Online intelligent monitoring and automatic control method and system for pet food processing

Abstract

The application relates to the technical field of advanced process control, in particular to an online intelligent monitoring and automatic control method and system for pet food processing, which comprises the following steps: acquiring a current time window current sequence, a reactor inner temperature sequence and an outer wall temperature sequence; separating the current sequence into a current component and a friction component, and calculating the current variance of the friction component; outputting a viscosity value; calculating the differential slope of the viscosity value in three continuous time windows, if the differential slope is greater than the upper limit of the slope and maintains a preset anti-shake period, extracting the decay coefficient corresponding to the viscosity value; reducing the preset upper limit of heating according to the decay coefficient, and prolonging the integral constant of the preset PID algorithm; calculating the temperature difference value of the outer wall temperature sequence and the inner temperature sequence, inputting the temperature difference value into the PID algorithm after the integral constant is reconstructed, and generating a heating instruction constrained by the upper limit of heating. The application can solve the problem of heating out of control caused by viscosity measurement distortion.

Description

Technical Field

[0001] This application relates to the technical field of advanced process control, and in particular to online intelligent monitoring and automatic control methods and systems for the processing of pet palatability enhancers. Background Technology

[0002] The processing of pet palatability enhancers is often accompanied by intense Maillard reactions. Under high-temperature stirring within the reactor, the materials undergo a biochemical phase change, and their physical viscosity increases non-linearly with the reaction progress. This change in rheological properties directly thickens the thermal boundary layer between the material and the reactor wall, exacerbating the thermal resistance between the fluid inside the reactor and the external heat source. To maintain the reaction environment required for the process, industrial sites typically rely on sensor nodes to collect equipment operating load and temperature data, which are then input into the process controller to dynamically adjust the output power of the jacketed heating valves and stirring mechanisms, ensuring the underlying stability of large-scale fluid processing.

[0003] Patent application CN121279661A discloses an intelligent braising control method for casual braised chicken feet snacks based on multi-source sensing. This technical solution deploys multi-source sensors to collect real-time data on temperature, humidity, and gas concentration in the processing space, and simultaneously acquires the hardness, color, and pH value of the materials. These multi-dimensional indicators are then input into a pre-defined denaturation kinetics model and a Maillard reaction kinetics coupling model for independent calculations. The kinetic analysis results for each dimension are standardized to form a correlation matrix. The control system extracts the element deviations from this matrix and directly generates target values ​​for heating power, circulation rate, and ventilation volume, performing feedforward adjustment of the equipment's operating status.

[0004] When the above mechanism is applied to the processing of pet palatability enhancers, its control logic is limited by the delayed characterization of macroscopic biochemical indicators, lacking real-time closed-loop feedback on changes in fluid physical viscosity and thermal conductivity. When the material enters the Maillard reaction phase transition, the fluid viscosity increases dramatically, causing a sharp drop in thermal conductivity. At this point, because the existing system fails to isolate the high-frequency interference generated by mechanical friction from the total operating load of the stirring motor, the control loop cannot obtain accurate real fluid viscosity parameters. Simultaneously, limited by its inherent static temperature control architecture, the system cannot dynamically adjust the thermal inertia parameters within the control algorithm when faced with sudden changes in fluid thermal conductivity. Under conditions of large hysteresis in the reactor's heat transfer, the distorted state assessment combined with the static heating logic causes a disconnect between the issued heating command and the rapidly decreasing heat exchange capacity within the reactor. This easily leads to continuous loading of the heating actuator during the impedance change period, causing temperature overshoot and coking of materials adhering to the reactor walls. Summary of the Invention

[0005] To address the problem of uncontrolled heating caused by viscosity measurement distortion, this application provides an online intelligent monitoring and automatic control method and system for the processing of pet palatability enhancers.

[0006] Firstly, the online intelligent monitoring and automatic control method for pet palatability enhancer processing provided in this application adopts the following technical solution: The online intelligent monitoring and automatic control method for pet palatability enhancer processing includes:

[0007] Obtain the current sequence, the internal temperature sequence of the reactor, and the external wall temperature sequence within the current time window;

[0008] The current sequence is separated into low-frequency current components with frequencies below a preset lower limit and high-frequency current components with frequencies within a preset range by a preset filtering algorithm, and the current variance of the high-frequency current components is calculated.

[0009] When the current variance is less than a preset safety threshold, the average temperature of the internal temperature sequence is extracted;

[0010] The average temperature and the low-frequency current component are input into a preset mapping table, and the viscosity value is output.

[0011] Calculate the differential slope of the viscosity value under three consecutive time windows. If the differential slope is greater than the upper limit of the slope and maintains the preset anti-shake period, then extract the attenuation coefficient corresponding to the viscosity value.

[0012] The preset heating limit is reduced according to the attenuation coefficient, and the integral constant of the preset PID algorithm is extended.

[0013] The temperature difference between the outer wall temperature sequence and the inner temperature sequence is calculated, and the temperature difference is input into the PID algorithm after reconstructing the integral constant to generate a heating command constrained by the heating upper limit.

[0014] Optionally, obtaining the current sequence, the internal temperature sequence of the reactor, and the external wall temperature sequence within the current time window includes:

[0015] The current time window is updated based on a preset step size;

[0016] Within the current time window, the motor current is extracted to generate the current sequence, and the internal temperature data and external wall data of the reactor are collected respectively.

[0017] The current sequence, the internal temperature data, and the external wall data are simultaneously stored in a buffer pool;

[0018] Using the timestamp of the current sequence as a reference, the internal temperature data and the external wall data are interpolated in the buffer pool to extract time-aligned discrete temperature points, which are then aggregated to generate the internal temperature sequence and the external wall temperature sequence, respectively.

[0019] Optionally, the preset mapping table includes a two-dimensional grid, and the step of inputting the average temperature and the low-frequency current component into the preset mapping table and outputting the viscosity value includes:

[0020] The average temperature value is combined with the low-frequency current component to form a double bond value;

[0021] Address and locate the target interval enclosing the double bond value within the two-dimensional grid;

[0022] The double bond value is interpolated based on the boundary nodes of the target interval, the characteristic viscosity is extracted, and the characteristic viscosity is used as the viscosity value.

[0023] Optionally, the step of reducing the preset heating upper limit according to the attenuation coefficient and extending the integral constant of the preset PID algorithm includes:

[0024] The target upper limit is generated by multiplying the current preset heating upper limit by the attenuation coefficient, and the target upper limit is set as the maximum duty cycle;

[0025] The target constant is generated by dividing the current integral constant of the preset PID algorithm by the attenuation coefficient, and the current integral constant is replaced with the target constant.

[0026] The attenuation coefficient is a real number that is greater than zero and less than one.

[0027] Optional, also includes:

[0028] When the current variance is greater than or equal to the preset safety threshold and the duration reaches the preset time threshold, an over-limit state is triggered.

[0029] In the over-limit state, the interpolation calculation of the double key value is suspended;

[0030] The average viscosity is calculated by extracting the viscosity values ​​within the time window prior to triggering the over-limit state.

[0031] The viscosity value during the over-limit state is locked as the average viscosity.

[0032] Optional, also includes:

[0033] In the over-limit state, the temperature difference value is compared with a preset limiting threshold.

[0034] When the temperature difference value is greater than the limiting threshold, the temperature difference value input to the preset PID algorithm is clamped to the limiting threshold.

[0035] Optionally, after generating the heating command subject to the said heating upper limit constraint, the method further includes:

[0036] When the viscosity value is greater than or equal to the preset wall-binding threshold, the difference between the inner temperature sequence and the outer wall temperature sequence is calculated to extract the inner temperature slope and the outer wall slope.

[0037] When the internal temperature slope is less than zero and continues for a preset determination period, and the external wall slope is greater than a preset abrupt change threshold, thermal inversion is determined to have occurred and a forced intervention flag is generated.

[0038] Optionally, the reactor includes a heating mechanism, a hinged cantilever, and a locking assembly connecting the reactor and the hinged cantilever. The hinged cantilever has a stirring motor connected to a stirring paddle extending into the reactor. After generating the forced intervention flag, the following steps are included:

[0039] Output a heating command matching a preset cutoff threshold to the heating mechanism;

[0040] A closing command is issued to the locking component;

[0041] Based on the closing command, the locking component is driven to lock, and the articulated cantilever is converted from a clearance state to a rigid state.

[0042] Optionally, after converting the articulated cantilever from a clearance state to a rigid state, the following steps are included:

[0043] Collect the confirmation signal fed back after the locking component is engaged;

[0044] A cell-wall breaking command is sent to the stirring motor only when the confirmation signal is received;

[0045] Based on the cell-wall breaking command, the stirring motor is driven to operate at a rate greater than the nominal threshold with a preset pulse width;

[0046] When the preset pulse width is reached, a speed reduction command is sent to the stirring motor, and an unlocking command is sent to the locking component simultaneously;

[0047] The unlocking command drives the locking component to release the lock, thereby restoring the articulated cantilever to the clearance state.

[0048] Secondly, the online intelligent monitoring and automatic control system for pet palatability enhancer processing provided in this application adopts the following technical solution: The online intelligent monitoring and automatic control system for pet palatability enhancer processing includes:

[0049] The data acquisition module is used to acquire the current sequence, the internal temperature sequence of the reactor, and the external wall temperature sequence within the current time window;

[0050] The feature separation module is used to separate the current sequence into low-frequency current components with frequencies below a preset lower limit and high-frequency current components with frequencies within a preset range using a preset filtering algorithm, and to calculate the current variance of the high-frequency current components.

[0051] The viscosity mapping module is used to extract the average temperature of the internal temperature sequence when the current variance is less than a preset safety threshold; input the average temperature and the low-frequency current component into a preset mapping table, and output the viscosity value.

[0052] The parameter reconstruction module is used to calculate the differential slope of the viscosity value under three consecutive time windows. If the differential slope is greater than the upper limit of the slope and maintains a preset anti-shake period, the attenuation coefficient corresponding to the viscosity value is extracted. The preset heating upper limit is reduced according to the attenuation coefficient, and the integral constant of the preset PID algorithm is extended.

[0053] The instruction generation module is used to calculate the temperature difference between the outer wall temperature sequence and the inner temperature sequence, input the temperature difference into the PID algorithm after reconstructing the integral constant, and generate a heating instruction subject to the heating upper limit constraint.

[0054] In summary, this application includes the following beneficial technical effects:

[0055] 1. Based on the frequency domain separation processing of the current sequence of the stirring motor, the health status of the mechanical transmission is judged by extracting the current variance of the friction component. Only when the variance is lower than the safety threshold, the viscosity value is output by looking up the table using the temperature mean and the current component, which effectively eliminates the interference of mechanical friction noise on viscosity measurement. Then, based on the maintenance conditions of the viscosity difference slope and the anti-shaking cycle under the continuous time window, the heating upper limit is dynamically reduced and the PID integral constant is extended, thereby avoiding the decoupling of the heating command and the heat exchange capacity of the material, and suppressing temperature overshoot and wall coking.

[0056] 2. When the current variance continuously exceeds the safety threshold and triggers the over-limit state, the viscosity lookup table interpolation is suspended and the output is locked to the average viscosity before the over-limit. At the same time, the temperature difference value input to the PID algorithm is limited. This mechanism prevents heating malfunctions caused by viscosity mapping distortion or excessive temperature difference feedforward during mechanical abnormalities, and improves the fault-tolerant operation capability of the process control system under abnormal sensing conditions.

[0057] 3. When the internal temperature sequence slope is negative and the external wall temperature sequence slope exceeds the abrupt change threshold, the heating output is forcibly cut off and the hinged cantilever is locked into a rigid state. Only after obtaining the locking confirmation signal is a wall-breaking pulse of higher than the rated rate is sent to the stirring motor to destroy the attached heat insulation coking layer by using fluid shear force, thus realizing adaptive physical intervention for the heat conduction failure state. Attached Figure Description

[0058] Figure 1 A flowchart of an online intelligent monitoring and automatic control method for processing pet palatability enhancers provided in an embodiment of the present invention. Detailed Implementation

[0059] The following combination Figure 1This application will be described in further detail.

[0060] This application discloses an online intelligent monitoring and automatic control method for the processing of pet palatability enhancers, which operates on a physical architecture consisting of reaction actuators, online monitoring sensor nodes, and edge process controllers.

[0061] The reaction actuator includes a Maillard reactor, a jacketed proportional heating valve wrapped around the outer wall of the Maillard reactor to control the heat source input, and a variable frequency stirring motor movably mounted on the free end of the articulated cantilever. The output shaft of the variable frequency stirring motor is connected downward to a stirring blade extending into the bottom of the Maillard reactor cavity. The bottom of the articulated cantilever is movably connected to the top of the Maillard reactor via a base. A locking component is installed at the joint between the articulated cantilever and the top of the Maillard reactor. When the locking component is not closed, the articulated cantilever is kept in a free state; when closed, the articulated cantilever is moved into a rigid fixed state.

[0062] The online monitoring sensing node includes a high-frequency current sampling module connected in series with the stator end of the variable frequency stirring motor, a jacketed wall temperature sensor attached to the outer wall of the vessel, and an in-vessel penetrating temperature sensor inserted into the fluid inside the vessel.

[0063] The edge process controller uses an industrial-grade programmable logic controller, which internally incorporates a timestamp-based data alignment buffer, a two-dimensional viscosity discrete mapping table, and a PID calculation module with a variable integral constant.

[0064] Combination Figure 1 As shown in the flowchart, the edge process controller cycles through the current time window with a window length of 500ms and a step size of 100ms, and returns to the data acquisition node after each heating command is generated. Within the current time window, the high-frequency current sampling module acquires the stator current generation current sequence of the variable frequency stirring motor at a sampling rate of not less than 20kHz, and the in-vessel penetrating temperature sensor and the jacket wall temperature sensor acquire the fluid temperature inside the vessel and the jacket wall temperature, respectively.

[0065] The current sequence, along with two temperature data streams, is synchronously written to the data alignment buffer according to their respective timestamps. The process controller uses the current sequence timestamp as a reference to linearly interpolate the two temperature data streams in the buffer, extracting discrete temperature points that are aligned point-by-point with the current timestamp, and aggregating them to generate the inner temperature sequence and the outer wall temperature sequence. The aligned current sequence enters the digital bandpass filter logic. The edge process controller calls a set of fourth-order Butterworth filters. Before being sent to the low-pass branch, the aligned current sequence is first subjected to effective value (RMS) calculation or envelope detection. Then, the extracted amplitude envelope signal is input to the low-pass branch for smoothing. The -3dB cutoff frequency of the low-pass branch is set to 0.5Hz to separate the low-frequency current components with a cutoff frequency below 0.5Hz. The passband of the bandpass branch is set to 5kHz to 8kHz to separate the high-frequency current components in this frequency band, and the current variance of the high-frequency current components is calculated to generate the high-frequency current characteristic variance.

[0066] The Butterworth filter order can be selected between second and sixth order. Orders below second order result in an excessively wide transition band and difficulty in cleanly cutting high and low frequency bands. Orders above sixth order cause the group delay to accumulate with the order, compressing the real-time margin of the 100ms step size. Fourth order achieves a balance between transition band steepness and phase delay. As a preferred implementation, the bandpass branch is replaced by a sliding discrete Fourier transform. Energy is calculated on the 5kHz to 8kHz spectrum and then converted into current variance, effectively obtaining the mechanical friction characteristics of this frequency band. Here, the 0.5Hz low-pass cutoff frequency is set based on the fundamental frequency of the agitator rotation speed. The separated low-frequency current component is used to characterize the steady-state physical load of the fluid inside the vessel on the agitator. The 5kHz to 8kHz frequency band extraction is based on the mechanical resonance frequency band excited when the motor bearing and main shaft are jammed or experiencing early wear. The separated high-frequency current component is used to calculate the current variance and generate the high-frequency current characteristic variance. By separating the frequency domain, the characteristic variance of the high-frequency current can be used as a criterion for the health status of the mechanical transmission. This can prevent the resistance caused by mechanical jamming from being mistaken for a sudden increase in fluid viscosity, thus providing an objective input source for subsequent viscosity mapping.

[0067] The low-frequency current component changes slowly with the fluid impedance, characterizing the actual load of the fluid on the agitator inside the vessel; the high-frequency current characteristic variance increases with mechanical deterioration such as bearing wear and spindle jamming, characterizing the health status of the mechanical transmission. The edge process controller compares the high-frequency current characteristic variance with a preset mechanical noise safety threshold. If the variance is strictly less than the safety threshold, the mechanical transmission is determined to be in a healthy state, and viscosity calculation is allowed. The preset mechanical noise safety threshold is calculated based on the background noise variance statistically analyzed under the motor's no-load operating condition, and is taken as 1.5 to 2.0 times the no-load reference value.

[0068] The edge process controller extracts the average temperature of the internal temperature sequence within the current time window, combines the average temperature with the low-frequency current component to form a double-key value, and inputs it into a two-dimensional viscosity discrete mapping table. The two-dimensional viscosity discrete mapping table constructs a grid with temperature as the row index and low-frequency current as the column index. Each node stores the calibrated characteristic viscosity. The node data is discretized and filled after fitting the temperature-viscosity compensation relationship according to the Arrhenius equation. The temperature range covers 80℃ to 140℃ with a step size of 5℃, and the low-frequency current range is divided according to the rated parameters of the motor.

[0069] The edge process controller addresses the target interval of the double bond values ​​within the grid, performs bilinear interpolation on the double bond values ​​using the four boundary nodes of the target interval, and extracts the characteristic viscosity as the true biochemical viscosity value. Compared with a single exponential fitting formula, table lookup interpolation avoids the fitting residual caused by the strong nonlinearity of viscosity-temperature during the phase transition period, and outputs a true biochemical viscosity value that is free from mechanical noise and has rheological temperature compensation.

[0070] The edge process controller calculates the differential slope of the actual biochemical viscosity values ​​over three consecutive time windows. If the discrete differential slope exceeds a preset upper limit for abrupt change, and the internal anti-jitter counter confirms that the breakthrough state is maintained for more than three sliding steps, thermal impedance decoupling is triggered. The edge process controller reads the viscosity range of the actual biochemical viscosity value and obtains the corresponding attenuation coefficient, which is a real number between 0.1 and 0.9.

[0071] The edge process controller multiplies the current upper limit of the jacketed proportional heating valve by the attenuation coefficient to obtain the reduced upper limit of the opening and sets it as the maximum duty cycle. It then divides the current integral time constant of the PID calculation module by the attenuation coefficient to obtain the extended integral time constant and replaces the original value. When the attenuation coefficient approaches 0.1, the upper limit of the opening shrinks drastically and the integral time lengthens significantly, adapting to the abrupt phase transition caused by a sudden increase in thermal resistance; when it approaches 0.9, the adjustment is gentle, adapting to gradually changing operating conditions. The anti-vibration counter uses three sliding steps as a threshold to filter out spikes in the viscosity value caused by single-window noise, preventing frequent vibrations in the upper limit of the opening and the integral time constant due to instantaneous spikes. Furthermore, the correspondence between the attenuation coefficient and the viscosity range is replaced by online calculation using a piecewise linear function instead of table lookup, adapting to transitional viscosity ranges not covered by the mapping table.

[0072] The edge process controller calculates the temperature difference between the jacket wall temperature sensor reading and the in-vessel penetration temperature sensor reading in real time, and inputs the temperature difference value into the PID calculation module to complete parameter reconstruction. The PID calculation module adopts an incremental discrete structure. Its core closed-loop error input is the deviation between the actual temperature of the fluid in the vessel and the preset process target temperature. The proportional term is calculated with this deviation as the core error input. At the same time, the temperature difference value is added as a superimposed input to the PID loop to participate in the composite calculation as a feedforward compensation term used to suppress large hysteresis overshoot of heat conduction. The integral term is accumulated under the extended integral time constant. The calculation result is limited to the maximum duty cycle after reduction, and a pulse width modulation (PWM) duty cycle command is generated.

[0073] The duty cycle command is sent to the jacketed proportional heating valve via the industrial bus, driving the jacketed proportional heating valve to inject heat source at a limited opening. The edge process controller interacts with the frequency converter and the jacketed proportional heating valve via Profinet or EtherCAT industrial bus, with a communication cycle of no more than 4ms, ensuring that the issuance delay of the duty cycle command and the overload pulse command matches the real-time requirements of the control loop. The system mechanism described above, which strips the variance of high-frequency current characteristics and extracts the attenuation coefficient based on the slope of the actual biochemical viscosity value, achieves adaptive coordination with the variable viscosity characteristics of the fluid. As viscosity increases and heat transfer resistance increases, the heating upper limit is reduced synchronously by the attenuation coefficient, and the integral constant is extended. This suppresses excessive heat injection caused by thermal inertia from the control source, overcoming the temperature control overshoot limitation of conventional large-hysteresis nonlinear systems during the biochemical phase transition period.

[0074] If the variance of the high-frequency current characteristic is greater than or equal to the mechanical noise safety threshold, and the over-limit state lasts for more than 2.0 seconds, the edge process controller determines that the mechanical transmission is jammed or worn prematurely. It then suspends the lookup interpolation of the two-dimensional viscosity discrete mapping table and locks the current output real biochemical viscosity value as the sliding average of the last five time windows before the over-limit.

[0075] The edge process controller synchronously puts the PID calculation module into a restricted follow-up heating mode, clamping the temperature difference value input to the PID calculation module within a 15℃ safety limit range. This serves as a limit protection for the feedforward compensation term. Specifically, the temperature difference value is compared with a preset limit threshold. When the temperature difference value exceeds the limit threshold, the temperature difference value input to the PID calculation module is clamped to the limit threshold; that is, when the temperature difference value exceeds 15℃, it is always calculated as 15℃. The aforementioned 15℃ limit threshold is derived and set based on the heat transfer limit of the stainless steel jacket and the critical temperature difference for material coking, limiting the blind loading of the heat source due to large temperature difference feedforward during periods of mechanical anomalies. The 2.0-second continuous judgment threshold corresponds to the cumulative observation of dozens of sliding windows. High-frequency variance spikes shorter than the threshold mostly originate from the instantaneous impact of material clumps on the agitator without triggering the limit exceedance. After the limit exceedance state is cleared, the edge process controller resumes lookup table interpolation and normal temperature difference input.

[0076] During the closed-loop execution, if the actual biochemical viscosity value exceeds the preset high viscosity danger threshold, and the edge process controller extracts the physical-thermodynamic inversion characteristic, it will intervene in the underlying hard-sequence self-rescue. The physical-thermodynamic inversion characteristic is defined as a downward trend in the reading of the in-vessel penetrating temperature sensor, while the upward discrete slope of the jacket wall temperature sensor reading exceeds the preset extreme value. Its physical essence is that the thickening of the attached coking layer blocks heat conduction, and the heat in the jacket accumulates on the wall and cannot penetrate to the fluid inside the vessel.

[0077] After recognizing this feature, the edge process controller forcibly outputs a 0% duty cycle, instantly cuts off the drive source of the jacket proportional heating valve to perform hardware-level intervention control, sends a closing command to the locking component, drives the locking component to lock, and constrains the hinged cantilever and the base, which are in a free state, into a rigid fixed state.

[0078] The edge process controller acquires the positioning confirmation signal fed back after the locking component engages. Only upon acquiring the positioning confirmation signal does it send an overload breaking pulse command to the variable frequency stirring motor, driving the motor at 120% of its nominal speed to perform constant maximum torque shearing for 3 seconds, using the fluid's ultimate shear force to destroy the attached heat-insulating coking layer. After 3 seconds, the edge process controller sends a deceleration command to the variable frequency stirring motor to restore it to the reference speed, and simultaneously sends an unlocking command to the locking component, driving it to release the lock and allowing the articulated cantilever to return to its clearance state. The positioning confirmation signal is a necessary trigger condition for the overload breaking pulse. Since 120% overload breaking generates a huge reaction torque in high-viscosity fluids, if executed in the clearance state, it is very easy to cause severe radial runout of the cantilever and trigger metal fatigue tearing. After rigid locking, the lateral assembly gap is effectively eliminated, allowing the reverse impact load generated by breaking to be safely and evenly distributed and transferred to the vessel base. This mechanism achieves a safe linkage between fluid breaking and structural damage prevention without changing the basic mixing structure of the reactor.

[0079] The locking assembly is an actuation mechanism that switches between a clearance state and a rigid state in response to control commands. In the clearance state, where no closing command is received, the actuating end of the locking assembly is in the released state, and the hinged cantilever retains lateral assembly clearance at the junction with the reactor. The hinged cantilever oscillates slightly with the stirring reaction force. When a closing command is received from the edge process controller and the assembly enters the rigid state, the locking assembly actuates to forcibly eliminate the lateral assembly clearance, thus constraining the hinged cantilever to a rigid fixed state.

[0080] The locking component can be an electromagnetic actuation element or equivalently replaced by a hydraulic locking mechanism, with the hydraulic cylinder clamping the hinged cantilever to achieve the same backlash elimination. In addition to using a current transformer connected in series with the stator, the high-frequency current sampling module obtains the stator current equivalently from the DC bus current inside the inverter through a phase current reconstruction algorithm, eliminating the need for an external current transformer.

[0081] Existing technologies directly map viscosity to the total motor current. High-frequency current distortion caused by mechanical jamming during the phase transition is mistakenly interpreted as viscosity increase. Combined with the thermal inertia of the static temperature control architecture, this leads to frequent full-load heating valves, temperature overshoot reaching +15℃, and coking of materials adhering to the wall. The edge process controller hard-strips the total current into high-frequency current characteristic variance and low-frequency current components within a 500ms sliding window. Only when the high-frequency variance does not exceed the limit is the true biochemical viscosity value output via double-bond value mapping, effectively reducing the mistransmission of mechanical artifacts to the temperature control loop. Based on the discrete differential slope combined with anti-jitter counting triggering thermal impedance decoupling, the attenuation coefficient is used to inversely adjust the upper limit of valve opening and the PID integral time constant, objectively avoiding the blind accumulation of thermal inertia integral during the phase transition, and reducing the temperature following error to ±1℃. After identifying the physical thermodynamic inversion characteristics, the edge process controller locks the execution premise of overload wall breaking with a positioning confirmation signal, ensuring that the articulated cantilever is not radially torn while removing the coking layer adhering to the wall.

[0082] This application also discloses an online intelligent monitoring and automatic control system for the processing of pet palatability enhancers, which includes a data acquisition module, a feature separation module, a viscosity mapping module, a parameter reconstruction module, and an instruction generation module, with each module deployed on an edge process controller.

[0083] The data acquisition module is communicatively coupled to the high-frequency current sampling module, the in-vessel penetrating temperature sensor, and the jacket wall temperature sensor. It is configured to acquire the current sequence, the internal temperature sequence, and the external wall temperature sequence within the current time window, and to interpolate and align the two temperature data streams in the buffer pool based on the current sequence timestamp.

[0084] The input of the feature separation module is coupled to the output of the data acquisition module. It is configured to separate the current sequence into a current component with a frequency below a preset lower limit and a friction component with a frequency in a preset range through digital bandpass filtering, and to calculate the current variance of the friction component.

[0085] The input of the viscosity mapping module is coupled to the feature separation module. It internally stores a two-dimensional viscosity discrete mapping table and is configured to extract the temperature mean of the internal temperature sequence when the current variance is less than a preset safety threshold. The temperature mean and the current component are used as double bond values ​​and the viscosity value is output by lookup table interpolation.

[0086] The input of the parameter reconstruction module is coupled to the viscosity mapping module and is configured to calculate the differential slope of the viscosity values ​​under three consecutive time windows. When the differential slope is greater than the upper limit of the slope and the preset anti-shake period is maintained, the attenuation coefficient corresponding to the viscosity value is extracted, and the preset heating upper limit is reduced and the integral constant of the preset PID algorithm is extended according to the attenuation coefficient.

[0087] The input of the instruction generation module is coupled to the parameter reconstruction module, and the output is coupled to the jacketed proportional heating valve via an industrial bus. It is configured to calculate the temperature difference between the outer wall temperature sequence and the inner temperature sequence. The temperature difference is input into the PID algorithm after reconstructing the integral constant. The core closed-loop error input of the PID algorithm is the deviation between the actual temperature of the fluid in the vessel and the preset process target temperature. The temperature difference is used as a feedforward compensation term superimposed into the PID loop to suppress the large hysteresis overshoot of heat conduction and participate in the calculation to generate a heating instruction constrained by the upper limit of heating.

[0088] The system also includes a locking component and a variable frequency stirring motor coupled to the instruction generation module. The locking component locks into the joint between the hinged cantilever and the top of the vessel, and is configured to constrain the hinged cantilever from a free-floating state to a rigidly fixed state when closed. The variable frequency stirring motor is movably mounted on the hinged cantilever and is connected to a stirring paddle that extends into the reactor. It is configured to perform a shearing motion with a preset pulse width at a rate exceeding the rated value after the locking component provides a confirmation signal of being in place.

[0089] Each module is physically represented by a processor and memory. The memory stores instructions that can be executed by the processor, and the processor executes the instructions to implement the functions of each module. The micro-execution details of frequency domain stripping, lookup table interpolation, differential triggering, and timing self-rescue have been described in detail in the aforementioned method embodiments and will not be repeated here.

[0090] The above are all preferred embodiments of this application, and are not intended to limit the scope of protection of this application. Therefore, all equivalent changes made in accordance with the structure, shape and principle of this application should be covered within the scope of protection of this application.

Claims

1. An online intelligent monitoring and automatic control method for the processing of pet palatability enhancers, characterized in that, include: Obtain the current sequence, the internal temperature sequence of the reactor, and the external wall temperature sequence within the current time window; The current sequence is separated into low-frequency current components with frequencies below a preset lower limit and high-frequency current components with frequencies within a preset range by a preset filtering algorithm, and the current variance of the high-frequency current components is calculated. When the current variance is less than a preset safety threshold, the average temperature of the internal temperature sequence is extracted; The average temperature and the low-frequency current component are input into a preset mapping table, and the viscosity value is output. Calculate the differential slope of the viscosity value under three consecutive time windows. If the differential slope is greater than the upper limit of the slope and maintains the preset anti-shake period, then extract the attenuation coefficient corresponding to the viscosity value. The preset heating limit is reduced according to the attenuation coefficient, and the integral constant of the preset PID algorithm is extended. The temperature difference between the outer wall temperature sequence and the inner temperature sequence is calculated, and the temperature difference is input into the PID algorithm after reconstructing the integral constant to generate a heating command constrained by the heating upper limit.

2. The online intelligent monitoring and automatic control method for processing pet palatability enhancers according to claim 1, characterized in that, The acquisition of the current sequence, the internal temperature sequence of the reactor, and the external wall temperature sequence within the current time window includes: The current time window is updated based on a preset step size; Within the current time window, the motor current is extracted to generate the current sequence, and the internal temperature data and external wall data of the reactor are collected respectively. The current sequence, the internal temperature data, and the external wall data are simultaneously stored in a buffer pool; Using the timestamp of the current sequence as a reference, the internal temperature data and the external wall data are interpolated in the buffer pool to extract time-aligned discrete temperature points, which are then aggregated to generate the internal temperature sequence and the external wall temperature sequence, respectively.

3. The online intelligent monitoring and automatic control method for processing pet palatability enhancers according to claim 2, characterized in that, The preset mapping table contains a two-dimensional grid. The step of inputting the average temperature and the low-frequency current component into the preset mapping table and outputting the viscosity value includes: The average temperature value is combined with the low-frequency current component to form a double bond value; Address and locate the target interval enclosing the double bond value within the two-dimensional grid; The double bond value is interpolated based on the boundary nodes of the target interval, the characteristic viscosity is extracted, and the characteristic viscosity is used as the viscosity value.

4. The online intelligent monitoring and automatic control method for processing pet palatability enhancers according to claim 1, characterized in that, The step of reducing the preset heating upper limit according to the attenuation coefficient and extending the integral constant of the preset PID algorithm includes: The target upper limit is generated by multiplying the current preset heating upper limit by the attenuation coefficient, and the target upper limit is set as the maximum duty cycle; The target constant is generated by dividing the current integral constant of the preset PID algorithm by the attenuation coefficient, and the current integral constant is replaced with the target constant. The attenuation coefficient is a real number that is greater than zero and less than one.

5. The online intelligent monitoring and automatic control method for processing pet palatability enhancers according to claim 3, characterized in that, Also includes: When the current variance is greater than or equal to the preset safety threshold and the duration reaches the preset time threshold, an over-limit state is triggered. In the over-limit state, the interpolation calculation of the double key value is suspended; The average viscosity is calculated by extracting the viscosity values ​​within the time window prior to triggering the over-limit state. The viscosity value during the over-limit state is locked as the average viscosity.

6. The online intelligent monitoring and automatic control method for processing pet palatability enhancers according to claim 5, characterized in that, Also includes: In the over-limit state, the temperature difference value is compared with a preset limiting threshold. When the temperature difference value is greater than the limiting threshold, the temperature difference value input to the preset PID algorithm is clamped to the limiting threshold.

7. The online intelligent monitoring and automatic control method for processing pet palatability enhancers according to claim 1, characterized in that, After generating the heating command subject to the aforementioned heating upper limit constraint, the process further includes: When the viscosity value is greater than or equal to the preset wall-binding threshold, the difference between the inner temperature sequence and the outer wall temperature sequence is calculated to extract the inner temperature slope and the outer wall slope. When the internal temperature slope is less than zero and continues for a preset determination period, and the external wall slope is greater than a preset abrupt change threshold, thermal inversion is determined to have occurred and a forced intervention flag is generated.

8. The online intelligent monitoring and automatic control method for processing pet palatability enhancers according to claim 7, characterized in that, The reactor has a heating mechanism, a hinged cantilever, and a locking assembly connecting the reactor and the hinged cantilever. The hinged cantilever has a stirring motor connected to a stirring paddle extending into the reactor. After generating the forced intervention flag, the following steps are included: Output a heating command matching a preset cutoff threshold to the heating mechanism; A closing command is issued to the locking component; Based on the closing command, the locking component is driven to lock, and the articulated cantilever is converted from a clearance state to a rigid state.

9. The online intelligent monitoring and automatic control method for processing pet palatability enhancers according to claim 8, characterized in that, After the articulated cantilever is converted from a clearance state to a rigid state, the following steps are included: Collect the confirmation signal fed back after the locking component is engaged; A cell-wall breaking command is sent to the stirring motor only when the confirmation signal is received; Based on the cell-wall breaking command, the stirring motor is driven to operate at a rate greater than the nominal threshold with a preset pulse width; When the preset pulse width is reached, a speed reduction command is sent to the stirring motor, and an unlocking command is sent to the locking component simultaneously. The unlocking command drives the locking component to release the lock, thereby restoring the articulated cantilever to the clearance state.

10. An online intelligent monitoring and automatic control system for pet palatability enhancer processing, used to implement the online intelligent monitoring and automatic control method for pet palatability enhancer processing as described in any one of claims 1-9, characterized in that, include: The data acquisition module is used to acquire the current sequence, the internal temperature sequence of the reactor, and the external wall temperature sequence within the current time window; The feature separation module is used to separate the current sequence into low-frequency current components with frequencies below a preset lower limit and high-frequency current components with frequencies within a preset range using a preset filtering algorithm, and to calculate the current variance of the high-frequency current components. The viscosity mapping module is used to extract the average temperature of the internal temperature sequence when the current variance is less than a preset safety threshold. The average temperature and the low-frequency current component are input into a preset mapping table, and the viscosity value is output. The parameter reconstruction module is used to calculate the differential slope of the viscosity value under three consecutive time windows. If the differential slope is greater than the upper limit of the slope and maintains a preset anti-shake period, the attenuation coefficient corresponding to the viscosity value is extracted. The preset heating upper limit is reduced according to the attenuation coefficient, and the integral constant of the preset PID algorithm is extended. The instruction generation module is used to calculate the temperature difference between the outer wall temperature sequence and the inner temperature sequence, input the temperature difference into the PID algorithm after reconstructing the integral constant, and generate a heating instruction subject to the heating upper limit constraint.

Images