Segmented feeding control method for cooking machine based on cooking process

Abstract

The invention discloses a cooking process-based cooking machine segmented feeding control method, and relates to the technical field of intelligent cooking control, and the method comprises the steps: building a feeding transaction, executing bidirectional micro-motion detection on a channel, evaluating the health degree of a weighing signal, and outputting a gating conclusion and a risk type; feeding according to a short coding beat pulse, collecting a single weighing signal and synchronizing related calculation to obtain a feeding amount estimated value and calculation confidence, stopping when the standard is reached, and supplementing feeding when the standard is insufficient; when the estimated value of the putting amount is stagnated or the confidence coefficient is reduced, fault typing is carried out based on an equivalent resistance index, a forward and reverse response asymmetry index and the consistency of the actuator, and the typing confidence coefficient is given; and implementing self-recovery or loss stopping according to the types, rechecking resistance falling, and submitting a transaction and recording a log after the delivery amount is increased and the confidence reaches the standard. Single-weighing verifiable metering is realized, misjudgment shutdown is reduced, and product consistency and continuous operation are improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent cooking control technology, specifically to a segmented feeding control method for a stir-fry machine based on cooking processes. Background Technology

[0002] To reduce reliance on skilled operators and ensure consistent product output, automated stir-frying or heating equipment is equipped with multiple feeding and batching units. Main ingredients, auxiliary ingredients, solid seasonings, and liquid sauces are stored in separate compartments, quantitatively supplied, and dispensed in stages. In actual operation, this equipment often operates under conditions of high temperature, oil fumes, steam, and continuous vibration. It is also required to operate continuously during peak hours and quickly switch between recipes and batches. The feeding media include powders, granules, flakes, and high-viscosity semi-fluids, and their physical properties vary with temperature, moisture content, particle size distribution, settling time, and pipeline back pressure.

[0003] Existing segmented feeding control typically measures feed based on predetermined times, valve opening times, pump or motor speeds or steps. Some systems also incorporate individual flow detection or use differential calculations of the total weight of multiple feed boxes to save on sensors. Under these conditions, powders may become damp and clump together, forming bridging; particles may become stuck in narrow channels; liquid sauces may have wall adhesion, backflow, or valve seat residue, leading to a narrowing of the effective flow path; pump actuators may experience idling, air resistance, or flow drift due to hose aging; weighing and flow signals are susceptible to vibration, thermal drift, and condensation, resulting in noise and zero-point offset. Therefore, deviations may occur during the segmentation process, where control commands are issued but the actual delivery is insufficient or non-delivered, accumulating in subsequent segments. For example, if a segment experiences partial blockage or idling and is judged as complete before proceeding to the next segment, it will cause deviations in proportion and cycle time. If judgment is based solely on no flow or simple weight difference, non-material shortage factors such as execution chain failure, valve jamming, or metering chain drift can also trigger false alarms and unnecessary downtime. Vibration rebound and wall residue may also mask the true under-delivery, making these deviations difficult to detect in a timely manner. The inability to identify and distinguish these deviations in a timely and reliable manner will lead to negative impacts such as raw material waste, reduced batch consistency, reduced equipment utilization, increased maintenance frequency, and process interruptions during peak periods.

[0004] Therefore, the current technical problem is that under complex working conditions and changes in physical properties, it is impossible to make an online and verifiable judgment on whether each feeding segment has occurred and whether the feeding amount meets the standard, and to separate the reasons for meeting the standard. Summary of the Invention

[0005] (a) Technical problems to be solved To address the shortcomings of existing technologies, this invention provides a segmented feeding control method for a cooking machine based on cooking processes. It obtains the estimated feeding amount and calculation confidence level by collecting a single weighing signal and simultaneously performing related calculations. The method stops feeding when the target is reached and replenishes feeding when insufficient. When the estimated feeding amount stagnates or the confidence level decreases, it performs fault classification based on equivalent resistance indicators, forward and reverse response asymmetry indicators, and actuator consistency, and provides a classification confidence level. Based on the classification, it implements self-recovery or loss prevention. After verifying the resistance decrease, feeding amount increase, and confidence level reaching the target, it submits the transaction and records the log, reducing false shutdowns and improving product consistency and continuous operation. This solves the technical problems described in the background art.

[0006] (II) Technical Solution To achieve the above objectives, the present invention provides the following technical solution: A segmented feeding control method for a cooking machine based on cooking technology includes: establishing a feeding transaction and generating target feeding quantity and channel and material type identifiers; applying a bidirectional micro-motion detection sequence to the channel to collect motor current and speed, calculating equivalent resistance index and forward and reverse response asymmetry index, and assessing the health of the weighing signal to obtain a gating conclusion; based on the gating conclusion, allocating short coded beats to the channel and encoding pulse feeding, collecting single weighing signals to obtain a weight change sequence, synchronously correlating the short coded beats with the weight change sequence to obtain a feeding quantity estimate and calculation confidence level, and stopping or supplementing feeding accordingly; When the feed volume estimate stagnates or the calculation confidence level falls below the threshold, a mutually exclusive fault classification is performed based on the equivalent resistance index, the forward and reverse response asymmetry index, the feed volume estimate growth trend, and the current and speed consistency index, and the classification confidence level is output. Self-recovery or stop-loss is performed based on the mutually exclusive fault classification and classification confidence level. The bidirectional micro-motion detection sequence and synchronous correlation calculation are reused. When the equivalent resistance index falls back, the feed volume estimate increases, and the calculation confidence level meets the standard, the feed transaction is submitted; otherwise, the log is recorded.

[0007] Furthermore, when establishing a material feeding transaction, the system generates the target feeding quantity, allowable error, maximum allowable duration, maximum number of retries, and cumulative feeding limit. After executing a bidirectional micro-motion detection sequence of forward short drive, pause, reverse short drive, pause, and forward short drive, the system outputs the gating conclusion and risk type.

[0008] Furthermore, the equivalent resistance index is calculated by integrating the response sequence of motor current and speed within the detection window, and the forward and reverse response asymmetry index is calculated by logarithmic transformation of the ratio of the forward and reverse equivalent resistance indexes. The controller uses the equivalent resistance index and the forward and reverse response asymmetry index as the basis for determining the risk type.

[0009] Furthermore, the weighing signal health assessment includes weighted accumulation of the change rate of the short-window weighing sequence and determination of the ratio of endpoint difference to fluctuation within the window. At the same time, saturation and jump detection are performed, and a gating conclusion is output indicating whether the metering chain is available or unavailable. The gating conclusion is then written into the feeding transaction.

[0010] Furthermore, the short coded beats are selected from the candidate beat table and bound to the channel identifier. When selecting, an upper limit threshold of cross-correlation is applied to the short coded beats of different channels. The coded pulse feeding drives the actuator to switch between the running segment and the pause segment according to the short coded beat, and records the short coded beat identifier for synchronous correlation calculation.

[0011] Furthermore, the weight change sequence is formed by taking the endpoint difference of a single weighing signal according to the boundary of the beat unit. The endpoint difference is the difference between the start and end endpoints of the beat unit, and the delivery direction is unified to positive according to the arrangement direction of the weighing sensor. The controller takes the median of the resampled values ​​at the endpoints of the beat unit and then generates the weight change sequence.

[0012] Furthermore, the confidence level is calculated by normalizing the contribution weights of the short coded beat and the weight change sequence. The contribution weights are determined based on the weight change amplitude after beat modulation. The confidence level is output using the information entropy normalization method and is used by the controller to determine whether to stop feeding, add a small amount of feed, or proceed to the next step.

[0013] Furthermore, the mutually exclusive fault classification includes high resistance blockage or jamming, bridging or intermittent feeding, actual material shortage or idling, execution chain failure and metering chain failure; the inputs of the mutually exclusive fault classification include equivalent resistance index, forward and reverse response asymmetry index, feed quantity estimate, solution confidence and consistency index, and output classification confidence.

[0014] Furthermore, the consistency index is obtained by aligning the running and stopping segments of the coded pulse feeding to collect motor current and speed, calculating the current response amplitude and speed response amplitude, and determining their degree of mismatch; step three combines the consistency index with the equivalent resistance index to distinguish between execution chain failure and high-resistance blockage or jamming.

[0015] Furthermore, the controller categorizes the processing authorization into automatic processing, conservative processing, and stop-loss processing based on the classification confidence level, and restricts the action sequence with the maximum number of retries, the maximum allowed duration, and the cumulative upper limit of delivery; it submits the delivery transaction when the submission conditions are met, and rolls back the delivery transaction and records the log when the submission conditions are not met and any limit is reached.

[0016] (III) Beneficial Effects This invention provides a segmented feeding control method for a cooking machine based on cooking technology, which has the following beneficial effects: By treating each feeding segment as a feeding transaction with a fixed target feeding amount, allowable error, maximum allowable duration, maximum number of retries, and cumulative feeding limit, consistent traceability is ensured, facilitating seamless integration of subsequent metering, classification, and disposal on the same object. Before feeding, a bidirectional micro-motion detection sequence is employed, combined with motor current and speed to obtain equivalent resistance and forward / reverse response asymmetry indicators. Gating conclusions and risk types are derived from the health of the weighing signal, exposing high resistance, bridging, idling, and metering chain anomalies before feeding, avoiding false alarms from blind feeding.

[0017] After the gate passes, a short coding cycle is used to feed the material using coded pulses. The single weighing signal is aligned with the weight change sequence and the correlation calculation is performed simultaneously to obtain the estimated feed amount and the calculation confidence level. This enables independent weighing without channels, verifiable metering with the flow meter, and stopping when the target is reached and supplementing feed when insufficient.

[0018] When the estimated feed volume stops or the confidence level decreases, the equivalent resistance index, the asymmetric index of positive and negative responses, the growth trend of the estimated feed volume, and the consistency index are used to output a classification confidence level through mutually exclusive fault classification. This allows for the differentiation of blockages, intermittent feeding, material shortage idling, execution chain failures, and metering chain failures, preventing misjudgments. Based on the mutually exclusive fault classification and classification confidence level, graded self-recovery or downgraded loss prevention is implemented, subject to constraints such as the maximum number of retries, the maximum allowable duration, and the cumulative feed volume cap. When the equivalent resistance index falls back, the estimated feed volume resumes growth, and the confidence level meets the standard, a feed volume transaction is submitted and logged, ensuring the continuous and stable operation of the closed-loop support system. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the process flow of the segmented feeding control method for a stir-fry machine based on cooking technology according to the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Please see Figure 1 This invention provides a segmented feeding control method for a stir-fry machine based on cooking technology, including: Step 1: Before entering a certain feeding segment, the controller creates a feeding transaction and uses active detection gating to determine whether the execution chain and metering chain are in a submittable state, thereby providing traceable evidence for subsequent verifiable metering and differentiated processing.

[0022] In a continuous production scenario in the kitchen, multiple feeding channels need to deliver main ingredients, auxiliary ingredients, solid seasonings, and liquid sauces in different process stages. Powders can clump together when damp and form bridging at the feeding port; particles may get stuck; viscous sauces can adhere to the walls and change back pressure; pumps and motors may experience idling, slippage, or brief jamming. At the same time, the weighing structure is prone to noise and zero-point offset due to vibration and thermal drift. If the controller directly feeds materials without confirming that the conditions for submission are met, even if timed, step-by-step, or single-path flow judgments are used, it will be difficult to distinguish between no feeding, insufficient feeding, execution chain failure, and metering chain failure. This will lead to a superposition of false alarms and invalid handling, disrupting the connection between process stages.

[0023] After receiving the material feeding trigger signal from the process section, the boundary parameters of the material feeding transaction are first locked. Then, bidirectional micro-motion detection is performed on the target channel and resistance evidence is extracted. Next, the health of the weighing signal is assessed to form the metering chain availability conclusion. Finally, the submittable flag and risk type mark are output as the conditional branch for entering step two or transitioning to step four.

[0024] A single segmented material feeding is abstracted into a feeding transaction, ensuring that the same set of boundary parameters and referential objects are maintained even when anomalies occur within the same feeding segment. The controller first reads the process segment identifier, channel identifier, and material type identifier, and then generates and fixes the target feeding quantity, allowable error, maximum allowable duration, maximum number of retries, and cumulative feeding limit, thereby restricting subsequent actions to within the budget boundaries of this feeding segment.

[0025] The target delivery volume and allowable error constitute the target range for this segment, the maximum allowable duration limits the handling budget for this segment, and the maximum number of retries and the cumulative delivery limit define the boundaries for replenishment and obstacle clearing. The material type identifier is used to map the detection drive gear and pause rhythm, so that the powder channel uses short drives in conjunction with pauses and falls, the particle channel uses slow starts and stops to reduce compression, and the viscous sauce channel uses longer pauses to release back pressure.

[0026] The above mapping couples the interpretation of the detection signal with the material state in the same direction, avoiding the misinterpretation of property changes as execution chain failures. To ensure that objects referenced in subsequent steps do not drift, the controller writes the process section identifier, channel identifier, material type identifier, and target feed quantity into the feeding transaction record using fixed fields. Allowable error Maximum allowed duration Maximum number of retries Cumulative disbursement limit Furthermore, the probe drive gear and pause time are fixed as the probe configuration for this transaction; therefore, whether it enters the formal feeding in step two or the preprocessing branch in step four, the boundary and probe conditions at that time can be reproduced with the same set of fields, thereby avoiding parameter inconsistencies caused by repeated parameter building or cross-referencing during the abnormal handling process.

[0027] Specifically, the controller locks the target delivery volume when the transaction is created. Allowable error Maximum allowed duration Maximum number of retries Cumulative disbursement limit The material type identifier is mapped to the detection drive gear and pause beat, which serves as the fixed boundary for subsequent gating.

[0028] When in use, anomaly handling is carried out within the same feeding transaction boundary to avoid overstepping boundaries during refeeding and retrying. Material type identification is bound to detection level to ensure that the interpretation of detection response is in the same direction as the change in physical properties, reducing misclassification. Boundary parameters are written into the feeding transaction record to provide unified input and consistent reference for subsequent steps.

[0029] A bidirectional micro-motion detection sequence is used to acquire load characteristics before material feeding, avoiding triggering actions only after complete stall. The detection sequence consists of a short forward drive, a pause, a short reverse drive, a pause, and another short forward drive. The drive amplitude is mapped from the material type identifier, making the detection produce an observable response without generating an accumulated feed amount.

[0030] The controller collects motor current and equivalent speed in two detection directions, respectively, and uses the integral ratio to form a directional resistance index, and the logarithmic ratio to form a forward and reverse response asymmetry index. The directional resistance index is used to characterize high resistance trends, and the forward and reverse response asymmetry index is used to distinguish between bridging, wall rebound, and transmission slippage. The equivalent speed can be converted from the encoder pulse interval or estimated from the back EMF; the integral can be implemented using trapezoidal integral, making the resistance evidence insensitive to instantaneous peak noise. To ensure the reproducibility of drive conditions, the start-stop slope is generated by piecewise linear interpolation, and the interpolation node is switched according to the material type identifier.

[0031] In setting the amplitude of the detection drive, the controller does not use the ability to drive material flow as the criterion, but rather uses an observable response without cumulative delivery as the criterion: for solid channels, the turning angle of the short detection drive is limited to the section where continuous delivery has not yet formed at the outlet, and the material is allowed to fall back under its own weight during the pause; for liquid channels, the pumping rate of the short detection drive is limited to the section that can only wet the outlet and is insufficient to form continuous dripping, and the local back pressure is released in the reverse short drive to prevent residual liquid from being squeezed out. Therefore, the detection sequence can be repeatedly executed without changing the calculation basis of the target delivery amount for this segment, while its response still reflects the changing trend of the current resistance state.

[0032]

[0033] Where: Directional resistance index : Characterize the first The converted resistance in each direction is a non-negative real number; direction number : Distinguish the detection direction, with a value range of . Direction number For positive direction, direction number For reverse; detection window duration : The length of the integration interval, whose values ​​are positive real numbers and satisfy the following conditions: Motor current sequence: Current changes over time during the detection period, non-negative real numbers, used as a torque proxy. Calculated velocity sequence The velocity during the detection period varies with time, is a non-negative real number, and serves as a displacement surrogate; time variable. The independent variable for integration takes values ​​in an interval. ; Zero bias protection Numerical stability constant, with a range of positive real numbers, used to suppress the denominator from approaching zero due to low-speed jitter;

[0034] Where: Asymmetric index of forward and reverse responses This represents the relative difference between positive and negative resistance, taking values ​​that are real numbers. Positive values ​​indicate greater positive resistance, and negative values ​​indicate greater negative resistance. It is used to indicate risk type; directional resistance indicator. Positive resistance indicator, non-negative real number; directional resistance indicator. : Reverse resistance indicator, non-negative real number; prevents zero bias As mentioned above, it is a positive real number; The controller executes a bidirectional micro-motion detection sequence to collect data. and And calculate the directional resistance index Directional resistance index Asymmetric indicators of positive and negative responses This information is then recorded in the material feeding log. A uniform sampling period is used for both motor current and calculated speed. Data collection is performed within the detection window. Each sampling time is The sampling point number A non-negative integer. Detection window. Corresponding number of sampling points .

[0035] When in use, evidence of the execution chain load is obtained before formal feeding, so that early signs of high resistance, bridging and idling can be recorded. The directional resistance index is formed by integral ratio, which reduces the impact of instantaneous peak value on interpretation and keeps it in the same direction as the load. The positive and negative response asymmetry index provides direct evidence for risk type marking, avoiding a single interpretation based solely on the absence of feeding.

[0036] After detection, a health assessment is performed on each weighing signal to avoid incorporating vibration, jitter, or thermal drift into subsequent delivery calculations. The controller selects an assessment window adjacent to the detection sequence, and uses the ratio of the rate of change energy of the weighing signal to the endpoint difference to distinguish between jitter-dominated and drift-dominated states, outputting a metering chain availability conclusion. When the weighing health index reflects a high proportion of jitter or non-negligible endpoint drift, the controller does not perform numerical correction on the weighing signal. Instead, it directly invalidates the metering chain availability flag and retains the signal segment index of the original assessment window, allowing subsequent steps to trace the trigger source of this gating decision even when degrading to open-loop operation.

[0037] The integral of the absolute value of the rate of change is the jitter energy surrogate, the difference between window endpoints is the drift surrogate, and the weighing health is the exponential weight. The exponential weight indicates the environment of the initial window, making it closer to the transient state after the detection ends. The rate of change is obtained by dividing the difference between adjacent samples by the sampling period.

[0038] Where: Weight health index : Represents the percentage of jitter, with values ​​ranging from non-negative real numbers, used for metering chain availability gating; weighing signal The weighing output changes over time and is a real number; the window starts at [time]. The evaluation window starts at the beginning of the evaluation window and takes the value of a real number; the evaluation window duration is also specified. : Evaluate the span, the range of values ​​is positive real numbers and satisfies ; Attenuation coefficient : The rate of decay of the exponential weights, a positive real number; derivative operator Weighing change rate, discretely implemented using differential substitution; prevents zero offset. Consistent with the above, the value range is positive real numbers; time variable. The independent variable for integration takes values ​​in an interval. ; The controller calculates the weighing health index within the evaluation window. The output metering chain availability flag is compared with the threshold and written to the feeding transaction record.

[0039] The weighing signal is denoted as When the weighing sensor is located on the feeding hopper / tank side, the occurrence of feeding will lead to... A decrease; when the weighing sensor is positioned on the receiving container side, the discharge will cause... Increase. To make the estimated deployment amount rise. The controller will always use the increase in delivery as the positive metering, and the controller will... The weight change of each beat unit is constructed as the change in the mass of the delivery. It is defined as the time interval within that beat unit. The net change is positiveized.

[0040] When weighing at the feeding side (when the weight decreases due to feeding):

[0041] When weighing on the receiving side (when the weight increases due to material feeding):

[0042] in For the first The start time of each beat unit For the first The end time of each beat unit. Therefore... The value is non-negative when the delivery occurs, and approaches zero during the pause period, thus ensuring... Can take non-negative values ​​and Positive values ​​are acceptable. Change in the quality of the input. : No. The quality change of each beat unit is a non-negative real number; weighing signal : Weighing output, value range is real number; Start time of cycle unit : Time stamp, with values ​​ranging from real numbers.

[0043] End time of beat unit: Time marker, value range is real number. Beat unit number : Index, with a value range of integers. .

[0044] When in use, the availability of the metering chain is determined before proceeding to subsequent measurements to avoid false compliance or undercompletion caused by weighing jitter. The weighing health index is formed by weighted integral of the rate of change index and endpoint difference to maintain comparability for different total weight levels. The metering chain availability flag serves as a gating input, providing a traceable basis for subsequent path selection.

[0045] The system integrates evidence of execution chain resistance and measurement chain availability to output a gating conclusion, which is then transformed into a conditional branch for subsequent steps. The gating conclusion includes a commit flag and a risk type flag. The risk type flag takes a unique value between the execution chain risk and the measurement chain risk, and includes a snapshot of key indicators, which includes a directional resistance indicator. Directional resistance index Asymmetric indicators of positive and negative responses Weight health indicators And transaction boundary parameters.

[0046] The controller first determines the high resistance trend based on the directional resistance index, then determines the bridging or transmission slippage trend based on the forward and reverse response asymmetry index, and then determines the availability of the metering chain by combining the weighing health index. When the metering chain is unavailable, the gating conclusion is fixed to the metering chain risk and the process moves to the conservative degradation branch in step four. When the execution chain has a high resistance or idling risk, the gating conclusion is fixed to the execution chain risk and the process moves to the preprocessing branch in step four. Only when the execution chain is within the commit range and the metering chain is available, the commit flag is set to valid and the process proceeds to step two.

[0047] The controller writes the gating conclusion into the material feeding transaction as a conditional branch for material feeding. If the commit flag is valid, it proceeds to step two; otherwise, it proceeds to step four.

[0048] For simplicity and clarity, for example: When the equipment enters the seasoning section to add liquid sauce, the controller creates a feeding transaction and locks the boundary parameters before driving the pump to perform bidirectional micro-motion detection. The sauce outlet shows a brief period of wetness but no continuous dripping. The directional resistance remains high, and the asymmetry index of the forward and reverse responses is high. The weighing health index meets the requirements for the metering chain to be usable. The controller outputs a high resistance risk in the execution chain, and the submission flag is invalid. It then proceeds to step two and enters the preprocessing branch of step four, recording a snapshot of key indicators. When detecting powder, the feed inlet shows slight vibration and stops before falling back. The asymmetry index of the forward and reverse responses is bridged. The controller marks the execution chain as a risk, not a material shortage risk, reserving conditions for the subsequent arch-breaking cycle.

[0049] The controller is based on the directional drag index. Directional resistance index Asymmetric indicators of positive and negative responses With weight health index The output can submit flags and risk type markers and form conditional branches, while writing key indicator snapshots to the feed transaction record for reference in subsequent steps.

[0050] When used, the gating conclusion provides both the committable and risk causes, so that subsequent steps do not rely on temporary assumptions when selecting measurement or disposal paths. Execution chain risks and measurement chain risks are mutually exclusive, avoiding misjudging weighing noise as execution chain failure or idling as material shortage. Snapshots of key indicators are retained with the transaction records, providing a consistent source of input and traceability for subsequent measurement and disposal.

[0051] Step 2: The submittable flag output in Step 1 is valid and the health index is weighed. When the metering chain is available, the controller drives the feeding actuator with short coded beats to form a visible run-stop rhythm, and responds to single weighing signals. By performing synchronous correlation calculations, we obtain the estimated delivery volume and the calculation confidence level, thereby transforming the change in total weight into delivery evidence attributable to the channel.

[0052] The actual operating conditions of segmented feeding involve both physical property fluctuations and structural disturbances: powders become damp and bridge at the feed inlet, particles get stuck in narrow channels, and viscous sauces adhere to the valve seat and pipe wall, altering the back pressure; equipment vibration affects the weighing signal. This results in rebound and drift. If the controller only targets the desired delivery volume... Allowable error Using continuous operation or simple timed and step-by-step operation, the actuator completing its action is not equivalent to completing the delivery. Furthermore, single-path flow judgment can easily misinterpret a fault in the actuator chain as a material shortage when the pump is running dry or the valve is stuck. Step one has already included the directional resistance index. Directional resistance index Asymmetric indicators of positive and negative responses This evidence is solidified as proof of material feeding transactions and measured using health indicators. Since the metering chain is given a usable conclusion, step two needs to make the single weighing output reusable evidence of the delivery quantity without adding channel-level sensors.

[0053] After the controller enters step two, it continues to use the target feeding quantity and allowable error as boundary parameters of the feeding transaction. Maximum allowed duration With the cumulative upper limit Subsequently, the controller generates a short-coded beat sequence for each channel participating in this feeding segment, and maps this beat sequence to the actuator's run-stop drive rhythm, so that the feeding action forms a weight ladder in phase with the beat in the weighing signal. The controller then uses beat-aligned differential and short-window integral to... The data is converted into a weight change sequence and synchronously correlated with the channel beat sequence to obtain the estimated feed rate. Then, a calculation confidence level is constructed to determine whether to allow closed-loop stopping or to enter micro-feeding. If the feed rate estimate stagnates or the calculation confidence level is insufficient, the controller writes the feed rate estimate, growth trend, and calculation confidence level into the feeding transaction record and submits it to step three for fault classification.

[0054] First, the issue of distinguishing cycle time sequences needs to be addressed. Since single weighing only observes changes in total weight, the contribution of each channel can only be separated through synchronous correlation calculations when the cycle time is distinguishable in terms of temporal structure. The controller has already locked the channel identifier and material type identifier in the feeding transaction record of step one. Therefore, it does not use random trial and error when generating cycle time sequences, but instead uses an engineering path indexed by a candidate table: candidate cycle time tables are pre-set for powder, particle, and liquid channels respectively. Each candidate cycle time consists of a finite-length run / pause unit. The controller first selects the candidate cycle time length according to the number of channels and uses bit-by-bit similarity to avoid long segments of phase co-occurrence in channel cycles. To match the cycle time with the mechanical response, the cycle time unit duration is mapped by the material type identifier: the powder channel uses short runs and long pauses to allow for a drop and create a gap; the liquid channel uses a slow start and stop and releases back pressure during the pause; and the particle channel uses a slow start to reduce compression. This mapping is saved with the feeding transaction record for easy reproduction.

[0055] The controller indexes short-encoded beat sequences from the candidate beat table based on the channel identifier and material type identifier, and uses bit-by-bit similarity constraints to filter channel beat combinations, so that the beat unit duration corresponds one-to-one with the material type before writing it into the feeding transaction record.

[0056] When in use, the beat distinguishability constraint separates the delivery rhythm of different channels in time structure, providing a separable input form for single weighing calculation. The candidate table index avoids the non-reproducibility caused by random generation, ensuring that the same recipe has a consistent beat definition on different batches of devices.

[0057] The short coded cycle sequence is implemented as the actuator's run-stop drive, creating a visible step on the weighing signal for the delivery action. The controller does not treat the cycle as a simple switching sequence, but rather uses the resistance evidence from step one as the selection condition for the drive gear: when the directional resistance index... and When approaching the upper edge of the gate, the running section adopts a low gear and extends the pause, allowing the material to fall back and reducing instantaneous impact; when the forward and reverse response are asymmetrical... When the positive resistance is too high, a starting ramp generated by piecewise linear interpolation is added before each running segment, and a longer release time is maintained during the pause segment to reduce material accumulation at the discharge port. During the drive process, the controller operates at the maximum allowable duration. Limit the number of beat cycles to the cumulative maximum deployment limit. Limiting the cumulative drive budget for a single channel ensures that the cycle feed does not exceed the boundary of the current feed transaction and intrude into the next process segment.

[0058] The visible actions on-site are as follows: intermittent dripping of liquid occurs in the running section, and dripping stops in the pause section; the solid screw pushes the powder closer to the discharge port in the running section, and the powder falls back in the pause section, creating a short-term gap. This running-pause development is used as the input for the weighing calculation in sub-step 202.

[0059] Among them, the controller is based on the directional resistance index. Directional resistance index Asymmetric indicators of positive and negative responses Set the running speed and pause duration, and set the maximum allowed duration. With the cumulative upper limit Constrain the beat cycle to make the actuator output a visible run-stop rhythm.

[0060] When in use, resistance evidence is involved in gear and ramp selection, making the cycle drive closer to the range that the execution chain can withstand, reducing the damage to the symmetrical weighted steps caused by start-stop impacts, and the feed transaction budget constraint makes the cycle cycle close within the same transaction boundary, providing clear upper limit conditions for subsequent stops and refeeding.

[0061] Single weighing signal The weight change is converted into a one-to-one correspondence with the beat position, and the total weight change is separated into estimated values ​​of the delivery volume for the target channel using synchronous correlation calculations. The controller operates on the weighing health index. When the gating conditions are met, the sampling interval is divided according to the beat unit, and within each beat unit, the sampling is performed... First, perform a first-order differential and then a limiting operation to suppress the start-stop rebound peak; then, perform a short-window integration on the differential value to obtain the weight change, so that the running segment shows a stable decrease while the stopping segment approaches zero.

[0062] Then, the controller reads the channel cycle sequence from the material feeding transaction record, multiplies and accumulates the weight changes, and estimates the feeding amount using the channel scale coefficient; the channel scale coefficient comes from assembly calibration or maintenance calibration, and the update time is limited to the maximum allowable duration. and cumulative upper limit In order to prevent scale factor drift caused by a single anomaly:

[0063] Where: relevant cumulative amount : No. The cumulative results of the relevant channels are real numbers; the estimated delivery volume is... : No. The estimated delivery volume for each channel is a non-negative real number, used for stopping, re-delivering, and as input in step three; the channel number is also included. A unique, positive integer number corresponding to each channel identifier, used to index the channel beat sequence and channel scale factor; channel scale factor. : Conversion factor, a positive real number, sourced from assembly calibration or maintenance calibration; In calibration mode, the controller only drives the first... The channel executes several tick units (which can be taken) All of The calibration beat was determined, and the change in calibration delivery quality was obtained using the endpoint difference method. Simultaneously, the calibration-related cumulative amount is calculated according to the same rhythm. Therefore, the following calculations can be performed:

[0064] Among them, zero bias protection This is used to avoid division by zero due to a lack of valid changes within the calibration window. After calibration is complete... Write to non-volatile memory and retrieve it by channel identifier in subsequent material feeding transactions.

[0065] Calibration of changes in delivery quality The quality change obtained by endpoint differential during the calibration phase, with values ​​ranging from non-negative real numbers; calibration-related cumulative amount. The cumulative sum obtained during the calibration phase, with values ​​ranging from real numbers; channel scale coefficient. : Conversion factor, with a range of positive real numbers.

[0066] Channel beat sequence : No. Channel 1 The sequence values ​​of the beat unit are within the range of... or It is also used for driving modulation and correlation calculation; weight change : No. The net weight change obtained from the construction within the cycle unit is a real number; the value is usually negative during the running segment and close to zero during the pause segment. beat length The number of tick units within a solution window, ranging from positive integers, determines the solution delay and discrimination capability; tick unit number. : Sequence position index, with values ​​ranging from an integer interval .

[0067] The controller constructs the weight change using beat-aligned differential, amplitude limiting, and short-window integration. And calculate the relevant cumulative amount based on the channel beat sequence. Compared with the estimated amount of delivery The estimated amount of delivery Write the material feeding transaction record and use it to stop the input in step three.

[0068] The controller will use the channel beat sequence Use a set of values ​​uniformly A binary sequence. If the original sequence is... The controller will then press A centering mapping is performed to make the mean of the beat sequence zero, reducing the bias of slow weighing drift on the relevant solutions. When the controller selects the beat sequence for each channel from the candidate beat list, it performs a round-robin test on any two channels. and Calculate the cross-correlation coefficient:

[0069] And require to meet ,in To pre-define the upper bound threshold for cross-correlation. Candidate beat list and threshold. The data can be written to the controller memory at the factory, or set during installation and commissioning according to the number of channels and the weighing noise level.

[0070] Channel beat sequence : No. Channel 1 The value of the beat unit is within the range of... Cross-correlation coefficients :aisle With channel The beats are cross-correlated, and their values ​​range from 1 to 2. Cross-correlation upper bound threshold The distinguishability constraint threshold has a range of values. Beat length : Number of beat units, value range is positive integer; Channel number : Index, with values ​​ranging from positive integers; beat unit number : Index, value range is Integers.

[0071] When used, the weight change is constructed according to the beat position, so that the total weight change is discretized into a traceable sequence, reducing the dependence on complex models. The synchronous correlation solution transforms the total weight change of a single weighing into an estimated value of the delivery volume that can be attributed to the channel, providing direct evidence for subsequent classification.

[0072] A confidence level is calculated to complement the estimated delivery volume, preventing the triggering of stop or re-delivery based solely on the estimated value during disturbances. The confidence level is constructed using the concentration of relevant contributions: if the weight change is mainly driven by the beat of this channel, the relevant contribution is concentrated at the beat position; if the weight change is mainly caused by touch, rebound, or superposition of other channels, the contribution is dispersed.

[0073] When calculating the relevant cumulative amount, the contribution amplitude of each beat unit is simultaneously calculated and normalized into a contribution weight, and then its concentration is evaluated in the form of information entropy. Information entropy can be implemented by fixed-point accumulation combined with logarithmic lookup table, which is convenient for execution on the controller. In closed-loop decision-making, the controller first checks whether the solution confidence level has reached the threshold, and then checks whether the estimated delivery amount has entered the target range. When both conditions are met, the actuator stops and the completion is recorded; if the estimated value is insufficient but the confidence level meets the threshold, a micro-feedback is initiated. The micro-feedback continues to use cycle-driven operation but shortens the cycle length and increases the pause ratio, while not exceeding the cumulative feeding limit. If the confidence level is lower than the threshold, no additional investment is made; instead, the estimated value and confidence level are submitted to step three for decision.

[0074] As an example: In the seasoning process section, the operator pours liquid sauce into the liquid tank and closes the lid, and pours solid seasoning powder into the screw hopper and closes the hopper lid. After the equipment passes through the gate in step one, it enters this step. The controller drives the pump and screw according to the cycle time. The displayed value of the weighing tray decreases in a stepwise manner during the running phase and remains flat during the pause phase. If the operator holds the tank during operation, causing a short-term jump in the weighing signal, the relevant contribution is dispersed at the cycle time position, the calculation confidence decreases, the controller maintains the current cycle time without triggering a stop or refill, and records this transaction for further classification in step three:

[0075] In the formula: Calculate the confidence level : Concentration of relevant contributions, value A larger value indicates a more reliable solution, and it is used to determine whether to stop the process and allow minor supplementary input; contribution weight. : No. Channel 1 The normalized result of the contribution amplitude of the beat unit, with values... And satisfy Its structure is ,in and Consistent with the above; Channel number : Positive integer, identifying the channel and used for indexing contribution weights and calculating confidence; beat cell number : Positive integers and their values ​​range from Used to indicate beat position; normalized summation sequence number : Positive integers and their values ​​range from This is the summation index used in the denominator of the contribution weights. (Temporal length) Consistent with the above, the value range is positive integers; zero bias protection. Numerical stability constant, a positive real number, used to avoid logarithmic zero input and achieve stable normalization.

[0076] Among them, the controller is based on contribution weight. Constructing the solution confidence level And by calculating the confidence level Compared with the estimated amount of delivery The decision to jointly drive the cessation and minor reinvestment will be calculated based on the confidence level. Estimated amount of deployment Its growth trend is recorded in the material input transaction record for reference in step three.

[0077] In practice, the concentration of relevant contributions allows for explicit verification of the solution's reliability, thereby suppressing false stops and false re-applications under disturbances such as touch and rebound. Both stops and re-applications are affected by the solution's confidence level. Target delivery volume Allowable error With the cumulative upper limit Consistent constraints ensure that the boundaries of material input transactions remain closed and do not spread across segments.

[0078] When using, If insufficient evidence is found, it is directly passed to step three, enabling the subtyping to utilize the causal clue of insufficient credibility rather than solely relying on the apparent insufficiency of the distribution volume. Step 3: Within the same material feeding transaction boundary, the insufficient feeding quantity estimate is broken down into mutually exclusive fault types, and the subsequent processing authorization is constrained by the type confidence level, so that Step 4 can proceed along a single chain without switching back and forth between multiple recovery paths.

[0079] Step 2 outputs the estimated deployment amount. With the solution confidence level It can provide channel attribution evidence for delivery, but under the combined conditions of oil fumes, high temperature and vibration, there may still be two types of deviations: the actuator operates according to the cycle but there is no continuous material falling from the outlet, and the weighing display jumps, which leads to a decrease in the confidence of the solution.

[0080] If the controller triggers an alarm or fixed clearing based solely on an insufficient estimated feed amount, it may misjudge a metering chain fault as an execution chain blockage, or misjudge an execution chain idleness as a genuine material shortage. This could trigger an invalid action in step four and disrupt the process segment connection. Therefore, step three requires adjusting the directional resistance index from step one. Directional resistance index Asymmetric indicators of positive and negative responses Weight health indicators Compared with the estimated amount of deployment in step two Calculate the confidence level The consistency of motor current and speed response is also included in the same decision chain.

[0081] After the controller enters step three, it continues to use the target feeding quantity as the boundary parameter of the feeding transaction. Allowable error Maximum allowed duration Maximum number of retries Cumulative disbursement limit The controller freezes snapshots of resistance evidence and solution evidence. Then, multiple solution windows are concatenated into a growth sequence to form a growth retention rate as growth evidence. Next, the cycle time segment is aligned with the motor current and speed samples to form consistency evidence. Once both types of evidence are complete, the controller outputs fault classifications according to mutually exclusive priorities and calculates the classification confidence level. Finally, the fault classification, classification confidence level, and key evidence are written into the material feeding transaction record as the input and authorization boundary for step four.

[0082] Instead of directly equating insufficient estimates of a single feed quantity with a failure, this approach connects multiple solution windows within the same feed transaction into a growth sequence, and uses the solution confidence level as the basis for the calculation. The contribution of each window is weighted, naturally reducing the weight of low-confidence windows caused by touches and bounces. The controller calculates the delivery increment for adjacent windows and takes the positive part of the delivery increment to obtain the growth contribution; at the same time, it takes the absolute value of the delivery increment to obtain the total change. After weight normalization, both are used to form the growth preservation degree, which is used to describe whether the delivery volume estimate accumulates in a single direction.

[0083]

[0084] Where: growth retention rate Growth stability index, value Used as a criterion for growth stagnation in fractal analysis; channel number : Channel index, a positive integer, used to index the delivery volume estimate and calculate the confidence level; Window number Window index, with a value range of: Integer; total number of windows The number of windows involved in the calculation, ranging from positive integers, and subject to the maximum allowed duration. Restrictions; Increased Deployment The difference between estimated drop volumes from adjacent windows is a real number used to describe growth or reversal; drop volume estimates : No. The estimated window delivery volume, a non-negative real number, comes from step two; calculate the confidence level. : No. The confidence level is calculated using a window, and the range of values ​​is [value missing]. From step two; zero bias prevention : Numerical stability constant, which is a positive real number.

[0085] The controller uses the estimated delivery volume. Window sequence and solution confidence Calculate growth preservation for input and maintain growth Write the material feeding transaction record for sub-step 302 to call.

[0086] When used, growth retention Transforming single-point deficiencies into time-series evidence ensures that the classification is based on continuous growth or stagnation rather than instantaneous fluctuations, thus improving confidence levels. The weighted participation suppresses the impact of low-confidence windows on growth evidence, thus maintaining growth stability. The criteria for reviewing the recovery and growth of the amount of funds allocated in step four are consistent, which facilitates the smooth flow of the steps.

[0087] Explicitly define command-response consistency to distinguish between execution chain with no valid output and metering chain with unreliable output. The controller aligns the cycle time segment of step two with the motor current and speed sampling, extracts the rising edge of the motor current and the rising edge of the speed within the cycle time, and combines this with the directional resistance index from step one. Directional resistance index Asymmetric indicators of positive and negative responses As an explanatory boundary: when the motor current increases during operation but the speed does not increase accordingly, and the directional resistance index snapshot is high, evidence of high load without displacement consistency is formed; when the motor speed increases during operation but the motor current approaches the no-load snapshot, and the increase maintains its value... When the value remains consistently low, evidence of consistent displacement under no-load conditions is formed; when both the motor current and speed responses are normal, the confidence level can be calculated. Consistently low and accompanied by weight health indicators When the snapshot is too high, it forms evidence of inconsistency between normal response and measurement instability.

[0088] The rising edge of the motor current can be obtained by three-point differential differentiation and sign-holding filtering, while the rising edge of the speed can be obtained by smoothing the inverse of the encoder pulse interval through a sliding window. If the speed information is obtained by Hall pulse interval or back EMF estimation, the same alignment and extraction process is used.

[0089] The controller aligns the cycle time segment with the motor current and speed samples to generate consistency evidence, and then compares the consistency evidence with the directional resistance index. Directional resistance index Asymmetric indicators of positive and negative responses Weight health indicators It should be written into the material feeding transaction record.

[0090] Within each cycle segment, the average motor current and average speed are calculated, and the difference between these values ​​and the corresponding average values ​​during the pause segment is used to obtain the current response amplitude and speed response amplitude. A consistency mismatch index is then constructed using a normalized approach. :

[0091] in For the first Channel current response amplitude, For the first Channel rotational speed response amplitude, To prevent zero bias. The larger the value, the more significant the mismatch between current and speed response, which is used to support the classification and authorization of execution chain faults.

[0092] Inconsistency mismatch index : The degree of consistency mismatch, with a value range of . Current response amplitude : The absolute value of the difference between the mean currents of the running and stopping phases, taking the range of non-negative real numbers; speed response amplitude. The absolute value of the difference between the average speeds during the running and stopping phases, taking the range of non-negative real numbers; zero offset protection. Numerical stability constant, taking values ​​of positive real numbers; channel number. : Index, with a range of positive integers.

[0093] When in use, the consistency evidence separates the actuator operation from the effective deployment, so that the clues to the causes of idling, slippage and high resistance jamming are preserved. The consistency evidence is constrained by the resistance snapshot, avoiding misinterpreting a simple increase in current or a simple fluctuation in speed as blockage or slippage.

[0094] Output mutually exclusive fault classification to ensure that step four is handled along a single chain. The controller uses a priority chain: when the weighing health index... The snapshot is too high and the solution confidence is too low. If the reading is low within a continuous window, and the evidence of consistency is normal but the measurement is unstable, the ruling is a measurement chain failure; when the measurement chain is available and the growth is maintained... When the level remains consistently low, and consistent evidence shows displacement during idling, the ruling is that there is a genuine material shortage or idling; when the metering chain is available and the growth rate is maintained... The value is low, and the evidence of consistency is high load with no displacement and directional drag index. Directional resistance index When the high-resistance snapshot is consistent, the determination is high-resistance blockage or jamming; when the estimated delivery volume is... The transitions between windows are discontinuous and the confidence level is calculated. Synchronous fluctuations, with simultaneous positive and negative responses from asymmetric indicators When the unidirectional resistance is too high, the ruling is bridging or intermittent feeding; when the evidence of consistency is that the command is issued but the response is abnormal and does not meet the resistance snapshot interpretation boundary, the ruling is execution chain failure.

[0095] As an example: In the seasoning process section, the screw in the solid seasoning channel operates on a rhythmic beat. On-site observation shows that the motor noise increases during the running section and stops during the pause section, with only occasional powder falling from the outlet; the weighing display shows a swing back after descending during the running section, indicating an estimated feed amount. It exhibits discontinuous jumps and solves the confidence level Fluctuations, growth retention Unstable, with asymmetric indicators of positive and negative responses. The forward resistance is too high. Based on this, the controller determines whether to proceed with bridging or intermittent feeding, and assigns this parting line to step four for cycle reconfiguration and arch-breaking micro-actions. In another scenario, operator contact with the hopper causes a weighing jump; the confidence level is calculated accordingly. If the reading remains consistently low while the motor response is normal, the controller determines it as a metering chain fault and restricts the triggering of obstacle clearing actions.

[0096] Among them, the controller is based on the weighing health index Calculate the confidence level Growth Maintenance Estimated amount of deployment Window performance and directional resistance indicators Directional resistance index Asymmetric indicators of positive and negative responses The failure classification is mutually exclusive with the consistency evidence output and written into the material feeding transaction record as the input for step four.

[0097] In use, the priority chain ensures that metering chain faults are eliminated first, reducing the likelihood of misinterpreting weighing jumps as blockages, and improving resistance snapshots and growth retention. Joint participation in the adjudication process will provide a clear boundary between high-resistance congestion and intermittent bridge construction.

[0098] The output classification confidence score is used to limit the automatic recovery permissions in step four. The controller constructs the evidence cost for the five fault classifications. The cost of a failure in the metering chain is determined by the weighing health index. With the solution confidence level Dominant, high-resistance blockage or stagnation costs are determined by directional resistance indicators. Directional resistance index With growth Dominant, bridging or intermittent material feeding costs are determined by asymmetric indicators of positive and negative responses. Compared with the estimated amount of delivery The intermittent nature of the trend is dominant, and the real cost of material shortages or idle operations is determined by consistent evidence and the degree of growth maintenance. The primary cause of execution chain failure costs is the mismatch between consistent evidence and resistance snapshots.

[0099] The evidence cost is exponentially normalized to obtain the subtyping weight, and the subtyping confidence is obtained from the information entropy of the subtyping weight. The exponential and logarithmic functions can be implemented using table lookup and piecewise linear interpolation, and the accumulation of information entropy can be implemented using fixed-point arithmetic.

[0100] Where: fractal weight : No. Class type weight, value And satisfy ; categorical confidence : Scale reliability index, with values... Used for the action authorization boundary in step four; type number : Fractal index, value range is This corresponds to high resistance blockage or jamming, bridging or intermittent feeding, actual material shortage or idling, execution chain failure, and metering chain failure; Classification number : Normalized summation index, with a value range of .

[0101] Cost of Evidence The degree of inconsistency of evidence, with values ​​ranging from non-negative real numbers, is determined by... , , , , , , Construction of consistent evidence; controller classification for each type of fault. The Cost of Constructing Evidence The cost of evidence is constituted by the sum of several soft threshold penalties, with the soft threshold being... The form allows the cost to change continuously as evidence approaches a threshold. Define the weighing health threshold. Forward and reverse asymmetric thresholds Consistency mismatch threshold Resistance threshold can be submitted. Solve the confidence threshold Growth maintenance threshold Zero bias protection : Numerical stability constant, a positive real number; Constant 5: Fractal number, used for normalization scale.

[0102] Among them, the controller constructs evidence cost And calculate the fractal weights With classification confidence Fault classification and classification confidence Write the material feeding transaction record and provide it to step four as the processing authorization boundary.

[0103] When using it, the classification confidence level Making the consistency of evidence explicit ensures that step four tends to downgrade and stop losses rather than clear obstacles when the evidence is similar, thus reducing the likelihood of deflation. The existing evidence constructed from steps one and two ensures that each disposal can be traced back to the same material feeding record. The classification weight and information entropy structure ensure that different classifications are comparable, providing a basis for a unified authorization scale across material types.

[0104] Step 4: The controller delivers the target amount. Allowable error Maximum allowed duration Maximum number of retries Cumulative disbursement limit Under the constraints, the mutual exclusion fault classification and classification confidence scores output in step three are... This is transformed into a sequence of actions and actions, and the directional resistance index is reused after each round of actions. Directional resistance index Asymmetric indicators of positive and negative responses Estimated amount of deployment Calculate the confidence level Growth Maintenance Complete the submission or rollback.

[0105] In segmented feeding scenarios, the role of mutual exclusion fault typing is to map different causes to different handling paths. If the controller ignores the typing confidence level... If the degree of dominance is determined by repeated clearing under any insufficient condition, it may pull the actual material shortage or idling into the clearing cycle, or amplify the weighing jump with mechanical action in the event of a metering chain failure, thereby occupying the maximum allowable time. This also disrupts the connection between process segments. Therefore, step four needs to organize the authorization of disposal, execution of actions, verification of evidence, and budget convergence into a single-chain closed loop, so that each action can be intercepted or released by the evidence gate.

[0106] The controller reads the mutual exclusion fault type and type confidence level from the material feeding transaction record. With evidence snapshots, and with fractal confidence levels The controller determines the authorization level for handling; after the authorization level is determined, the controller selects the action sequence according to the classification and executes the review chain after each round of actions; if the review chain meets the submission conditions, the material feeding transaction is submitted and allowed to enter the next process segment; if the review chain does not meet the conditions and the budget is reached, the material feeding transaction is rolled back and the process is downgraded to stop loss.

[0107] First, set the confidence level of the fractal. As an automatic processing permission gate, this permission gate is then bound to a hard budget boundary, ensuring that actions have clear boundaries in both the executable and haltable dimensions. The controller will then assign fractal confidence levels. The mapping is divided into three levels of authorization: automatic handling, conservative handling, and stop-loss handling. Automatic handling allows the writing of obstacle clearing or rhythm reconstruction actions, conservative handling only allows pause and review, and stop-loss handling directly enters the degradation path of sub-step 402.

[0108] To prevent authorization from existing but budget from being invisibly consumed, the controller treats each action as a retry and deducts the maximum number of retries. The cumulative duration of the action is included in the maximum allowed duration. And the estimated deployment amount formed in step two The incremental increase is the accumulated budget usage; once the budget usage reaches the cumulative spending limit... The controller prohibits the addition of new materials, allowing only actions that do not introduce new material input, such as releasing back pressure or extending the pause. Therefore, the authorization for disposal and budget allocation are recorded in the material input transaction log.

[0109] Among them, the controller uses fractal confidence level Generate authorization levels and retry the maximum number of times. Maximum allowed duration Cumulative disbursement limit Restrict writing and execution of actions in the action queue, and record the authorization status and budget usage in the material feeding transaction record. Define the authorization threshold as an automatic authorization threshold. With conservative authorization threshold ,satisfy Authorization levels are determined by classification confidence level. The piecewise function is determined as follows: If Authorization is set to automatic processing; if Authorization is to handle the matter conservatively; if Authorization to handle losses.

[0110] In practice, the authorization level is hard-bounded with the budget, ensuring that actions are paused for review rather than entering a clearing loop when evidence is insufficient, and that the budget allocation is based on the estimated allocation volume. The incremental measurement ensures that supplementary contributions are not allowed to be made in accordance with verifiable evidence.

[0111] When authorized, an action sequence is generated based on mutually exclusive fault classifications, requiring that the action parameters converge as evidence falls back, avoiding secondary accumulation caused by fixed reversals. For high-resistance blockage or jamming classifications, the controller uses a cycle of short reverse pulse-pause-short forward pulse, and re-triggers the bidirectional micro-motion detection in step one after each cycle, so that the directional resistance index... Directional resistance index Asymmetric indicators of positive and negative responses New evidence of resistance is formed; when the evidence of resistance falls and the estimated amount of injection... If the signal is still insufficient, the controller resumes the short coding cycle from step two for supplementary verification, and maintains the growth rate. Whether to continue supplementary feeding is restricted. For bridging or intermittent feeding types, the controller prioritizes cycle time reconstruction: shortening the running segment and lengthening the pause segment, causing the material outlet to drop during the pause segment, and then calculating the confidence level. Determine whether the weighing steps are back in phase with the beat.

[0112] To map the resistance evidence into executable parameters, the controller calculates the obstacle clearance intensity coefficient. And piecewise linear interpolation was used to adjust the obstacle clearance intensity coefficient. The mapping is to the inverse pulse width and the pause duration; the interpolation node is determined by the material type identifier, and the slope required for interpolation is obtained through three-point difference.

[0113] If the beat reconstruction needs to be selected from a set of candidate pause proportions, the controller uses a binary search to select the growth preservation degree. Increase and solve confidence level The selected pause ratio is recorded in the feeding transaction log without decreasing. Short reverse pulses can be achieved by reverse motor drive or valve body suction action.

[0114]

[0115] Where: Clearance intensity coefficient Clearance intensity coefficient Used to map the reverse pulse width and pause duration; directional resistance index Positive resistance indicator, non-negative real number; directional resistance indicator. : Reverse resistance indicator, non-negative real number; fractal confidence level The degree of evidence dominance is: Used to limit the force of obstacle removal; prevents zero offset Numerical stability constant, a positive real number, used to prevent the denominator from approaching zero; For each type of material, identify a predefined set of breakpoints. And preset the set of reverse pulse widths corresponding to the breakpoint. Pause duration set positive pulse width set and the set of repetitions .when At that time, the reverse pulse width is calculated using piecewise linear interpolation:

[0116] Pause duration With positive pulse width Interpolate using the same formula; the number of repetitions can be taken from the nearest breakpoint value or the interpolated value rounded up and subject to the maximum number of retries. Limitations. Reverse pulse width. Pause duration Positive pulse width Action timing parameters, with values ​​ranging from positive real numbers. Breakpoints Preset strength breakpoints, with a value range of [value range missing]. Interpolation segment index The range of values ​​is Integers.

[0117] As an example: the solid seasoning channel was classified as either bridging or intermittent feeding, and the classification confidence level was... When the screw is in a dominant position, the controller lengthens the cycle pause. During this pause, the screw pushes the powder towards the discharge port, and during the pause, the powder falls back, creating a gap. The weighing step descent changes from intermittent to continuous. The controller reads and calculates the confidence level. Rebound and confirm growth sustainability The process transitions from a swing-back pattern to a unidirectional accumulation pattern, and then proceeds to the next step of submission for review.

[0118] The controller generates a clearing or cycle reconstruction sequence based on the mutually exclusive fault classification, and uses the clearing intensity coefficient as the basis for the sequence. Motion parameters are determined by combining piecewise linear interpolation, three-point difference, and binary search, while the directional resistance index is reused after each round of motion. Directional resistance index Asymmetric indicators of positive and negative responses Estimated amount of deployment Calculate the confidence level Growth Maintenance The action converges and invalid loops are terminated. A pre-defined set of candidate pause proportions, ordered in ascending order of pause percentage, is used. For each candidate pause ratio The controller performs a short-cycle trial run and calculates the estimated release quantity. Calculate the confidence level With growth Then calculate the submission potential function. The controller selects to submit the potential function. Minimum candidate pause ratio As a new beat parameter. If the candidate set is large and empirically indicates... Follow If the result is approximately monotonic, a binary search can be used on the candidate set index to reduce the number of trials.

[0119] In use, the motion parameters are derived from resistance evidence and classification confidence levels. Common constraints allow the clearance effort to naturally decrease as evidence diminishes, preventing secondary accumulation caused by fixed actions. Beat reconstruction is used to calculate confidence levels. With growth The criteria for judgment are to make the proceedings revolve around evidence of the restoration of cumulative growth.

[0120] For segments where further clearing is not advisable, a stop-loss path is provided, ensuring that the stop-loss path remains within the material feeding transaction boundary. For segments with genuine material shortages or idle operations, the controller stops the channel and freezes replenishment authorizations, limiting the cumulative feeding cap. The material is not further consumed, and a replenishment notification is written to the feeding transaction record. If a spare channel for the same material exists, the controller switches the channel identifier and re-executes step one (gating) to adjust the new directional resistance index. Directional resistance index Weight health indicators After becoming a new snapshot of evidence, proceed to step two. The backup channel can be a backup material tank channel or a dual discharge channel of the same material tank.

[0121] For the fault classification of the execution chain, the controller puts the channel into an isolated state and prohibits entry into the second step of the operation to prevent valve jamming or transmission slippage from causing residual material to be squeezed out and overflow. The controller also records the isolation cause, mutually exclusive fault classification, and classification confidence level. Compared with the estimated amount of deployment at that time Calculate the confidence level Write the material feeding transaction record. For metering chain fault classification, the controller does not perform numerical correction on the weighing signal, but instead switches to a conservative open-loop approach for this segment's feeding: shortening the running segment and lengthening the pause segment, and forcibly verifying the weighing health indicators after each running segment. If within the maximum allowed duration Internal weighing health index If it is still unavailable, roll back the material feeding transaction and request manual intervention.

[0122] The controller performs shutdown and loss mitigation, channel isolation and conservative open-loop control respectively for three types of failures: actual material shortage or idling, execution chain failure and metering chain failure. It also fixes the status field and evidence snapshot in the material feeding transaction record. If necessary, it switches to the backup channel and re-enters step one gate control.

[0123] When in use, stop-loss and isolation prevent unsuitable cleaning segments from consuming the cleaning budget, avoid repeated operations due to material shortages or mechanical failures, and conservatively use open-loop weighing to assess health indicators. The availability of the signal is used as a verification condition, so that the handling of metering chain faults does not depend on the numerical correction of the weighing signal.

[0124] A unified submission review criterion is provided, ensuring that the execution of an action does not necessarily mean permission to proceed to the next process stage. The controller calculates the submission potential function after each round of processing. Submit the potential function Directional resistance indicator Directional resistance index Resistance has fallen, estimated release volume The target interval is achieved, and the confidence level is calculated. With growth The threshold is achieved, converting to the same scalar cost; the soft threshold is constructed using logarithms and exponentials, and engineering implementation employs table lookup and piecewise linear interpolation. Table lookup can be implemented using read-only memory or piecewise polynomial approximation. When the potential function is submitted... Submit the material feeding transaction and clear the current segment's budget usage when the preset submission threshold is met and the budget has not been reached; when the potential function is submitted... The preset submission threshold is not met and the maximum number of retries has been reached. Or maximum allowed duration or cumulative upper limit Roll back the material feeding transaction and classify the rollback reason, mutual exclusion fault type, and type confidence level. Directional resistance index Directional resistance index Estimated amount of deployment Calculate the confidence level Growth Maintenance Write this into the material feeding transaction record for reference in subsequent process stages:

[0125] Where: Submitted potential function : Verification value, a non-negative real number, used for commit or rollback determination; can be a commit resistance threshold. : Upper resistance threshold, a non-negative real number, determined by the material type identifier; directional resistance index : Positive resistance indicator, non-negative real number; Directional resistance index : Reverse resistance indicator, a non-negative real number; solve for the confidence threshold. : Calculate the lower confidence limit, ; Solving for confidence : Channel solution confidence level, is Growth maintenance threshold : The lower limit of stable growth Growth retention Growth stability indicators Target delivery volume The target delivery amount for this segment is a non-negative real number; allowable error. The target interval half-width, whose values ​​are non-negative real numbers and satisfy the following conditions: Estimated amount of delivery : Estimated delivery volume for the channel, with values ​​ranging from non-negative real numbers; The controller submits the potential function. Unified review of directional resistance indicators Directional resistance index Estimated amount of deployment Calculate the confidence level Growth Maintenance The decision and evidence snapshot will be written into the material feeding transaction record.

[0126] When using it, submit the potential function. By merging multi-condition reviews into scalar criteria, submission and rollback are controlled by the same evidence gate. The soft threshold form reduces repeated switching near the boundary, making it easier for reviews to converge to submission or rollback. The rollback record fields are fixed, ensuring consistent input for maintenance positioning and subsequent process segment strategy references.

[0127] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0128] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0129] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0130] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0131] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A cooking process-based control method for a stir-fry machine, characterized in that: comprising, establishing a feeding transaction and generating a target feeding amount and a channel and material type identifier; applying a bidirectional micro-motion detection sequence to the channel to collect motor current and rotation speed, calculate equivalent resistance indicators, and positive and negative response asymmetry indicators, and evaluate the health of the weighing signal to obtain a gating conclusion; based on the gating conclusion, assigning a short coding beat to the channel and encoding a pulse feeding, collecting a single weighing signal to obtain a weight change sequence, synchronously correlating the short coding beat and the weight change sequence to obtain a feeding amount estimate and a solving confidence, and stopping or supplementing feeding accordingly; when the feeding amount estimate is stagnant or the solving confidence is below a threshold, performing mutual exclusion fault typing based on the equivalent resistance indicators, the positive and negative response asymmetry indicators, the feeding amount estimate growth trend, and the current and rotation speed consistency indicators, and outputting a typing confidence; performing self-recovery or stop-loss according to the mutual exclusion fault typing and the typing confidence; reusing the bidirectional micro-motion detection sequence and synchronous correlation solving; if the equivalent resistance indicators fall back and the feeding amount estimate grows and the solving confidence meets the standard, the feeding transaction is submitted, otherwise the log is recorded.

2. The segmented feeding control method of the cooking process-based wok machine according to claim 1, wherein: when establishing a feeding transaction, generating a target feeding amount, an allowable error, a maximum allowable time, a maximum number of retries, and a cumulative feeding upper limit, and executing a bidirectional micro-motion detection sequence according to forward short-acting driving, pause, reverse short-acting driving, pause, forward short-acting driving, and outputting a gating conclusion and a risk type.

3. The segmented feeding control method of the cooking process-based wok machine according to claim 2, wherein: the equivalent resistance indicators are calculated by integrating the response sequence of the motor current and rotation speed within the detection window, the positive and negative response asymmetry indicators are calculated by taking the ratio of the positive and negative equivalent resistance indicators and performing logarithmic transformation, and the controller uses the equivalent resistance indicators and the positive and negative response asymmetry indicators as the basis for determining the risk type.

4. The segmented feeding control method of the cooking process-based wok machine according to claim 3, wherein: the weighing signal health assessment includes calculating the change rate of the short window weighing sequence, combining the endpoint difference and the window fluctuation ratio, detecting saturation and jump, outputting a gating conclusion of the metering chain being available or unavailable, and writing the gating conclusion into the feeding transaction.

5. The segmented feeding control method of the cooking process-based wok machine according to claim 4, wherein: the short coding beat is selected from a candidate beat table and bound to the channel identifier, and a cross-correlation upper limit threshold is applied to the short coding beats of different channels during selection; the encoding pulse feeding drives the actuator according to the short coding beat switching between the running segment and the pause segment, and records the short coding beat identifier for synchronous correlation solving.

6. The segmented feeding control method of the cooking process-based wok machine according to claim 5, wherein: The weight change sequence is formed by taking the end-point difference of the single weighing signal at the beat unit boundary, the end-point difference is the difference between the start end-point and the end end-point of the beat unit, and the drop direction is unified as a positive direction according to the arrangement direction of the weighing sensor; the controller takes the median value of the weighing sample value at the end of the beat unit to generate the weight change sequence.

7. The cooking process-based segmented feeding control method of the wok machine according to claim 6, characterized in that: The solving confidence is calculated by normalizing the beat-by-beat contribution weight of the short coding beat and the weight change sequence, the contribution weight is determined based on the weight change amplitude after beat modulation, the solving confidence is output in an information entropy normalization manner, and is used by the controller to determine stopping feeding, micro-supplement feeding or entering the subsequent step.

8. The cooking process-based segmented feeding control method of the wok machine according to claim 7, characterized in that: The mutually exclusive fault typing includes high-resistance blockage or sticking, bridging or intermittent feeding, real material shortage or idling, execution chain fault and metering chain fault; the input of the mutually exclusive fault typing includes the equivalent resistance index, the positive and negative response asymmetry index, the estimated feeding amount, the solving confidence and the consistency index, and outputs the typing confidence.

9. The cooking process-based segmented feeding control method of the wok machine according to claim 8, characterized in that: The consistency index is obtained by respectively aligning the motor current and speed collected in the running section and the pause section of the coding pulse feeding, calculating the current response amplitude and speed response amplitude, and obtaining the mismatch degree; step three uses the consistency index and the equivalent resistance index to distinguish the execution chain fault and the high-resistance blockage or sticking.

10. The cooking process-based segmented feeding control method of the wok machine according to claim 9, characterized in that: The controller divides the disposal authorization into automatic disposal, conservative disposal and loss disposal according to the typing confidence, and limits the action sequence with the maximum retry number, the maximum allowed time and the cumulative feeding upper limit; when the submission condition is met, the feeding transaction is submitted, when the submission condition is not met and any limit is reached, the feeding transaction is rolled back and the log is recorded.

Images