A high-efficiency pretreatment method before solid-liquid separation of vinegar dregs

Abstract

The present application relates to the technical field of data processing, and especially to a kind of efficient pretreatment method before solid-liquid separation of vinegar dregs of fermented vinegar. Including: distributed sensor acquisition attribute constructs coupling feature vector, after dynamic weighting by information entropy, match optimal trajectory to anchor initial state;In virtual simulation environment, take this as starting point, with time length and target state as hard constraint, construct functional containing energy consumption and shear penalty and introduce dynamic threshold, optimize to generate power and static sequence and ideal envelope surface;When executing, according to real-time state and envelope surface error, adopt Lyapunov control law to calculate feedforward compensation, if there is shear out-of-limit risk, limit power to within safety threshold, exceed energy into heat energy or extend static compensation. The present application improves the adaptability to batch material state change, effectively reduces the separation quality loss caused by excessive shear.

Description

Technical Field

[0001] This invention relates to the field of data processing technology. More specifically, this invention relates to an efficient pretreatment method for vinegar mash before solid-liquid separation. Background Technology

[0002] In vinegar brewing, the efficiency of solid-liquid separation of vinegar mash is highly dependent on the material control effect in the pretreatment stage. Existing processes mostly rely on operational experience and adopt a fixed combination of stirring and settling, which is difficult to adapt to changes in the initial physicochemical properties caused by batch differences in raw materials, uneven fermentation, and fluctuations in ambient temperature. This open-loop control strategy cannot match the appropriate mechanical action process for different batches of materials within a fixed production cycle, resulting in increased pressure filtration resistance and unstable juice yield in the subsequent process.

[0003] In terms of condition monitoring, existing technologies mostly use single-point or a small number of instruments to obtain macroscopic parameters, which cannot accurately characterize the overall state of heterogeneous multiphase substances such as vinegar mash. Initial state determination is prone to introducing random errors, resulting in distortion of subsequent control boundaries.

[0004] Because vinegar mash is a thixotropic material, its rheological properties dynamically evolve with mechanical action. Existing fixed threshold protection mechanisms lack the ability to perceive and proactively avoid real-time changes in safety boundaries. This often leads to prolonged processing time due to conservative control, or damage to the microstructure of the vinegar mash due to localized over-shearing, resulting in the loss of beneficial flavor compounds and affecting product quality consistency. Therefore, there is an urgent need for a pretreatment method that can accurately sense the initial state of the material, dynamically match safety thresholds, and adaptively regulate the mechanical action process. Summary of the Invention

[0005] To address the technical challenge of balancing damage risk and processing efficiency in thixotropic materials like vinegar mash, where the rheological properties dynamically evolve with mechanical action, and where simple fixed threshold protection mechanisms are insufficient, this invention provides solutions in the following aspects.

[0006] In a first aspect, the present invention provides a highly efficient pretreatment method for vinegar mash before solid-liquid separation, the method comprising: The biochemical and physical properties of the vinegar mash to be processed are collected synchronously by a distributed sensor array, and an intrinsic physicochemical coupled feature vector is constructed. After weighting each dimension of the feature vector based on the local information entropy dynamic weight model, the optimal evolution trajectory is matched in the historical multidimensional physicochemical evolution database to complete the initial state anchoring. In the virtual simulation environment, taking the intrinsic physicochemical coupling feature vector as the evolution starting point, and using the fixed total preprocessing time and the final target separation state as hard constraints, a target loss functional containing mechanical energy consumption and shear failure penalty terms is constructed. A dynamic transient shear failure threshold is introduced as a barrier, and optimization is performed through a sequential quadratic programming solver to generate a planning execution sequence of mechanical power and resting time and an ideal state evolution envelope. During actual operation, based on the error between the real-time perceived state and the envelope of the ideal state evolution, the Lyapunov control law is driven to calculate the feedforward compensation energy and determine whether the required mechanical compensation energy triggers the risk of exceeding the transient shear failure threshold. If there is a risk of exceeding the threshold, the mutual exclusion backoff logic is triggered: the actual mechanical power command is limited to below the safety threshold, and the excess energy demand is converted into heat energy input and / or the resting time is extended for equivalent compensation, so that the state converges to the target trajectory.

[0007] Preferably, the distributed sensor array includes a pH sensor, a temperature sensor, a density sensor, and a near-infrared spectroscopy sensor deployed in the material conveying pipeline and buffer silo.

[0008] Preferably, the biochemical properties include pH value and starch residue value; the physical properties include initial bulk density value and initial ambient temperature value.

[0009] Preferably, the local information entropy dynamic weight model includes: dynamically allocating weight coefficients by calculating the degree of information entropy fluctuation of each feature dimension in historical data, so as to eliminate the scale difference of features with different dimensions in distance measurement.

[0010] Preferably, the fixed total preprocessing time and the separation state of the final target are hard boundary conditions for trajectory optimization in the virtual simulation environment.

[0011] Preferably, the target loss functional includes a penalty term for the accumulation of mechanical energy consumption and a penalty term for the predicted shear rate exceeding the transient shear failure threshold.

[0012] Preferably, the transient shear failure threshold is a variable that is dynamically calculated and updated based on the real-time evolution of the material in the virtual simulation environment.

[0013] Preferably, in step three, the real-time sensing status is obtained by collecting torque feedback data from the stirring motor and internal temperature data of the material and then fusing and reconstructing them.

[0014] Preferably, the mutual exclusion avoidance logic includes: limiting the actual mechanical power command to below a safety threshold, and converting the excess energy demand into equivalent compensation for the thermal energy input of the material and / or the extension of the resting time by converting it into a thermal energy input and / or the extension of the resting time through the thermo-mechanical efficiency conversion coefficient.

[0015] Preferably, the heat input is achieved by controlling the opening degree of the steam valve in the equipment jacket.

[0016] The embodiments of the present invention have at least the following beneficial effects: 1. By synchronously collecting biochemical and physical multidimensional attributes through a distributed sensor array, an intrinsic physicochemical coupled feature vector is constructed. A local information entropy dynamic weight model is introduced to automatically assign weights based on the degree of information fluctuation of each feature dimension in historical data, eliminating scale differences in distance measurement between different dimensions. The weighted vector is then used to match the optimal evolutionary trajectory in the historical database to complete the initial state anchoring. This extends the initial state determination from a single macroscopic parameter to a high-dimensional space, fundamentally preventing the risk of subsequent control mismatch caused by initial state distortion and solving the state perception deviation problem caused by batch-to-batch nonlinear fluctuations.

[0017] 2. In a virtual simulation environment, starting from an anchored initial state and with fixed total preprocessing time and target separation state as hard constraints, a target loss functional is constructed, incorporating mechanical energy consumption and shear failure penalties. A transient shear failure threshold, dynamically updated based on real-time material density and temperature, is introduced as a safety barrier. Through optimization, a planned execution sequence of mechanical power and settling time, along with an ideal state evolution envelope, is generated. This ensures the planned trajectory always satisfies the material's rheological tolerance boundary, guaranteeing the conditions required for subsequent solid-liquid separation while avoiding microstructural damage and flavor loss due to over-shearing. This achieves synergistic optimization of processing efficiency and quality stability within a fixed cycle time.

[0018] 3. During actual operation, based on the error between the real-time perceived state and the ideal state evolution envelope, the Lyapunov control law is driven to calculate the feedforward compensation energy, and it is determined in real time whether the compensation energy will trigger the transient shear failure threshold exceeding the limit. Once there is a risk of exceeding the limit, the mutual exclusion backoff logic is triggered to limit the actual mechanical power command below the safety threshold, while the excess energy demand is converted into heat energy input and / or the resting time is extended for equivalent compensation. Thus, under the strict constraint of fixed total time, the forced merging of the state trajectory is achieved by equivalent heat-mechanical energy substitution and resting relaxation, avoiding the engineering risks of breaking through the shear thickening threshold due to blindly increasing mechanical power, and ensuring the consistency and robustness of batch processing results. Attached Figure Description

[0019] Figure 1 The flowchart illustrates the steps of the efficient pretreatment method for solid-liquid separation of vinegar mash in this 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, not all, of the embodiments of the present invention. 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] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0022] S1: The biochemical and physical properties of the vinegar mash to be processed are collected synchronously through a distributed sensor array to construct an intrinsic physicochemical coupled feature vector. After weighting each dimension of the feature vector based on the local information entropy dynamic weight model, the optimal evolution trajectory is matched in the historical multidimensional physicochemical evolution database to complete the initial state anchoring.

[0023] Before entering the solid-liquid separation process, the internal biochemical reaction process and macroscopic physical state of the vinegar mash to be processed exhibit significant batch-to-batch nonlinear fluctuations.

[0024] Since data obtained from a single or a few monitoring points cannot represent the non-uniformity of material distribution in space, initial state determination based on such low-dimensional, single-point data will introduce random errors, resulting in distortion when providing control boundaries for subsequent control.

[0025] Therefore, a distributed sensor array was pre-deployed in the material conveying pipeline and buffer silo. This array includes acid and alkalinity sensors, temperature sensors, density sensors, and near-infrared spectroscopy sensors to collect synchronous data on the material in a spatially comprehensive manner.

[0026] The sensor array synchronously collects pH values ​​that reflect the fermentation maturity of the materials. Starch residue values ​​were obtained by peak detection using a near-infrared spectroscopy sensor. And the initial bulk density value reflecting the material state is collected by a density sensor. ,unit The initial ambient temperature value obtained by the temperature sensor. , in °C.

[0027] The four types of parameters collected together constitute the basic dataset describing the intrinsic state of the current batch of vinegar mash. This dataset is organized into a four-dimensional column vector, namely the intrinsic physicochemical coupling feature vector. .

[0028] Due to the composition The various dimensions of characteristics have different physical meanings and dimensions, and their numerical ranges vary greatly, for example... Typically in the hundreds, while It is usually in the single digits.

[0029] If the original feature vectors are used directly for similarity measurement, high-amplitude features will completely dominate the distance calculation results, causing low-amplitude but potentially informative features to be overlooked.

[0030] To eliminate the interference of differences in dimensions and scales on state matching, a local information entropy dynamic weight model is used to weight each dimension of the feature vector.

[0031] Local information entropy Used to measure the The degree of fluctuation in the values ​​of each feature dimension across numerous batches of data recorded in the historical database; the greater the fluctuation, the higher its information entropy. The larger the value.

[0032] According to each dimension Calculate its dynamic weights The calculation formula is: .

[0033] in, For the first The local information entropy of each feature dimension is calculated using a well-known technique, which will not be elaborated here. The empirical smoothing coefficient is a hyperparameter used to prevent excessive weight allocation to a feature with extremely low entropy. Its empirical value is [value missing]. It is dimensionless and can be adjusted by the implementer according to the specific implementation scenario. The number 4 corresponds to four dimensions.

[0034] This results in higher weights for features with lower information entropy, i.e., those that are relatively stable across different batches, while the weights for features that fluctuate wildly are reduced, thus highlighting those dimensions that can more robustly distinguish the state of materials in the distance metric.

[0035] The weighted feature vector As the query value for the current batch, where It is determined by the weights of each dimension. The constructed diagonal matrix is ​​the diagonal matrix. Multiply by column vector ,get It means assigning weights to each dimension. Multiply each of these values ​​by the corresponding eigenvalues.

[0036] In a pre-stored historical multidimensional database, the weighted Mahalanobis distance is used. The distance metric is used to perform the search in order to obtain the covariance relationship between different feature dimensions.

[0037] Calculate the weighted Mahalanobis distance between the current state and the starting state of each historical trajectory in the database. The weights for different dimensions of data are used in calculating the weighted Mahalanobis distance. The value is determined, and the historical trajectory with the smallest distance is selected as the optimal matching evolution trajectory. The Mahalanobis distance is a well-known technical content and will not be elaborated further.

[0038] Confidence of matching results From the formula Calculation, where The optimal historical feature vector that was matched.

[0039] when When the value exceeds a preset threshold, the initial state is determined to be successfully anchored, and the subsequent evolution sequence of the optimal historical trajectory is used as a reliable reference to output an anchoring completion signal. The preset threshold is 0.7, which can be adjusted by the implementer according to the specific implementation scenario.

[0040] S2: In the virtual simulation environment, the intrinsic physical-chemical coupling eigenvector is used as the starting point for evolution. With a fixed total preprocessing time and the final target separation state as hard constraints, a target loss functional containing mechanical energy consumption and shear failure penalty terms is constructed. A dynamic transient shear failure threshold is introduced as a barrier. The optimization is performed through a sequential quadratic programming solver to generate the planning execution sequence of mechanical power and resting time and the ideal state evolution envelope.

[0041] The virtual simulation environment receives the initial state anchoring signal and the intrinsic physicochemical coupling eigenvector. .

[0042] Because the pretreatment process must be embedded in the fixed production cycle at the workshop level, a total pretreatment time has been set. Its empirical value is 2700 seconds, which can be adjusted by the implementer according to the specific implementation scenario.

[0043] Simultaneously, a pre-defined final target separation state vector is used. This vector represents the rheological critical point that the material should reach before solid-liquid separation, including target values ​​such as specific viscosity and free liquid phase saturation.

[0044] The virtual evolution trajectory within the virtual simulation environment must be based on Starting from, and in Always satisfied The terminal hard constraint means that any trajectory that does not satisfy this dual constraint of time and state is considered invalid.

[0045] Within a virtual simulation environment, a nonlinear differential equation describing the evolution of the material's state is established: ,in, For a moment The applied mechanical power, measured in watts (W); For a moment The working condition switching function is defined as follows: 0 represents the mechanical work state, and 1 represents the static relaxation state without mechanical input. The unit is seconds (s). The function f characterizes the dynamic law of the continuous evolution of the material state under two discrete working conditions: mechanical work and static relaxation.

[0046] The function f is implemented using a data-driven radial basis function neural network model. Specifically, this surrogate model uses the state vectors recorded at each time point in the historical production batches. Mechanical power In addition, operating condition indicators are used as input features to complete a control cycle. Rate of change of state after The output labels are obtained through offline training and fitting.

[0047] In a virtual simulation environment, the sequential quadratic programming solver can forward deduce the state evolution trajectory under any control input sequence without relying on explicit mathematical analytical expressions by repeatedly calling the trained and converged surrogate model, thus providing a computational basis for optimization. The modeling method of using neural networks for state prediction is a well-known technology in this field, and the specific network topology will not be described here.

[0048] To find the optimal control trajectory that consumes little energy and meets the safety boundary, a target loss functional J is constructed to evaluate the trajectory. Its expression is as follows: in, The target loss functional is dimensionless. This is a mechanical energy consumption penalty term, used to penalize excessive energy input. (Parameters...) This is the energy consumption weighting coefficient, empirically set to 0.01. , is a hyperparameter.

[0049] This is a penalty for shearing damage.

[0050] To determine based on the current state and control input Predicted transient shear rate, in units of per second ( ).

[0051] This is the dynamic transient shear failure threshold, also measured in units of... This threshold varies with the material state. Changes in real time.

[0052] function The value is 0 when the predicted shear rate does not exceed the threshold, and positive when it does.

[0053] parameter The shear damage penalty weighting coefficient is empirically set to a value of [value missing]. is a very large constant hyperparameter designed to severely punish any behavior that exceeds the threshold.

[0054] Terminal Item This is a penalty function for the terminal state, ensuring that the final state approximates the target.

[0055] Dynamic transient shear failure threshold The calculation depends on the real-time status of the material.

[0056] According to the material rheology model, A function established as a state vector. ,in and State vector The corresponding components in The model coefficients, calibrated based on the rheological properties of the material, are hyperparameters, where... Values ​​are taken from The unit is , Values ​​are taken from The unit is , Values ​​are taken from The unit is The hyperparameter empirical values ​​can be adjusted by the implementer according to the specific implementation scenario.

[0057] This model makes the safety threshold decrease as the material density increases and may increase appropriately as the temperature rises, thus dynamically reflecting the current physical limits that the material can withstand.

[0058] The above-mentioned constrained optimal control problem is solved using a sequential quadratic programming solver, where sequential quadratic programming is a well-known technique and will not be described in detail here.

[0059] The solver is Minimize as the objective, while satisfying the state differential equation and time boundary conditions. and terminal status Iterative calculations are performed under hard constraints.

[0060] After the optimization converges, the solver outputs the program execution sequence. ,in For discrete time nodes, The sequence length is given.

[0061] This sequence specifies the mechanical power to be applied and the resting time at each moment from start to finish.

[0062] Simultaneously, output the state evolution sequence corresponding to the optimal trajectory. , which serves as the ideal state evolution envelope for the entire preprocessing cycle.

[0063] S3: In the actual operation phase, based on the error between the real-time perceived state and the envelope of the ideal state evolution, the Lyapunov control law is driven to calculate the feedforward compensation energy and determine whether the required mechanical compensation energy triggers the risk of exceeding the transient shear failure threshold.

[0064] In the actual operation phase, obtain the planned execution sequence. Evolution envelope of ideal state .

[0065] Considering the response lag of the actual transmission mechanism and the uneven mixing of materials, the actual running trajectory will deviate from the planned trajectory.

[0066] Torque feedback data is acquired in real time at a frequency of 1 Hz using a torque sensor array deployed on the mixing shaft and a temperature sensor array embedded inside the material. With multi-point temperature data The units are Newton-meters (N·m) and degrees Celsius (°C).

[0067] Based on this sensing data, the real-time state vector is reconstructed using a pre-calibrated state observer model. Specifically, the state observer model is a nonlinear mapping model based on a multilayer perceptron, which uses real-time torque feedback data. With temperature data In addition to the online near-infrared spectral feature values ​​and in-situ pH sensor data deployed in the mixing chamber as inputs, the system is pre-trained and calibrated based on the mapping relationship between sensor input data accumulated in historical production batches and physicochemical states obtained from offline testing. This allows for online real-time estimation and output of the current bulk density, temperature, pH, and starch residue values, reconstructing a four-dimensional real-time state vector consistent with the planned spatial dimension. Among them, the multilayer perceptron is a well-known technology and will not be described in detail. Implementers can replace it with other state observer models according to the specific implementation scenario.

[0068] Calculate the current time Real-time state vector With ideal envelope The error between the expected state vectors at the same time. The calculation formula is: .

[0069] in, The real-time state error vector has the same dimension as the intrinsic eigenvector. same; This is a real-time state vector reconstructed from sensor data; The ideal state evolution envelope surface output in step two at time... The expected state vector.

[0070] The error vector The deviation of the actual running trajectory from the planned trajectory was calculated.

[0071] To drive the real-time state to converge to the planned trajectory, a control law based on Lyapunov stability is used to calculate the required feedforward compensation energy. Lyapunov stability theory is a well-known technique and will not be elaborated further.

[0072] Select a positive definite Lyapunov function candidate ,in Let be a positive definite diagonal weight matrix, which is a hyperparameter, and its empirical value is the identity matrix.

[0073] To achieve asymptotic stability, the following conditions must be met: Based on this, the controller calculates a theoretical total power demand command. To achieve the mapping from multidimensional physical state errors to a single scalar power command, a dimensionless all-1 row vector is introduced. Its expression is: in, Total power demand in real time (unit: W); The baseline mechanical power (in W) in the planning sequence; for Positive definite diagonal proportional gain matrix, with diagonal elements corresponding to the proportional compensation coefficients of each state dimension; for Positive definite diagonal nonlinear gain matrix; for The real-time state error column vector is generated by multiplying the gain matrix and the error vector. The power compensation column vector is then multiplied by the row vector M to complete the scalar summation of the multidimensional compensation terms, thus enabling the formula to achieve a strict closed loop in terms of dimension and scale.

[0074] This is a positive definite diagonal proportional gain matrix, where each diagonal element is an independent constant, empirically taken as 50. The unit of each element is determined according to the physical dimensions of the corresponding error component; for example, the gain unit corresponding to the packing density error component is... The gain unit corresponding to the temperature error component is The gain unit corresponding to the dimensionless error component is W, thus ensuring that the linear compensation term is uniformly dimensional in watts.

[0075] This is a diagonal nonlinear gain matrix, where each diagonal element is an independent constant, empirically taken as 30. The units of each element are determined based on the physical dimensions of the corresponding error component. The powers are determined separately, for example, the gain unit corresponding to the packing density error component is... The gain unit corresponding to the temperature error component is The gain unit corresponding to the dimensionless error component is W; It is a non-linear exponent, and its empirical value is [value missing]. Dimensionless; For symbolic functions, It is an absolute value function.

[0076] The item provides linear error compensation. This term is used to deal with larger nonlinear deviations.

[0077] It is necessary to determine Will using all of the energy input as mechanical compensation cause transient damage to the material structure?

[0078] Therefore, the controller invokes the same rheological model as in step S2, based on the current real-time state. Calculate the current transient shear failure threshold And predict if power is applied The resulting transient shear rate .

[0079] The judgment logic is: if If it does, there is a risk of triggering a transient shear failure threshold exceeding the limit; otherwise, there is no risk.

[0080] and The unit for all values ​​is per second.

[0081] When a risk of exceeding limits is identified, direct issuance should be avoided. This causes the material structure to break, triggering the mutual exclusion backoff logic: The logic first determines a safe mechanical power threshold. The threshold is satisfied The maximum mechanical power under certain conditions can be obtained by inverse rheological modeling, and the unit is watt (W). Inverse rheological modeling is a well-known technique and will not be described in detail here.

[0082] Subsequently, the actual mechanical power command issued to the mixing motor will be... The restriction is set below this safety threshold, specifically through a truncation function: ,in, To find the minimum value function, the mechanical actuator should always operate within the safe boundary of the material.

[0083] because After being restricted, there is a gap in the theoretical compensation energy. The unit is watt (W).

[0084] An equivalent compensation is needed to maintain the total energy input so that the state converges.

[0085] Therefore, the excess energy demand can be converted into one or two of the following equivalent forms of compensation.

[0086] The first method is to convert it into heat energy input, based on the heat engine efficiency conversion coefficient. Calculate the required additional heat energy ,in The heat engine efficiency conversion coefficient is empirically taken as [value missing]. , dimensionless, is a hyperparameter that represents the conversion ratio of equivalent thermal energy to unit mechanical energy deficit; To control the cycle, the empirical value is taken as follows: Seconds (s), where hyperparameters can be adjusted by the implementer according to the specific implementation scenario.

[0087] The unit is joule (J), and this heat is injected by adjusting the opening of the steam valve in the equipment jacket.

[0088] The second compensation method is to extend the settling time. Based on the energy-time equivalence model, the additional settling time required is calculated. And add it to the static phase of the current execution sequence.

[0089] By executing the above mutual exclusion and equivalent compensation logic, the state error is compensated by comprehensively utilizing three factors: mechanical power, heat input, and resting time, while avoiding mechanical shear exceeding the limit.

[0090] This makes real-time state within a fixed time window Internally, it can asymptotically converge to the ideal state evolution envelope. The defined target trajectory will eventually be in Approaching the target separation state at all times .

[0091] Therefore, after pretreatment using the above control strategy, the rheological and micro-biochemical properties of the vinegar mash fall precisely within the optimal separation range when entering the subsequent solid-liquid separation process. This invention ensures high efficiency in the pretreatment stage while effectively avoiding local over-shear damage caused by traditional blind stirring, reducing fluid resistance in the subsequent pressure filtration process, and improving the juice yield and flavor retention rate, thus achieving the ultimate goal of improving the overall efficiency of solid-liquid separation in vinegar mash.

[0092] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, specific embodiments have been described above. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps described in the claims can be performed in a different order than that shown in the embodiments and still achieve the desired result. Additionally, the processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0093] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0094] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A highly efficient pretreatment method for vinegar mash before solid-liquid separation, characterized in that, Includes the following steps: The biochemical and physical properties of the vinegar mash to be processed are collected synchronously by a distributed sensor array, and an intrinsic physicochemical coupled feature vector is constructed. After weighting each dimension of the feature vector based on the local information entropy dynamic weight model, the optimal evolution trajectory is matched in the historical multidimensional physicochemical evolution database to complete the initial state anchoring. In the virtual simulation environment, taking the intrinsic physicochemical coupling feature vector as the evolution starting point, and using the fixed total preprocessing time and the final target separation state as hard constraints, a target loss functional containing mechanical energy consumption and shear failure penalty terms is constructed. A dynamic transient shear failure threshold is introduced as a barrier, and optimization is performed through a sequential quadratic programming solver to generate a planning execution sequence of mechanical power and resting time and an ideal state evolution envelope. During actual operation, based on the error between the real-time perceived state and the envelope of the ideal state evolution, the Lyapunov control law is driven to calculate the feedforward compensation energy and determine whether the required mechanical compensation energy triggers the risk of exceeding the transient shear failure threshold. If there is a risk of exceeding the threshold, the mutual exclusion backoff logic is triggered: the actual mechanical power command is limited to below the safety threshold, and the excess energy demand is converted into heat energy input and / or the resting time is extended for equivalent compensation, so that the state converges to the target trajectory.

2. The efficient pretreatment method for solid-liquid separation of vinegar mash according to claim 1, characterized in that, The distributed sensor array includes pH sensors, temperature sensors, density sensors, and near-infrared spectroscopy sensors deployed in material conveying pipelines and buffer silos.

3. The efficient pretreatment method for solid-liquid separation of vinegar mash according to claim 2, characterized in that, The biochemical properties include pH value and starch residue value; the physical properties include initial bulk density value and initial ambient temperature value.

4. The efficient pretreatment method for solid-liquid separation of vinegar mash according to claim 1, characterized in that, The local information entropy dynamic weight model includes: dynamically allocating weight coefficients by calculating the degree of information entropy fluctuation of each feature dimension in historical data, so as to eliminate the scale difference of features with different dimensions in distance measurement.

5. The efficient pretreatment method for solid-liquid separation of vinegar mash according to claim 1, characterized in that, The fixed total preprocessing time and the separation state of the final target are the hard boundary conditions for trajectory optimization in the virtual simulation environment.

6. The efficient pretreatment method for solid-liquid separation of vinegar mash according to claim 5, characterized in that, The target loss functional includes a penalty term for the accumulation of mechanical energy consumption and a penalty term for the predicted shear rate exceeding the transient shear failure threshold.

7. The efficient pretreatment method for solid-liquid separation of vinegar mash according to claim 6, characterized in that, The transient shear failure threshold is a variable that is dynamically calculated and updated based on the real-time evolution of the material in the virtual simulation environment.

8. The efficient pretreatment method for solid-liquid separation of vinegar mash according to claim 1, characterized in that, In step three, the real-time sensing status is obtained by collecting torque feedback data from the stirring motor and internal temperature data of the material and then fusing and reconstructing them.

9. The efficient pretreatment method for solid-liquid separation of vinegar mash according to claim 1, characterized in that, The mutual exclusion avoidance logic includes: limiting the actual mechanical power command to below a safety threshold, and converting the excess energy demand into equivalent compensation for the thermal energy input of the material and / or the extension of the resting time by converting it into a thermal energy input and / or the extension of the resting time through the thermo-mechanical efficiency conversion coefficient.

10. The efficient pretreatment method for solid-liquid separation of vinegar mash according to claim 9, characterized in that, The heat input is achieved by controlling the opening of the steam valve in the equipment jacket.

Images