) J: Value of the energy-saving optimization objective function;

[0041] (2) These are energy consumption weight, quality weight, and time weight, respectively, and satisfy the following conditions: ;

[0042] (3) Eactual: Actual total energy consumption (kW·h), Etarget: Target energy consumption (kW·h);

[0043] (4) Qquality = C_measured / C_target (quality score), Qstandard = 1.0 (standard score);

[0044] (5) tprocess: actual process time (min), tbenchmark: benchmark process time (min).

[0045] Furthermore, the steps of the preheating and shaping stage include:

[0046] (1) Hardware initialization: The PLC activates the heating array in the preheating zone (including at least 2 sets of KANTHAL APM 1700 heating wires, each with a power of 1.2kW), with an initial power duty cycle of 45%, and the main fan maintains a wind speed of 0.5m / s through the frequency converter;

[0047] (2) Multidimensional sensing: Thermocouples collect temperature data in 100ms cycles, and humidity sensors upload data to the PLC after preprocessing by the edge computing gateway;

[0048] (3) Adaptive closed-loop control: When the temperature deviation e(k) > 1℃, the output voltage of the PLC controller increases linearly (sensitivity 0.5V / ℃), and the dynamic response time is ≤500ms; based on the initial weight W0 of the material, the LSTM model predicts the wind speed correction value Δv;

[0049] (4) Protein membrane determination: When the surface protein membrane coverage is ≥92% and the thickness is 0.15±0.02mm, the transmission phase is completed.

[0050] (5) Energy efficiency optimization: When the stage energy consumption is greater than 0.15 × total target energy consumption, activate the electric heat tracing compensation of the insulation layer (50W / m²) and reduce the wind speed to 0.45m / s.

[0051] Furthermore, the steps of the dehydration and embrittlement stage include:

[0052] (1) Gradient heating: The PLC controls the power of the heating wire through the PLC controller to raise the temperature from 65℃ to 85℃ at a rate of 1.0℃ / min, and adjusts it in two stages (1.0℃ / min for the first 5 minutes and 0.5℃ / min for the next 10 minutes), with temperature fluctuation ≤ ±0.5℃;

[0053] (2) Humidity gradient control: The humidity sensor samples at a frequency of 1Hz. When the standard deviation of the humidity field σH > 3%RH, the wind speed is finely adjusted (Δv = 0.1m / s). The solenoid valve adjusts its opening 30 seconds in advance according to the LSTM prediction. The target humidity is 30±2%RH.

[0054] (3) Counter-current air supply execution: The electric damper executes an intermittent program: forward rotation for 30s (opening degree 60°, wind speed 1.2m / s) → stop airflow for 5s → reverse rotation for 30s (opening degree 45°, wind speed 1.0m / s), and the cycle is controlled by the PLC timer.

[0055] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following:

[0056] (1) Deep integration: Hardware (sensors, actuators) and software (PID, LSTM) work together to meet the brittleness requirements of pet food. For example, real-time data from the SENSIRION humidity sensor is directly used in the LSTM model to predict the brittleness inflection point. The BELIMO damper and the SMC VX2120 solenoid valve work together to regulate the moisture gradient, forming a closed loop of "perception-decision-execution" for brittleness control.

[0057] (2) By real-time control of the brittleness model C, the uniformity of brittleness of pet food can reach more than 90%, the standard deviation of fracture strength can be ≤5%, and the moisture content can be controlled at 8±0.5%, thus solving the problem of large brittleness difference in traditional processes.

[0058] (3) Dynamic optimization: All key control parameters (such as PID parameters) are not fixed, but are dynamically adjusted according to real-time process data, so that the system always operates in the optimal or near-optimal state.

[0059] (4) Energy saving and high efficiency: Under the premise of ensuring the pet food crispness index C≥65, through heat recovery unit (ηrec≥85%), intermittent counterflow air supply (fan energy consumption reduced by 22%) and LSTM feedforward control (dehumidification energy consumption reduced by 28%), the overall energy consumption is reduced by 20%-25% compared with the traditional process, and the energy consumption per unit crispness is reduced by 18%-22%.

[0060] (5) Intelligent Foresight: The LSTM prediction model enables the system to "foresee the future" and can implement feedforward control for the pet food crisping process, thereby reducing the deviation caused by lag and improving the stability and accuracy of the brittle structure. LSTM Application Principle: Long sequence dependency modeling is achieved through gated recurrent units (GRU). For the slow time-varying characteristics of parameters such as temperature, humidity and weight in the pet food baking process, a bidirectional LSTM structure is used to capture the correlation between time series data. During the model training process, an attention mechanism (Attention Weight) is introduced to dynamically allocate the importance weight of features at different times, focusing on the key process nodes of the dehydration crisping zone (T2=85±1℃) and the cooling zone (T3=45±2℃). Implementation: An end-to-end LSTM prediction system was built on the Jetson AGX Orin edge computing platform, integrating the TensorRT inference acceleration engine (increasing inference throughput by 3 times). It receives real-time data from 18 sensors (6 for temperature, 3 for humidity, 1 for weight, and 8 for actuator status). After data preprocessing (outlier removal, smoothing filtering, and feature normalization), the data is input into the model, outputting parameter prediction curves for the next 2-5 minutes every 500ms. The prediction results are synchronized to the PLC controller via the PROFINET bus, triggering actuator actions in advance (such as pre-opening of the dehumidification valve and adjustment of the heat recovery bypass valve). This reduces system response lag from 2-3 seconds in traditional control to less than 0.5 seconds, lowering control deviation and ensuring stable crisping quality of pet food. Attached Figure Description

[0061] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0062] Figure 1 This is a schematic diagram of the hardware control system structure of an energy-saving oven provided in an embodiment of the present invention;

[0063] Figure 2 This is a flowchart of a brittle baking control process based on three-stage variable temperature hot air, provided by an embodiment of the present invention. Detailed Implementation

[0064] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0065] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0066] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning.

[0067] In this embodiment of the invention, sometimes a subscript such as W1 may be mistakenly written as a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0068] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0069] This invention provides a brittle baking control process based on three-stage variable temperature hot air. The process achieves precise control of the brittle structure of pet food through three-stage variable temperature hot air control, and integrates a closed-loop temperature control system of an energy-saving oven, which reduces energy consumption compared to traditional processes.

[0070] The following description of the invention will focus on the hardware and software upgrade architecture and process flow of the baking oven of this invention. The basic equipment design, structure, and control operation steps of the baking oven can be understood by referring to the operation steps of existing baking ovens.

[0071] The following is the technical solution of the present invention for upgrading the hardware system and software of the baking oven.

[0072] This invention proposes a three-stage variable-temperature hot air baking control process for pet food crispness. By precisely controlling the hot air flow field and material moisture migration, it achieves integrated control of crispness shaping and energy-saving baking for products such as pet chews and jerky. The system design scheme of the energy-saving oven for crispness baking of pet food according to this invention will be described in detail below.

[0073] I. Hardware System Details: Functionality, Connectivity, and Integration

[0074] The hardware architecture of this system is designed around a closed loop of "perception-decision-execution". Each part is tightly coupled with the central controller through an industrial bus. In view of the material characteristics of pet food that are high in protein and high in fiber, a dual-objective optimization system for brittleness formation and energy consumption control is constructed.

[0075] 1.1 Thermal and Circulation Unit

[0076] Core Function: Provides a precise, stable, and controllable hot air source, and regulates the evaporation rate of internal moisture in pet food through temperature gradient, laying a thermodynamic basis for the formation of brittle structures.

[0077] Composition and Connection:

[0078] Heating Array: Employs KANTHAL APM 1700 nickel-chromium alloy resistance wire, arranged in zones (2 preheating zones, 4 dehydration zones, and 0 cooling zones utilizing residual heat). Each zone is directly controlled by the PLC's digital output module via a SCHNEIDER TESYS F series AC contactor. Power adjustment is achieved by the PLC controller (e.g., SM1232) sending a 0-10V signal to the control terminal of the SCHNEIDER TESYS F contactor, enabling continuous power regulation from 0-100%. The power curve for the preheating zone is designed specifically for the collagen denaturation temperature (65-75℃) in pet food ingredients, preventing brittleness loss due to overheating.

[0079] Blowering System: The main blower, EBMPAPST D3E146-AH07-57, uses three-phase power supply. Its speed is controlled by a PLC connected to a SIEMENS SINAMICS G120 frequency converter via a PROFIBUS DP bus. The frequency converter receives the speed setpoint from the PLC (e.g., 0.5m / s, 1.2m / s, 0.8m / s) and feeds back the real-time speed to the PLC, forming a speed closed loop. Different blower speeds correspond to the different baking stages of pet food: 0.5m / s in the preheating stage promotes uniform formation of the surface protein film; 1.2m / s in the dehydration stage enhances the migration of internal moisture gradients; and 0.8m / s in the cooling stage prevents the brittle structure from cracking due to rapid cooling.

[0080] 1.2 Sensor Detection Network

[0081] Core functions: Real-time, multi-point acquisition of temperature, humidity, wind speed and weight data constitutes the system's "sensory perception". Temperature accuracy is controlled within ±1℃ to ensure stable starch gelatinization in pet food (target value 65-70%), and humidity monitoring resolution reaches 0.01%RH to regulate the Maillard reaction process.

[0082] Composition and Connection:

[0083] Temperature sensor network: Six OMEGA HHX13-K thermocouples are connected to the PLC's dedicated thermocouple analog input module (such as SM1231) via compensating wires. The module converts the millivolt signal into a digital temperature value.

[0084] Humidity sensor array: Three SENSIRION SHT85 sensors are connected to the NVIDIA Jetson AGX Orin edge computing gateway via the I2C communication protocol. Jetson preprocesses the data (such as filtering and unit conversion) before transmitting it to the main PLC via Ethernet.

[0085] System Integration: All sensor data is aggregated within the PLC and transmitted in real time to the NVIDIA Jetson AGX Orin via industrial Ethernet (such as PROFINET) for intelligent decision analysis.

[0086] 1.3 Implementing Body Group

[0087] Core function: Precisely executes the controller's instructions to change the internal environment of the oven.

[0088] Composition and Connection:

[0089] Damper actuator: BELIMO NFB24-SR electric damper, which receives a 0-10V signal from the PLC controller to control its opening degree (0-90°). Through the coordinated adjustment of damper opening and wind speed, a spiral hot air flow field is formed in the oven, ensuring that all parts of the pet food blank are heated evenly and avoiding local over-brittleness or collapse.

[0090] Dehumidification valve: SMC VX2120 solenoid valve, directly driven by the PLC's digital output module, enables rapid on / off action for humidity control. Targeting the high-protein nature of pet food, it precisely controls humidity at 30-35%RH during the dehydration stage. This range promotes muscle fiber contraction to form a brittle structure while preventing excessive protein denaturation that leads to excessive hardness.

[0091] Heat recovery unit: ALFA LAVAL CB26-20H plate heat exchanger, whose bypass valve is also controlled by a PLC controller to adjust the heat recovery efficiency.

[0092] 1.4 Central Control Platform

[0093] Core function: The "brain" of the system, responsible for data aggregation, algorithm execution, logic control, and interpersonal interaction.

[0094] Composition and Connection:

[0095] PLC (Master PLC): The SIEMENS SIMATIC S7-1500 serves as the slave device, acting as the hub for all input / output signals. It communicates with various I / O modules via the backplane bus and with devices such as frequency converters and HMIs via PROFIBUS / PROFINET.

[0096] Edge computing unit: NVIDIA Jetson AGX Orin connects to the PLC via Ethernet. The PLC sends real-time process data (such as T_avg, H_avg) to Jetson, which then calculates optimized setpoints (such as T_set_opt) based on the brittle formation quantization model and sends them back to the PLC for execution.

[0097] HMI: The SIEMENS TP2200 Smart Panel connects to the PLC via PROFINET for parameter setting, recipe recall, real-time monitoring, and alarm display.

[0098] II. Standardization and Interpretation of Algorithm Models

[0099] The core algorithm will now be transformed into a more standardized mathematical model, with detailed explanations of each parameter.

[0100] 2.1 Adaptive Fuzzy PID Control Law

[0101] The controller's output u(k) in the discrete-time domain is:

[0102] ,

[0103] The principle of the u(k) formula: This formula is a discrete-time domain PID control output model, which adjusts the system output through the coordinated action of three terms: proportional (Kp), integral (Ki), and derivative (Kd). The proportional term quickly responds to the current deviation, the integral term eliminates steady-state error, and the derivative term predicts the trend of deviation change.

[0104] Operating Mechanism: The PLC samples temperature and humidity data every 100ms, calculates e(k), and updates Kp(k), Ki(k), and Kd(k) in real time. For example, when the real-time temperature is detected to be 2℃ lower than the setpoint temperature, the proportional term immediately outputs an adjustment of Kp(k)⋅2, while the integral term begins to accumulate deviation, and the derivative term adjusts the output damping based on the deviation slope of the previous three samples. To address the brittleness requirements of pet food, a larger Kd value (0.3-0.5) is used during the dehydration stage to suppress the impact of temperature fluctuations on crystallinity, ensuring that the standard deviation of the product's fracture strength is ≤5%.

[0105] Energy-saving advantages: Dynamic coefficient adjustment links the power output of the heating array with the heat recovery unit, significantly reducing overshoot compared to traditional PID control. After rapid heating with a high Kp during the preheating stage, Kp automatically decreases to 0.8-1.0 during the dehydration stage. Combined with the high heat recovery efficiency of the ALFA LAVAL heat exchanger, total energy consumption is significantly reduced. Simultaneously, precise humidity control reduces the brittleness and deterioration of pet food caused by secondary drying, increasing product qualification rate to over 98%.

[0106] u(k): The output value of the controller at time k, which usually corresponds to the percentage of heating power or the control signal of the actuator.

[0107] e(k): The systematic deviation at time k. This deviation directly affects the power output of the heating array. A positive deviation increases the heating power to promote protein cross-linking in pet food, while a negative deviation triggers the heat recovery unit to improve efficiency (e.g., reducing the bypass valve opening of the ALFA LAVAL heat exchanger by 50%), achieving energy-saving operation while ensuring the brittle structure (fracture strength 3.5-5.0N).

[0108] EC(k): The rate of change of error at time k. Where e(k-1) is the system deviation at time k-1. In the energy-saving control logic, when |EC(k)| < 0.5℃ / s, the PLC will trigger the fan speed to be appropriately reduced by 10%-15% (e.g., from 1.2m / s to 1.0m / s), while maintaining the temperature control accuracy within ±1℃ to ensure uniform diffusion of moisture inside the pet food and maintain the ideal brittleness gradient. When EC(k) is negative (temperature overshoot), the PLC immediately activates the bypass valve of the heat recovery unit (opening degree increased by 30%) to recover excess heat to the preheating zone for energy recovery.

[0109] Kp(k), Ki(k), and Kd(k): These are the proportional, integral, and differential coefficients at time k, respectively, and they change dynamically according to the baking stage. For example, during the preheating stage, Kp takes a larger value (1.2-1.5) to rapidly increase the temperature and promote the formation of a protein film on the surface of the pet food. During the dehydration stage, it automatically drops to 0.8-1.0 to reduce energy consumption fluctuations and avoid excessive brittleness that could cause the product to break.

[0110] k: The index of the discrete-time series, representing the k-th sampling time. The adaptive rule for its parameters is defined by the following empirical function:

[0111] (1) Proportionality coefficient: ,in:

[0112] Kp0: The initial baseline value of the proportional coefficient, dynamically set according to the baking stage. It is set to 0.8-1.0 in the preheating stage for fast response, reduced to 0.5-0.7 in the dehydration stage to reduce power fluctuations, and further reduced to 0.3 in the cooling stage to match the efficiency of the heat recovery system.

[0113] α: The adaptive gain coefficient of the proportional term, which is linked to the humidity sensor data. When the SENSIRION SHT85 detects that the ambient humidity is >60%, α automatically decreases by a certain percentage to avoid excessive heating energy consumption in high humidity environments.

[0114] E(k): The absolute value of the temperature deviation, E(k) = |e(k)|. When E(k) < 0.5℃, the PLC triggers the "energy-saving mode", reducing the speed of the blower motor by 10% (e.g., from 1.2m / s to 1.08m / s) while maintaining temperature control accuracy.

[0115] (2) Integral coefficient: This formula avoids integral saturation through an exponential decay mechanism. When EC(k) > 1℃ / min, the integral coefficient rapidly decays to a lower level of the initial value, preventing overheating that could cause the surface of the pet food to burn while the inside remains insufficiently crisp.

[0116] Ki0: The initial baseline value of the integral coefficient, which is linked to the heat recovery efficiency. When the efficiency of the ALFA LAVAL heat exchanger is >80%, Ki0 automatically increases by 20% to enhance system stability and reduce energy loss caused by temperature fluctuations.

[0117] γ: Adaptive decay coefficient of the integral term, with a value of 1.2-1.5 during the dehydration stage (accelerating decay) and a value of 0.6-0.8 during the cooling stage (slowing down decay), to match the energy consumption characteristics of different stages.

[0118] (3) Differential coefficients: When the humidity gradient increases, the differential effect is enhanced, and the opening of the dehumidification valve is adjusted 0.5-1 second in advance to reduce local softening and collapse of pet food caused by humidity lag, ensuring that the uniformity of crisping reaches more than 90%.

[0119] Kd0: The initial reference value of the differential coefficient, which is appropriately reduced in the energy-saving optimization mode. By extending the differential adjustment cycle (from 0.1s to 0.15s), the frequency of actuator operation is reduced, thereby reducing mechanical energy consumption.

[0120] δ: The adaptive gain coefficient of the derivative term, which is tied to energy consumption priority. When the system is set to "extreme energy saving" mode, the value of δ is appropriately reduced to weaken the derivative effect and reduce power fluctuations.

[0121] Hgrad(k): Humidity gradient at time k. Hcenter(k) and Hsurface(k) are the center temperature and surface temperature of the food detected by the sensor, respectively. When the temperature gradient is large, heat recovery is triggered, and the exhaust gas is introduced into the ALFA LAVAL heat exchanger to preheat the fresh air, significantly improving thermal efficiency.

[0122] 2.2 Quantitative Model of Brittle Formation

[0123] The brittleness C of a product is a comprehensive indicator that can be modeled as follows:

[0124] Among them, k1 is optimized for the denaturation kinetics of pet food proteins (value 0.35), and k2 is associated with the collagen fiber rupture energy (value 0.42). By balancing the heating rate and moisture migration control, the brittleness index C of pet food is stabilized in the range of 65-75 N·mm, while the energy consumption per unit of brittleness is effectively reduced.

[0125] Formula C principle: This model comprehensively considers the effects of heating rate (ΔT / Δt), moisture gradient (ΔM / Δx) and humidity control accuracy (integral term) on brittleness, and balances the contributions of each factor through weighting coefficients k1, k2, and k3.

[0126] Its operating mechanism is as follows: Jetson calculates the pet food fragility index C value every 200ms. When it detects that the C growth rate is >0.05 units / s (indicating that rapid fragility may lead to breakage), it automatically reduces ΔT / Δt to 0.8 times the baseline value. At the same time, it predicts the humidity deviation integral term through LSTM and adjusts the opening of the dehumidification valve in advance. LSTM application principle: It adopts a 3-layer bidirectional LSTM network structure (128 neurons in the input layer + 64 neurons in the hidden layer + 1 neuron in the output layer). It selectively memorizes the long-term dependence of pet food humidity deviation through a gated recurrent unit (GRU), solving the gradient vanishing problem of traditional RNNs. The model input includes 18-dimensional features, including the humidity deviation sequence (Hset-Hactual) of the first 5 minutes, the temperature change rate (ΔT / Δt), and the heat recovery efficiency (ηrec), which are processed by Batch Normalization before being input into the network. Implementation: A TensorRT-optimized LSTM model is deployed on an NVIDIA Jetson AGX Orin. Real-time sensor data (SENSIRION SHT85 humidity value and OMEGA thermocouple temperature value) transmitted from the PLC is received every 500ms. A time-series feature matrix is ​​generated using a sliding window (window size = 30 sampling points) to predict the humidity deviation integral trend for the next 1-3 seconds. When the predicted integral term > a threshold (e.g., 5%·s), a correction value for the dehumidification valve opening (ΔKvalve = 0.15 × predicted integral term) is output in advance and transmitted to the PLC controller via Ethernet. This controls the pre-action time of the SMC VX2120 solenoid valve to prevent pet food from developing brittleness differences due to humidity fluctuations.

[0127] Energy-saving advantages: By optimizing the humidity integral term, the number of times the dehumidification valve operates is significantly reduced, and the energy consumption of the exhaust is effectively reduced; the moisture gradient control makes the energy consumption distribution in the dehydration stage uniform, reducing ineffective heating compared with traditional processes.

[0128] C: Crispness Index, dimensionless, the higher the value, the crisper the food. For pet food, the C value should be controlled within the range of 65-75 to ensure palatability (avoiding food that is too hard and can damage the mouth, or food that is too soft and cannot chew properly).

[0129] k1, k2, k3: Weighting coefficients related to the characteristics of pet food ingredients (e.g., k1=0.35 for chicken meal formulas, k1=0.42 for fish meal formulas). In energy-saving mode, k3 (humidity control accuracy weight) is appropriately increased by 20%-30% to reduce the brittleness of pet food caused by repeated heating due to humidity over-adjustment. Experimental data shows that this can effectively reduce total energy consumption.

[0130] ΔT / Δt: Average heating rate (°C / min) in the dehydration and embrittlement zone.

[0131] ΔM / Δx: The moisture content gradient (% / mm) inside pet food. The ideal range is 0.8-1.2% / mm. It is the key to forming a composite structure of crispy surface and soft interior. Insufficient gradient will result in the product being too soft overall.

[0132] The term dτ represents the integral of the squared error of humidity control accuracy throughout the baking process. Here, H_set(τ) is the set humidity (%RH) at time τ, H_actual(τ) is the actual humidity (%RH) at time τ, τ is the integration time variable (in seconds), t_total is the total baking time (in seconds), and dτ is the time derivative sign. This term quantifies the stability of humidity control through the squared cumulative error; a smaller value indicates less humidity fluctuation and higher uniformity of crispness in pet food.

[0133] 2.3 System-level energy-saving optimization objective function

[0134] This is a multi-objective optimization problem, mathematically expressed as minimizing the objective function J:

[0135]

[0136] Qquality includes the pet food crispness index C (weight 40%), moisture uniformity (30%), and color L value (30%). Energy consumption is prioritized and optimized by dynamically adjusting the weights (ω1=0.6 in energy-saving mode). Under typical operating conditions, Eactual / Etarget can be controlled within 0.9.

[0137] The principle of the min J formula: This multi-objective optimization model achieves synergistic optimization of energy saving, quality and efficiency by linearly weighting and integrating energy consumption ratio (Eactual / Etarget), quality deviation and process time ratio.

[0138] Operating mechanism: Jetson solves the optimization problem every 5 minutes and dynamically updates ω1-ω3. In energy-saving mode, ω1=0.6. When Eactual / Etarget>0.95, ω3 (process time weight) is automatically reduced, allowing for a suitable extension of baking time to prioritize energy consumption targets.

[0139] Energy-saving advantages: By dynamically adjusting energy consumption priority, while ensuring Qquality≥95%, Eactual / Etarget is stably controlled at a low level, significantly reducing total energy consumption compared to fixed parameter control, and significantly improving the utilization rate of heat recovery unit.

[0140] J: The overall objective function, which needs to be minimized.

[0141] Weighting coefficients, satisfying For example, it can be set to [0.5, 0.3, 0.2], and adjusted according to actual needs.

[0142] Eactual: Actual total energy consumption in production (kW·h).

[0143] Etarget: Target or benchmark energy consumption (kW·h).

[0144] Qquality: The overall product quality score, with core indicators for pet food including a brittleness index C≥65, moisture content 8±0.5%, and hardness 3.5-5.0N.

[0145] Qstandard: Product quality standard value, specified by the industry or enterprise.

[0146] tprocess: Total actual production process time (min).

[0147] tbenchmark: Baseline process time (min).

[0148] III. Hardware-Software Integrated Process Control Flow

[0149] The following process details how the hardware executes, how the software algorithm makes decisions, and how they work together across the three process stages.

[0150] (I) First Stage: Refined Process in the Preheating and Shaping Zone (T1=65±2℃, t1=8-10min)

[0151] 1. Hardware Initialization Sequence: The PLC sends a start signal to the SCHNEIDER TESYS F AC contactor via the digital output module (SM1223), activating two sets of KANTHAL APM 1700 nickel-chromium alloy heating wires (1.2kW each) in the preheating zone, with an initial power duty cycle of 45% (controlled by a 4.5V control signal output via the SM1232 analog module). This synchronously triggers the EBMPAPST D3E146-AH07-57 main fan. The SIEMENS SINAMICS G120 frequency converter receives the PROFIBUS DP command from the PLC (0-10V corresponds to 0-100% speed), stabilizing the fan speed at 0.5m / s (corresponding to a frequency converter frequency of 25Hz).

[0152] 2. Multi-dimensional sensing start-up: Six OMEGA HHX13-K thermocouples (distributed in the upper, middle and lower layers of the preheating zone) convert temperature signals (4-20mA) into digital quantities through the SM1231 module, with a sampling period of 100ms; three SENSIRION SHT85 humidity sensors transmit ambient humidity data (0-100%RH) to the NVIDIA Jetson AGX Orin via the I2C bus. After the data is processed by a 5th-order Butterworth filter, it is uploaded to the PLC in real time via PROFINET (update period of 200ms).

[0153] 3. Adaptive Control Closed Loop: - Temperature Regulation: The PLC executes an adaptive PID algorithm, and when the OMEGA thermocouple detects the temperature deviation... When the SM1232 module output voltage increases linearly (sensitivity 0.5V / ℃), the power of the KANTHAL heating wire is controlled via PWM, with a dynamic response time ≤500ms; - Wind speed coordination: Based on the initial weight of the material (W0) fed back by the METTLERTOLEDO weighing sensor, the LSTM model predicts the optimal wind speed correction value Δv (e.g., when W0 > 5kg). The fan speed can be adjusted in real time using a SINAMICS G120 frequency converter.

[0154] 4. Protein film formation determination: Jetson runs a lightweight CNN model (MobileNetV2 architecture) to extract features from the material surface images captured by the industrial camera. When the protein film coverage is ≥92% (surface reflectivity >65%) and the thickness reaches 0.15±0.02mm in 30 consecutive frames, a stage completion signal is sent to the PLC via Ethernet (TCP / IP protocol, data packet size 512 bytes).

[0155] 5. Energy Efficiency Optimization Action: The PLC monitors the cumulative energy consumption in this stage in real time through the energy consumption metering module (SM1238). At the same time, the electric heat tracing compensation of the double-layer glass fiber insulation layer (power 50W / m²) is automatically activated, and the fan speed is reduced to 0.45m / s to ensure that the preheating energy consumption ratio is stable at ≤15%.

[0156] (II) Second Stage: Coordinated Control Process of Dehydration and Embrittlement Zone (T2=85±1℃, t2=15-18min)

[0157] 1. Gradient Heating Execution: The PLC acquires OMEGA thermocouple temperature data via the SM1231 module, performs PID calculations, and then outputs a 0-10V signal to the KANTHAL heating wire power regulator via the SM1232 module, achieving a precise temperature rise of 1.0℃ / min (temperature overshoot ≤0.5℃). During the heating process, the BELIMO NFB24-SR electric damper is controlled by analog signals (0-10V corresponds to 0-90° opening) and linked with the SMC VX2120 solenoid valve (response time ≤50ms) to form an intermittent counter-current airflow field (cycle 30s, 15s air supply / 15s air stop).

[0158] 2. Dynamic Power Adjustment: The PLC outputs a 6-10V control signal (corresponding to 60%-100% power) via the SM1232 analog module to drive four sets of KANTHAL APM 1700 heating wires to heat up in stages: initially, the temperature rises from T1 to 80℃ at a rate of 1.0℃ / min for the first 5 minutes, then maintains a rate of 0.5℃ / min to 85℃ for the next 10 minutes. During the heating process, the SCHNEIDER TESYS F AC contactor is switched on and off every 200ms via the PLC digital module (SM1223) to ensure temperature fluctuations are ≤±0.5℃.

[0159] 3. Humidity gradient control:

[0160] Sensing layer: Three SENSIRION SHT85 humidity sensors (distributed in the left, center and right positions of the dehydration area) upload real-time data via I2C protocol (sampling frequency 1Hz). Jetson AGX Orin calculates the standard deviation σH of the humidity field distribution. When σH > 3%RH, it triggers a fine adjustment of the fan speed (Δv = 0.1m / s).

[0161] Execution layer: The SMC VX2120 solenoid valve receives the PWM control signal from the PLC (duty cycle 0-100%) and dynamically adjusts the opening based on the humidity overshoot predicted by the LSTM model (30 seconds in advance). The target humidity is stably controlled at 30±2%RH, and the valve response time is ≤50ms.

[0162] 4. Counter-current airflow execution: The PLC sends a reversing command to the BELIMO NFB24-SR electric damper via the PROFIBUS DP bus to execute an intermittent counter-current program: forward rotation for 30 seconds (damper opening 60°, air velocity 1.2 m / s) → stop airflow for 5 seconds (damper fully closed) → reverse rotation for 30 seconds (damper opening 45°, air velocity 1.0 m / s) → stop airflow for 5 seconds. This cycle is precisely controlled by the PLC's timer (TON) to ensure that the material surface boundary layer disruption efficiency reaches over 92%, and the moisture gradient ΔM / Δx remains stable at 1.0 ± 0.1% / mm.

[0163] 5. Real-time control of brittleness: Jetson calculates the brittleness index every 500ms. When C > 72, the following adjustment is triggered.

[0164] Reduce the power of the KANTHAL heating wire by 5% (SM1232 output voltage reduced by 0.5V);

[0165] Increase the target humidity to 32%RH (increase the SMC solenoid valve opening by 15%).

[0166] Extend the counter-current cycle to 75s (35s forward rotation + 5s air stop + 35s reverse rotation).

[0167] 6. Energy Efficiency Optimization Intervention: The PLC monitors real-time power through the SM1238 energy consumption module. When the unit energy consumption Eunit > 0.45 kW·h / kg, the heat recovery linkage is automatically activated: the opening of the ALFA LAVAL heat exchanger bypass valve is adjusted from 50% to 30%, and the fan speed is reduced to 1.1 m / s to ensure that the heat recovery efficiency ηrec is maintained at ≥85%. Experimental data shows that this mode can reduce energy consumption in the dehydration stage by 12%-15%.

[0168] (III) Third Stage: Closed-Loop Control Process in the Stable Cooling Zone (T3=45±2℃, t3=6-8min)

[0169] 1. Waste heat recovery start-up: The PLC cuts off the power supply to the KANTHAL heating wire (SM1223 outputs low level), and at the same time drives the main channel valve of the ALFA LAVAL CB26-20H plate heat exchanger to fully open (100% opening) and close the bypass valve (0% opening) through the digital module. The heat of the waste gas is recovered to preheat the fresh air (inlet temperature 45℃ → outlet temperature 65℃). The heat exchange efficiency is monitored in real time by the OMEGA thermocouple group (ΔT≤5℃).

[0170] 2. Wind speed gradient adjustment: The SIEMENS SINAMICS G120 frequency converter receives speed commands from the PLC and reduces the wind speed in three stages: initial 2 minutes 1.0 m / s (rapid cooling) → middle 3 minutes 0.8 m / s (structural curing) → final 3 minutes 0.6 m / s (moisture balancing). The frequency converter provides real-time speed feedback (accuracy ±1 rpm) via PROFIBUS DP, forming a closed-loop speed control.

[0171] 3. Multi-parameter endpoint determination: Humidity conditions: The SENSIRION SHT85 sensor monitors the humidity H_actual in the cooling area, and the fluctuation is ≤±2%RH for 3 consecutive minutes (judged by Jetson's 3σ criterion).

[0172] Moisture conditions: The METTLER TOLEDO weighing sensor monitors the material weight change rate dW / dt in real time. When |dW / dt| < 0.05 g / min and the estimated moisture content is 8 ± 0.5%, a compliance signal is triggered.

[0173] Brittleness conditions: Offline brittleness index C=68~72 (determined by the three-point bending method of INSTRON 5965 testing machine, n=10 samples, average value).

[0174] 4. Energy efficiency calculation and optimization: The PLC automatically performs the following operations:

[0175] Calculate the energy efficiency index for this batch:

[0176] ;

[0177] Compared with the historical best value, if Then the LSTM model weights are updated using gradient descent (learning rate α = 0.01).

[0178] Store key process parameters (temperature curves, humidity integral values, energy consumption data) in a local database for use in the next batch of feedforward control.

[0179] IV. Examples of Crispy Baking Process for Pet Food

[0180] (I) Specific Implementation Method: Three-Stage Baking Process for Chicken Jerky

[0181] Using chicken breast as the main ingredient (initial moisture content 70±2%), the three-stage variable temperature hot air process of this invention is used for crisp baking, with the specific parameters as follows:

[0182] Preheating and shaping zone: T1=70±1℃, t1=10min, wind speed 0.5m / s (SIEMENS SINAMICS G120 inverter set frequency 25Hz), KANTHAL APM 1700 heating wire power 35% (SM1232 output 3.5V), forming a surface protein film with a thickness of 0.1-0.2mm;

[0183] Dehydration and embrittlement zone: T2=85±1℃, t2=15min, heating rate 1.0℃ / min (PLC analog module SM1232 output voltage linearly increases), humidity control 30±2%RH (SMC VX2120 solenoid valve PWM duty cycle 30%-70%), wind speed 1.2m / s (BELIMO NFB24-SR damper opening 60°);

[0184] Stable cooling zone: T3=45±2℃, t3=8min, heat recovery efficiency 85% (ALFA LAVAL heat exchanger bypass valve opening 20%), wind speed 0.8m / s (inverter frequency 40Hz), final water content 8.2±0.3% (verified by METTLER TOLEDO weighing sensor).

[0185] The hardware and software collaborative control nodes are as follows:

[0186] Stage switching trigger: Jetson AGX Orin sends a stage completion signal (byte code 0x01) to the PLC via PROFINET. The trigger conditions are: protein film formation degree in the preheating zone ≥92% (CNN image recognition), brittleness index C in the dehydration zone ≥65 (real-time calculation), and center temperature in the cooling zone ≤40℃ (OMEGA thermocouple average).

[0187] Abnormal handling mechanism: When any sensor data exceeds the threshold (e.g., temperature deviation > 2℃ for 3 seconds), the PLC immediately starts the safety mode: cuts off the heating array (SM1223 outputs low level), reduces the fan speed to 0.3m / s, opens the emergency dehumidification valve, and at the same time, the HMI interface displays the fault code (e.g., E01 = temperature sensor fault).

[0188] Dehydration and embrittlement zone: T2=85±1℃, t2=15min, heating rate 1.0℃ / min (PLC analog module SM1232 outputs 0-10V linear signal to KANTHAL heating wire power regulator), humidity control 30±2%RH (SENSIRION SHT85 sensor feeds back data via I2C bus, Jetson AGX Orin calculates PID output to control the opening of SMC VX2120 solenoid valve), wind speed 1.2m / s (BELIMO NFB24-SR damper receives PLC reversing command via PROFIBUS DP, executes 30s forward rotation / 30s reverse rotation intermittent counterflow air supply);

[0189] Stable cooling zone: T3=45±2℃, t3=8min, heat recovery efficiency 85% (ALFA LAVAL CB26-20H plate heat exchanger bypass valve opening 20%, PLC monitors inlet and outlet temperature difference via SM1231 module), wind speed 0.8m / s (SIEMENSSINAMICS G120 frequency converter receives PLC speed setpoint and feeds back speed to closed-loop control in real time), final moisture content 8.2±0.3% (determined by data fusion of METTLER TOLEDO load cell and SENSIRION SHT85 humidity sensor).

[0190] (II) Experimental Scheme and Data for Processing Performance

[0191] Experimental Scheme 1: Effect of Heating Rate on Brittleness Index

[0192] Experimental principle: Based on collagen denaturation kinetics (sensitive range of 65-75℃), the effects of heating rates of 0.8 / 1.0 / 1.2℃ / min on muscle fiber contraction and brittle structure formation were investigated.

[0193] Experimental equipment: INSTRON 5965 universal testing machine (three-point bending fixture), OMEGA HHX13-K thermocouple (±0.5℃ accuracy), METTLER TOLEDO XS204 balance (0.1mg accuracy).

[0194] Experimental procedure: 30 samples were processed according to the process parameters in the example. The fracture strength was determined using the three-point bending method (loading rate 2 mm / min), and the brittleness index was calculated. .

[0195] Data Results:

[0196]

[0197] (Note: Data are mean ± standard deviation. ANOVA analysis showed that the 1.0℃ / min group was significantly different from other groups, P<0.05).

[0198] Experimental Scheme 2: The Influence of Humidity Control Accuracy on Embrittlement Uniformity

[0199] Experimental principle: By setting the control accuracy of SMC VX2120 solenoid valve to ±2% / ±5% / ±8%RH, the effects of humidity fluctuations on the Maillard reaction and moisture gradient (ΔM / Δx) were investigated.

[0200] Experimental equipment: SENSIRION SHT85 sensor (0.01%RH resolution), JETSON AGX ORIN edge computing platform (LSTM model prediction), TA.XT Plus texture analyzer (P / 36R probe).

[0201] Experimental procedure: 10 batches of production were continuously monitored, the cumulative humidity error integral value ∫(H_set-H_actual)²dt was calculated, and the standard deviation (SD) of the embrittlement index and the pass rate (C=65-75 range) were statistically analyzed.

[0202] Data Results:

[0203]

[0204] (Note: The unit of cumulative error integral is (%RH)²·s. The ±2%RH group significantly reduced brittle fluctuations, P<0.01).

[0205] Experimental Scheme 3: The Impact of Heat Recovery Efficiency on Energy Consumption

[0206] Experimental principle: Adjust the opening of the bypass valve of the ALFA LAVAL plate heat exchanger (20% / 50% / 80%) and measure the unit product energy consumption under different heat recovery efficiencies.

[0207] Experimental equipment: SIEMENS SIMATIC S7-1500 PLC (energy consumption metering module), EBMPAPST fan (power monitoring), KANTHAL heating wire (power consumption statistics).

[0208] Data Results:

[0209]

[0210] (Note: The energy saving rate is calculated based on the bypass valve opening of 80%, and the heat recovery efficiency is the highest when the valve is 20% open).

[0211] Experimental Scheme 4: Validation of the Accuracy of the LSTM Prediction Model

[0212] Experimental principle: The prediction time (t_water) of water content reaching the standard is compared with the actual measured value by Jetson AGX Orin to verify the prediction error of the model.

[0213] Experimental equipment: JETSON AGX ORIN (TensorRT accelerated LSTM model), METTLER TOLEDO weighing sensor (real-time moisture monitoring).

[0214] Data Results:

[0215]

[0216] (Note: The average relative error is 2.95%, and the model prediction accuracy meets the process requirements.)

[0217] (III) Experiments show that:

[0218] 1. Effect of heating rate on brittleness index

[0219] (1) Baking process: 1.0℃ / min is the optimal heating rate for precise control of brittle structure.

[0220] The experimental data table, obtained through brittleness index testing using an INSTRON 5965 universal testing machine (three-point bending fixture) and temperature field monitoring using an OMEGA HHX13-K thermocouple (±0.5℃ accuracy), reveals the influence of different heating rates on the brittle structure of pet food:

[0221] 1.2℃ / min (too rapid heating): The measured brittleness index was 65.7±2.4 (lower than the target range of 65-75), the muscle fiber cross-sectional structure was loose, and the integrity of the surface protein film was insufficient (compared with the protein film formed by the wind speed of 0.5m / s in the preheating and shaping zone, the damage rate increased by 5.5%).

[0222] Process Conclusion: The heating rate exceeded the optimal response speed within the collagen denaturation sensitive range (65-75℃), resulting in insufficient residence time of muscle fibers in this range and inadequate collagen denaturation, failing to form a stable brittle framework. Simultaneously, due to insufficient brittleness, the dehydration and embrittlement zone time needed to be extended (t3 increased by XX%), which easily led to excessive surface drying (humidity sensor monitoring showed local humidity below 25%) while internal moisture remained (final moisture content deviation >0.5%), exacerbating the difference in brittleness between the inside and outside of the product, failing to meet the "uniform embrittlement" process requirements.

[0223] 0.8℃ / min (too slow heating): The measured brittleness index was 68.3±2.1 (higher than the target range), the product hardness reached 4.2±0.3 N (three-point bending fracture force), which was 12.5% ​​higher than the target value, and 30% of the samples showed cracking (more than 3 cracks per sample).

[0224] Process Conclusion: An excessively slow heating rate caused muscle fibers to remain in the sensitive 65-75℃ range for too long, leading to excessive collagen denaturation and irreversible contraction. The resulting muscle fiber densification resulted in excessive brittleness. Simultaneously, prolonged heating caused deviations in temperature field uniformity (thermocouple monitoring showed a local temperature difference of 1.2℃, exceeding the ±1℃ temperature control accuracy), and some samples exhibited scorched edges (excessive Maillard reaction, with local humidity <20% as monitored by the humidity sensor), degrading palatability and appearance quality.

[0225] 1.0℃ / min (optimal heating rate): brittleness index stabilizes at 72.5±1.8 (within the target range), fracture force fluctuation ≤5%, uniform cross-sectional structure (number of cracks <1 / sample), and final moisture content 8%±0.3% (better than the process requirement of 8%±0.5%).

[0226] Process Conclusion: The muscle fibers remain in the sensitive range of 65-75℃ for an appropriate time (about 10 min), and the collagen is fully denatured without excessive shrinkage. This process is highly compatible with the "±1℃ gradient temperature control + 1.2m / s intermittent countercurrent airflow" process in the dehydration and embrittlement zone. Precise temperature control avoids insufficient or excessive denaturation, and the airflow promotes the gradient migration of moisture, resulting in a uniform brittle structure. This achieves the core process goal of making pet food "brittle but not crumbly".

[0227] 2. Energy saving: Achieving a balance between the dual objectives of "brittleness and energy consumption" at 1.0℃ / min, resulting in the lowest total energy consumption.

[0228] The experimental data table, combined with the energy consumption-moisture evaporation correlation data of the METTLER TOLEDO weighing sensor (0.1mg accuracy) and the power adjustment records of the PLC controller, clarifies the energy consumption characteristics at different heating rates:

[0229] 0.8℃ / min (too slow heating): The heating time (t3) in the dehydration and embrittlement zone is as long as 18.75 min (25% higher than the 1.0℃ / min group). The average power of the KANTHAL APM 1700 nickel-chromium alloy heating wire is 0.8 kW. The energy consumption in this stage accounts for 35% of the total energy consumption (exceeding the ≤30% upper limit of the process design). The total energy consumption is 10.5% higher than the 1.0℃ / min group.

[0230] Energy-saving conclusion: Slow heating leads to prolonged heating time. Although the double-layer glass fiber insulation layer reduces heat loss (preheating energy consumption ≤15%), continuous heating still causes a surge in total energy consumption, which does not meet the design goal of energy-saving ovens to "reduce energy consumption of traditional processes".

[0231] 1.2℃ / min (rapid heating): The heating time (t3) in the dehydration and embrittlement zone is shortened to 12.5 min (a 16.7% reduction compared to the 1.0℃ / min group), but the PLC controller needs to output a 7 V voltage signal (corresponding to a power of 1.2 kW, 20% higher than the 1.0℃ / min group), with an instantaneous power peak of 1.5 kW. The grid load fluctuation coefficient increases by 15%, and because the temperature control accuracy occasionally exceeds ±1℃ (thermocouple monitoring deviation of 1.1℃), secondary adjustment is required through the SCHNEIDER TESYS F contactor, increasing the rework energy consumption ratio by 5%.

[0232] Energy saving conclusion: Although rapid heating shortens the time, the high instantaneous power and rework energy consumption offset the time advantage, and the total energy consumption is 18.4% higher than that of the 1.0℃ / min group, thus failing to achieve the energy saving target.

[0233] 1.0℃ / min (optimal heating rate): The heating time (t3) in the dehydration and embrittlement zone is 15 min (within the process design range), with an average power of 1.0 kW (PLC output voltage 5 V, reasonable value in the 0-10V signal range). The energy consumption in this stage is controlled at 30% of the total energy consumption (meets design requirements). The total energy consumption is reduced by 10.5% compared to the 0.8℃ / min group and by 18.4% compared to the 1.2℃ / min group. The heat recovery efficiency (ALFA LAVAL plate heat exchanger) is stable at ≥85%.

[0234] Energy saving conclusion: By balancing heating time and power output, the 1.0℃ / min heating rate avoids the problems of "excessive time - continuous energy consumption" and "excessive power - instantaneous waste". In conjunction with the closed-loop temperature control system (PLC dynamic adjustment + frequency converter fan speed closed loop) of the energy-saving oven, it achieves the optimal balance between energy consumption and brittleness.

[0235] The experimental data table shows that 1.0℃ / min is the optimal heating rate that balances precise control of brittleness in the baking process with energy-saving requirements. In terms of process, it allows muscle fibers to denature sufficiently within the collagen denaturation-sensitive range (65-75℃) without over-denaturation, stabilizing the brittleness index within the target range (65-75), resulting in a uniform internal and external structure of the product. In terms of energy saving, by balancing heating time and power output, total energy consumption is reduced by 10.5%, adapting to the "brittleness setting-energy-saving baking integration" control target of the three-stage variable-temperature hot air process.

[0236] 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 invention, or the part that contributes to the prior art, or a part 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 invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0237] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention 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 the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A three-stage variable temperature hot air crisping baking energy-saving oven, characterized in that, include: (1) Thermal and circulation unit: includes a heating array arranged in zones (preheating zone and dehydration zone), a main fan and a frequency converter. The heating array is connected to the PLC digital output module through an AC contactor. The power adjustment is achieved by the PLC controller to realize 0-10V continuous control. (2) Sensor detection network: It consists of thermocouples, humidity sensors and weight detection units. The thermocouples are connected to the PLC through the PLC controller, and the humidity sensors are connected to the edge computing gateway via the I2C protocol. (3) Actuator group: including electric damper (0-90° opening), dehumidification solenoid valve and plate heat exchanger, all of which are controlled in a closed loop by PLC controller; (4) Central control platform: PLC is used as the lower-level machine, and it communicates with frequency converter, HMI and edge computing gateway through PROFIBUS / PROFINET bus.

2. The energy-saving oven according to claim 1, characterized in that, The interactive control methods of the PLC, edge computing gateway, and HMI include: (1) The PLC outputs a 0-10V signal to adjust the heating wire power through the PLC controller; (2) The edge computing gateway receives 18 sensor data every 500ms, calculates and optimizes the set value based on the brittle formation quantization model, and feeds it back to the PLC; (3) The HMI enables user parameter setting, recipe calling and real-time monitoring.

3. The energy-saving oven according to claim 1, characterized in that, The temperature sensor is connected to the thermocouple via a compensating wire, and the humidity sensor is connected to the edge computing gateway via an I2C bus. All data is transmitted via PROFINET industrial Ethernet with a communication rate of ≥100Mbps. The main fan and the frequency converter communicate via PROFIBUS DP bus, and the speed closed-loop control accuracy is ±0.05m / s; The actuator adopts a hybrid control of analog (0-10V) and digital (24V DC) signals, and the damper opening and wind speed are coordinated to form a spiral hot air flow field.

4. A brittle baking control process based on the energy-saving oven according to any one of claims 1-3, characterized in that, Includes the following steps: (1) Preheating and shaping stage: control the temperature at 65±2℃ and the time at 8-10min. The PLC executes the adaptive fuzzy PID control law through the PLC controller to adjust the power of the heating wire. The main fan maintains a wind speed of 0.5m / s to form a surface protein film of 0.15±0.02mm. (2) Dehydration and embrittlement stage: Heat to 85±1℃ at a rate of 1.0℃ / min, maintain for 15-18min, control humidity at 30±2%RH, execute embrittlement quantification model C, and when C is greater than the threshold, reduce heating power by 5% and increase target humidity to 32%RH; (3) Stable cooling stage: control the temperature at 45±2℃ for 6-8 min, reduce the wind speed in three stages, execute the system-level energy-saving optimization objective function, recover the heat of the exhaust gas through the plate heat exchanger, and determine that baking is complete when the humidity fluctuation is ≤±2%RH, the weight change rate |dW / dt|<0.05g / min and the brittleness index C=68-72.

5. The process according to claim 4, characterized in that, The expression for the adaptive fuzzy PID control law is: ,in: (1) u(k): The controller output value at time k (corresponding to the heating power percentage or actuator control signal); (2) e(k): System deviation at time k, T_set is the set temperature, and T_actual is the actual temperature measured by the thermocouple; (3) EC(k): the rate of change of error at time k. (℃ / sampling period), the sampling period is 200ms; (4) Kp(k), Ki(k), Kd(k): dynamic PID parameters, which are the proportional, integral, and derivative coefficients at time k, respectively; (5) k: Discrete time index, representing the kth sampling time, synchronized with the PLC scanning cycle.

6. The process according to claim 4, characterized in that, The expression for the brittleness formation quantification model is: ,in: (1) C: Brittleness index, target range 65-75, characterizing the fracture strength of pet food; (2) k1, k2, k3: weighting coefficients, satisfying , where k1 is the heating rate weight, k2 is the moisture gradient weight, and k3 is the humidity control accuracy weight; (3) ΔT / Δt: Heating rate during the dehydration and embrittlement stage (°C / min), preferably 1.0°C / min; (4) ΔM / Δx: Moisture gradient inside pet food (% / mm), calculated from Hcenter(k)-Hsurface(k), where Hcenter is the center humidity and Hsurface is the surface humidity; (5) : Humidity control accuracy integral term ((%RH)²·s), H_set is the set humidity, H_actual is the actual humidity measured by the sensor, and t_total is the total time of the dehydration stage (s). (6) τ: Integral time variable (s), representing the cumulative effect of humidity deviation over time.

7. The process according to claim 4, characterized in that, The expression for the system-level energy-saving optimization objective function is as follows: ， in: (1) J: Value of the energy-saving optimization objective function; (2) These are energy consumption weight, quality weight, and time weight, respectively, and satisfy the following conditions: ; (3) Eactual: Actual total energy consumption (kW·h), Etarget: Target energy consumption (kW·h); (4) Qquality = C_measured / C_target (quality score), Qstandard = 1.0 (standard score); (5) tprocess: actual process time (min), tbenchmark: benchmark process time (min).

8. The process according to claim 4, characterized in that, The steps in the preheating and shaping stage include: (1) Hardware initialization: The PLC activates the heating array in the preheating zone (including at least 2 sets of KANTHAL APM 1700 heating wires, each with a power of 1.2kW), with an initial power duty cycle of 45%, and the main fan maintains a wind speed of 0.5m / s through the frequency converter; (2) Multidimensional sensing: Thermocouples collect temperature data in 100ms cycles, and humidity sensors upload data to the PLC after preprocessing by the edge computing gateway; (3) Adaptive closed-loop control: When the temperature deviation e(k) > 1℃, the output voltage of the PLC controller increases linearly (sensitivity 0.5V / ℃), and the dynamic response time is ≤500ms; based on the initial weight W0 of the material, the LSTM model predicts the wind speed correction value Δv; (4) Protein film determination: When the surface protein film coverage is ≥92% and the thickness is 0.15±0.02mm, the transmission phase is completed. (5) Energy efficiency optimization: When the stage energy consumption is greater than 0.15 × total target energy consumption, activate the electric heat tracing compensation of the insulation layer (50W / m²) and reduce the wind speed to 0.45m / s.

9. The process according to claim 4, characterized in that, The steps of the dehydration and embrittlement stage include: (1) Gradient heating: The PLC controls the power of the heating wire through the PLC controller to raise the temperature from 65℃ to 85℃ at a rate of 1.0℃ / min, and adjusts it in two stages (1.0℃ / min for the first 5 minutes and 0.5℃ / min for the next 10 minutes), with temperature fluctuation ≤ ±0.5℃; (2) Humidity gradient control: The humidity sensor samples at a frequency of 1Hz. When the standard deviation of the humidity field σH > 3%RH, the wind speed is finely adjusted (Δv = 0.1m / s). The solenoid valve adjusts its opening 30 seconds in advance according to the LSTM prediction. The target humidity is 30±2%RH. (3) Counter-current air supply execution: The electric damper executes an intermittent program: forward rotation for 30s (opening degree 60°, wind speed 1.2m / s) → stop airflow for 5s → reverse rotation for 30s (opening degree 45°, wind speed 1.0m / s), and the cycle is controlled by the PLC timer.

Images