An automatic control method, device and medium in livestock feed processing process
By generating cross-domain process evolution sequences and constructing cross-domain transfer potential fields, the challenges of cross-domain transfer correlation and multi-objective control in livestock feed processing were solved, achieving more accurate control decisions and improved stability.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- LIAONING MEILONG FEED TECHNOLOGY CO LTD
- Filing Date
- 2026-03-12
- Publication Date
- 2026-06-05
AI Technical Summary
The current livestock feed processing process lacks cross-domain transmission and correlation mechanisms and the ability to make unified judgments on multi-objective control, resulting in regulatory lag, control conflicts, and overall imbalance.
By collecting industrial operation data from multiple work sections and identifying the work sections, a cross-domain process evolution sequence is generated. Multi-dimensional state folding mapping is performed to construct a cross-domain transfer potential field. Traction correlation calculation is executed to obtain cross-domain distribution characteristics, determine the dominant control target and adjustment position, and generate a process control instruction set.
It improves the control continuity, process adaptability and overall operational stability in the livestock feed processing process, and reduces regulation lag and control conflicts.
Smart Images
Figure CN122151775A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of automation control technology, and in particular to an automation control method, equipment and medium for livestock feed processing. Background Technology
[0002] With the development of large-scale farming, refined feeding, and diversified feed formulations, livestock feed processing has gradually evolved from simple mechanical processing to a continuous industrial process encompassing pretreatment, conditioning, extrusion pelleting, and finished product quality control. Early methods primarily relied on setting parameters for individual equipment and manual adjustments based on experience. Typical control targets included steam addition, moisture content adjustment, die cavity pressure control, spindle load monitoring, and finished pellet quality inspection. However, with the introduction of sensors, industrial controllers, and data acquisition technologies, current methods can now collect data online on temperature, humidity, pressure, rotational speed, current, and finished product durability, and achieve a certain degree of automatic adjustment based on rule-based thresholds, local feedback control, or empirical models.
[0003] Existing methods suffer from at least two shortcomings. Firstly, while most control schemes can collect operational signals from pretreatment and forming sections, they typically treat each section's data as parallel monitoring quantities, lacking mechanisms for time-series binding, folding mapping, and cross-domain correlation reconstruction of multi-section industrial operational data. Secondly, existing automated controls often rely on threshold-based adjustments based on single quality, energy consumption, or load indicators, lacking a unified ability to determine the competitive relationships among multiple objectives. This leads to problems such as adjustment lag, control conflicts, or overall imbalance due to local optimization, hindering the achievement of stable forming, load balancing, and efficiency synergy in livestock feed processing. Summary of the Invention
[0004] In view of the aforementioned existing problems, the present invention is proposed.
[0005] Therefore, this invention provides an automated control method for livestock feed processing to solve the problems of difficulty in forming cross-domain transmission associations and difficulty in unifying and coordinating multi-objective control.
[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides an automated control method for livestock feed processing, comprising: collecting multi-segment industrial operation data and processing segment identifiers, and binding the data streams in a time sequence to generate a cross-domain process evolution sequence; performing multi-dimensional state folding mapping on the cross-domain process evolution sequence to extract state action trajectories and forming response trajectories to generate a cross-domain corresponding trajectory group; employing a cross-domain transfer potential field construction algorithm to perform traction correlation calculation on the cross-domain corresponding trajectory group to construct a cross-domain transfer potential field; based on the cross-domain transfer potential field, performing traction main domain aggregation to obtain cross-domain distribution characteristics, and performing sub-domain aggregation to generate a target traction distribution; and based on the target traction distribution and combined with the cross-domain transfer potential field, determining the current dominant control target and corresponding adjustment position to generate a process control instruction set.
[0007] As a preferred embodiment of the automated control method in the livestock feed processing process described in this invention, the multi-stage industrial operation data includes thermodynamic state variables, pressure state variables, actuator load feedback, and finished product quality feedback signals. The processing section identifier includes a pre-processing control field and a forming control field.
[0008] As a preferred embodiment of the automated control method for livestock feed processing described in this invention, the specific steps for generating the cross-domain process evolution sequence are as follows: Record the start and end times of the preprocessing control domain and the forming control domain respectively, obtain the data acquisition window, and establish a unified batch identifier; Based on the data acquisition window, industrial operation data from multiple work sections are bound across domains according to a unified batch identifier to generate a cross-domain process evolution sequence.
[0009] As a preferred embodiment of the automated control method in the livestock feed processing process described in this invention, the specific steps of performing multi-dimensional state folding mapping on the cross-domain process evolution sequence, extracting the state action trajectory and the forming response trajectory, and generating a cross-domain corresponding trajectory set are as follows: Based on the cross-domain process evolution sequence, the industrial operation data of multiple sections are decoupled and mapped into the control domain according to the processing section identifier to generate the front-end state sequence and the back-end response sequence. Based on the preceding state sequence, a collaborative folding algorithm is used to collaboratively compress and merge the thermodynamic and pressure state variables to generate state action trajectories. Based on the subsequent response sequence, the response compression and fluctuation merging of the execution axis load feedback and finished product quality feedback signals are performed through the fluctuation folding algorithm to generate the molding response trajectory. Perform cross-domain flow grouping on the state action trajectory and the formed response trajectory to generate cross-domain corresponding trajectory groups.
[0010] As a preferred embodiment of the automated control method in the livestock feed processing process described in this invention, the method employs a cross-domain transfer potential field construction algorithm to perform traction correlation calculations on cross-domain corresponding trajectory groups to construct a cross-domain transfer potential field. The specific steps are as follows: Based on the cross-domain corresponding trajectory group, the cross-domain action direction of the state action trajectory and the forming response trajectory is determined by the traction direction identification algorithm, and a traction direction sequence is generated. Based on the traction direction sequence, the cross-domain action intensity of the corresponding trajectory group is continuously quantified to generate a traction intensity distribution chain. Based on the traction direction sequence and traction intensity distribution chain, a cross-domain potential field is constructed by reconstructing the distribution of cross-domain action direction and cross-domain action intensity through a cross-domain potential field aggregation algorithm.
[0011] As a preferred embodiment of the automated control method in the livestock feed processing process described in this invention, the specific steps for obtaining cross-domain distribution characteristics by performing traction main domain aggregation based on the cross-domain transfer potential field are as follows: The main domain condensation is traction-driven on the cross-domain transfer potential field, and the dominant convergence is identified by the cross-domain action direction and cross-domain action intensity to generate the main domain traction region sequence. The cross-domain distribution features of the main domain traction region sequence are extracted and the spatial distribution is summarized to obtain the cross-domain distribution features.
[0012] As a preferred embodiment of the automated control method in the livestock feed processing process described in this invention, the target traction distribution is generated by classifying and merging cross-domain distribution features using a domain convergence algorithm to generate a domain traction set, and then performing target traction reconstruction.
[0013] As a preferred embodiment of the automated control method for livestock feed processing described in this invention, the specific steps for determining the current dominant control target and corresponding adjustment position based on the target traction distribution and combined with the cross-domain transfer potential field, and generating a process control instruction set, are as follows: Based on the target traction distribution and cross-domain transfer potential field, the dominant target competition judgment is performed, and the degree of dominance is calculated for competition comparison to determine the current dominant control target; Extract the current dominant control target in the cross-domain transfer potential field corresponding to the cross-domain action concentration region, and perform adjustment position anchoring to determine the adjustment position of the preprocessing control domain and the adjustment position of the shaping control domain; Based on the adjustment positions of the preprocessed control domain and the shaping control domain, and combined with the target traction distribution, cross-domain adjustment channel allocation is performed to generate an adjustment channel configuration sequence; The gating instruction sequence of the regulation channel is arranged, and the regulation gating configuration is completed according to the current dominant control objective to generate the process control instruction set.
[0014] In a second aspect, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, wherein when the computer program is executed by the processor, it implements any step of the automated control method in the livestock feed processing process as described in the first aspect of the present invention.
[0015] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the automated control method in the livestock feed processing process as described in the first aspect of the present invention.
[0016] The beneficial effects of this invention are as follows: by binding multi-segment industrial operation data with processing segment identifiers in a time sequence, a cross-domain process evolution sequence is constructed, and a state action trajectory and a forming response trajectory are formed through multi-dimensional state folding mapping; by using a cross-domain transfer potential field construction algorithm to reconstruct the cross-domain interaction relationship between the preceding state and the subsequent response, and combining target traction distribution, dominant target competition judgment, and gating instruction arrangement, process control decision output is realized; it can more accurately identify cross-segment transmission influences, more effectively determine the dominant control target and adjustment position, thereby improving the control coherence, process adaptability, and overall operational stability in the livestock feed processing process. Attached Figure Description
[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0018] Figure 1 This is a flowchart of an automated control method for livestock feed processing.
[0019] Figure 2 A flowchart for generating cross-domain corresponding trajectory groups.
[0020] Figure 3 A flowchart for constructing a cross-domain transport potential field.
[0021] Figure 4 A flowchart for generating the target traction distribution.
[0022] Figure 5 This is a comparison chart of load feedback.
[0023] Figure 6 This is a comparison chart of quality feedback signals. Detailed Implementation
[0024] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.
[0025] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.
[0026] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.
[0027] Reference Figures 1-6 As one embodiment of the present invention, this embodiment provides an automated control method for livestock feed processing, comprising the following steps: S1: Collect multi-segment industrial operation data and processing segment identifiers, and bind the data stream time sequence to generate a cross-domain process evolution sequence.
[0028] S1.1: Multi-stage industrial operation data includes thermodynamic state variables, pressure state variables, actuator load feedback, and finished product quality feedback signals.
[0029] Specifically, temperature change data obtained through a temperature acquisition sensor and humidity change data obtained through a humidity acquisition sensor are combined to obtain the thermodynamic state variable composed of the temperature change data and humidity change data.
[0030] Pressure change data is obtained by pressure acquisition sensors, and the pressure change data reflecting the changes in the pressure state of the material are integrated to form a pressure state variable.
[0031] By using a load acquisition sensor located at the transmission position of the actuating axis of the machining equipment, load change data during the operation of the actuating axis is continuously collected to obtain load feedback of the actuating axis.
[0032] By acquiring finished product quality signal acquisition devices at the discharge position after processing, the particle durability detection data of the processed material is collected to obtain finished product quality feedback signals.
[0033] S1.2: The processing section identifier includes the pretreatment control field and the forming control field.
[0034] Specifically, based on the pretreatment processing area through which the material passes before entering the molding process, the time when the material enters the pretreatment processing area and the time when it leaves the pretreatment processing area are recorded. Then, the multi-stage industrial operation data collected within this time interval are mapped to the same processing area and marked as the pretreatment control domain.
[0035] Record the time when the material enters the molding processing position and passes through the molding processing area until the finished product is output, and record the time when the finished product is output. Then, map the multi-stage industrial operation data collected within the time interval to the same processing area and mark it as the molding control domain.
[0036] S1.3: Record the start and end times of the preprocessing control domain and the forming control domain respectively, obtain the data acquisition window, and establish a unified batch identifier.
[0037] Specifically, the time when the material enters the pretreatment control domain is recorded as the start time of the pretreatment control domain, and the time when the material leaves the pretreatment control domain is recorded as the end time of the pretreatment control domain; the time when the material enters the molding control domain is recorded as the start time of the molding control domain, and the time when the material completes the output of the finished product is recorded as the end time of the molding control domain.
[0038] The continuous time interval between the start time of the pretreatment control domain and the end time of the molding control domain is defined as the data acquisition window; the thermodynamic state variables, pressure state variables, actuator load feedback and finished product quality feedback signals corresponding to the same data acquisition window are classified into the same material batch, and a unique number is assigned to each data acquisition window to establish a unified batch identifier.
[0039] S1.4: Based on the data acquisition window, the industrial operation data of multiple work sections are bound across domains according to a unified batch identifier to generate a cross-domain process evolution sequence.
[0040] Specifically, thermodynamic state variables, pressure state variables, actuator load feedback, and finished product quality feedback signals with uniform batch identifiers are read from the data acquisition window and arranged in order according to the time sequence of the pre-processing control domain and the molding control domain. The thermodynamic state variables and pressure state variables located in the time interval of the pre-processing control domain are used as the operating information of the previous stage, and the actuator load feedback and finished product quality feedback signals located in the time interval of the molding control domain are used as the operating information of the subsequent stage.
[0041] According to the binding rules that the same unified batch identifier corresponds to the same material batch and the same data acquisition window corresponds to the same continuous processing process, the operation information of the preceding section and the operation information of the following section are connected across domains in the order of time progression; the operation information of the preceding section and the operation information of the following section are unfolded in the processing order of pre-processing control domain first and forming control domain later, forming a cross-domain process evolution sequence.
[0042] S2: Perform multi-dimensional state folding mapping on the cross-domain process evolution sequence, extract the state action trajectory and forming response trajectory, and generate the cross-domain corresponding trajectory group.
[0043] S2.1: Based on the cross-domain process evolution sequence, the industrial operation data of multiple work sections are decoupled and mapped in the control domain according to the processing section identifier to generate the front-end state sequence and the back-end response sequence.
[0044] Specifically, thermodynamic and pressure state variables with pre-processing control domain identifiers are read from the cross-domain process evolution sequence, and execution axis load feedback and finished product quality feedback signals with forming control domain identifiers are read from the cross-domain process evolution sequence.
[0045] According to the processing section identifier, the thermodynamic state variables and pressure state variables are assigned to the pre-processing control domain data set, and the actuator load feedback and finished product quality feedback signals are assigned to the forming control domain data set. According to the time progression order in the cross-domain process evolution sequence, the thermodynamic state variables and pressure state variables in the data set corresponding to the pre-processing control domain are continuously arranged to generate the pre-stage state sequence. The actuator load feedback and finished product quality feedback signals in the data set corresponding to the forming control domain are continuously arranged to generate the post-stage response sequence.
[0046] S2.2: Based on the previous state sequence, the thermodynamic state variables and pressure state variables are compressed and their effects are merged through a cooperative folding algorithm to generate state action trajectories.
[0047] Specifically, thermodynamic state variables and pressure state variables are read sequentially according to the time progression order in the preceding state sequence, and thermodynamic state variables and pressure state variables at adjacent time positions are arranged accordingly; the changing trends of thermodynamic state variables and pressure state variables at each group of adjacent time positions are compared synchronously, and positions where the changing trends of thermodynamic state variables and pressure state variables are in the same direction and continuous and located in adjacent time slices within the same data acquisition window are divided into the same coordinated segment.
[0048] Based on the collaborative folding algorithm, the thermodynamic and pressure state variables corresponding to multiple continuous acquisition locations are grouped into a continuous action segment. According to the time progression order in the previous state sequence, all continuous action segments are sequentially connected and unfolded to form a state action trajectory that can characterize the continuous action process of thermodynamic and pressure state variables in the preprocessed control domain.
[0049] It should be noted that the cooperative folding algorithm refers to a processing method that cooperatively compresses and merges the thermodynamic state variables and pressure state variables in the preceding state sequence. By dividing adjacent state records that maintain the same continuous trend and are in the same continuous processing stage into the same cooperative segment, and merging the continuous state change process within the same cooperative segment into continuous action segments, a state action trajectory that can characterize the continuous state action process within the preprocessed control domain is formed.
[0050] S2.3: Based on the subsequent response sequence, the response compression and fluctuation merging of the execution axis load feedback and finished product quality feedback signals are performed through the fluctuation folding algorithm to generate the molding response trajectory.
[0051] Specifically, according to the order of the subsequent response sequence, the load feedback of the execution axis corresponding to each acquisition moment and the finished product quality feedback signal corresponding to the same acquisition moment are combined into a set of response records, and then a continuous response record column is formed in sequence.
[0052] Starting from the first group of response records in the continuous response record column, compare the direction of change in the execution axis load feedback and the direction of change in the finished product quality feedback signal in the current response record and the next group of response records. When the execution axis load feedback continuously increases and the finished product quality feedback signal continuously decreases, or vice versa, the current response record and the next group of response records are grouped into the same fluctuation segment. When the direction of change in the execution axis load feedback in the next group of response records reverses the direction of change in the execution axis load feedback in the previous group of response records, and the direction of change in the finished product quality feedback signal reverses, the next fluctuation segment is redefined from the position of change.
[0053] Based on the load feedback and finished product quality feedback signals of the execution axis within each fluctuation segment, the response state corresponding to the first acquisition position, the response state corresponding to the last acquisition position, and the intermediate continuous change process are merged into a continuous response segment; according to the order in the subsequent response sequence, each continuous response segment is connected in sequence to form a shaped response trajectory.
[0054] It should be noted that the fluctuation folding algorithm refers to a processing method that continuously compresses and merges fluctuations in the execution axis load feedback and finished product quality feedback signals in the subsequent response sequence. By dividing adjacent response records that maintain a continuous correspondence in the direction of change into the same fluctuation segment, and merging the continuous response process within the same fluctuation segment into continuous response segments, a molding response trajectory that can characterize the continuous response change process within the molding control domain is formed.
[0055] S2.4: Perform cross-domain flow grouping on the state action trajectory and the formed response trajectory to generate cross-domain corresponding trajectory groups.
[0056] Specifically, first, read each continuous action segment in the state action trajectory according to the time progression order in the cross-domain process evolution sequence, and then read each continuous response segment in the forming response trajectory according to the time progression order corresponding to the subsequent response sequence.
[0057] Using the end position of each continuous action segment in the state action trajectory as the basis for preceding positioning, the continuous response segment located at a subsequent time position in the forming response trajectory and corresponding to the continuous time progression relationship within the same unified batch identifier and the same data acquisition window is searched; the found continuous response segments are paired one by one with the corresponding continuous action segments; each group of connected continuous action segments and continuous response segments is grouped and arranged in the order of preceding level before following level to form multiple sets of trajectory pairings with clear preceding and following connections, and all trajectory pairing sets are arranged sequentially according to the overall time progression order in the cross-domain process evolution sequence to generate cross-domain corresponding trajectory groups.
[0058] It should be noted that the cross-domain corresponding trajectory group is a set of trajectories formed by grouping the state action trajectory and the forming response trajectory according to the time progression relationship between the preceding and following work sections. The function of the cross-domain corresponding trajectory group is to establish a clear correspondence between the continuous state change process in the preprocessing control domain and the continuous response change process in the forming control domain, so as to provide a direct basis for subsequent determination of the cross-domain action direction, quantification of the cross-domain action intensity and construction of the cross-domain transfer potential field.
[0059] S3: Employ a cross-domain transfer potential field construction algorithm to perform traction correlation calculations on the corresponding cross-domain trajectory groups and construct a cross-domain transfer potential field.
[0060] S3.1: Based on the cross-domain corresponding trajectory group, the cross-domain action direction of the state action trajectory and the forming response trajectory is determined by the traction direction identification algorithm, and a traction direction sequence is generated.
[0061] Specifically, the state action trajectory and the shaping response trajectory in each trajectory pairing set are read sequentially according to the grouping order in the cross-domain corresponding trajectory group. Each continuous action segment is read sequentially from the state action trajectory in each trajectory pairing set, and each continuous response segment corresponding to each continuous action segment in the time progression position is read sequentially from the shaping response trajectory in the same trajectory pairing set.
[0062] Following the arrangement of the state action trajectory first and the forming response trajectory second, each group of continuous action segments and continuous response segments is compared and matched, and the state change trend corresponding to the continuous action segment and the response change trend corresponding to the continuous response segment are recorded. The traction direction identification algorithm is used to determine the direction of the state change trend and the response change trend. When the state change trend and the response change trend maintain a continuous change relationship in the same direction, it is determined to be a traction direction in the same direction. When the state change trend and the response change trend maintain a continuous change relationship in opposite directions, it is determined to be a traction direction in opposite directions.
[0063] Arrange the same-direction traction direction and the opposite-direction traction direction in sequence according to the time progression order in the corresponding cross-domain trajectory group to generate a traction direction sequence.
[0064] It should be noted that the traction direction identification algorithm is a processing method for determining the direction of the correspondence between the state action trajectory and the forming response trajectory. By comparing the corresponding changes in the state change trend and the response change trend in the continuous processing process, it identifies the unidirectional or reverse traction relationship formed by the change in the previous state on the change in the subsequent response, providing a directional basis for the subsequent quantification of cross-domain action intensity and the construction of cross-domain transfer potential field.
[0065] S3.2: Based on the traction direction sequence, the cross-domain action intensity of the corresponding trajectory group across the cross domain is continuously quantified to generate the traction intensity distribution chain.
[0066] Specifically, the trajectory pairing set corresponding to each traction direction is read sequentially according to the arrangement order in the traction direction sequence, and the continuous action segment in the state action trajectory and the continuous response segment in the forming response trajectory are read from the corresponding trajectory pairing set.
[0067] Based on the correspondence of continuous action segments preceding continuous response segments, the duration of consistent state change trends between adjacent positions of continuous action segments during time progression is statistically analyzed to obtain the degree of change continuity; the time interval between the end position of a continuous action segment and the start position of a continuous response segment is recorded to obtain the degree of time continuity; and the difference in changes in the execution axis load feedback and finished product quality feedback signals between the preceding and following positions within a continuous response segment is statistically analyzed to obtain the response fluctuation amplitude.
[0068] The quantification criteria are uniformly defined as the degree of continuity of change in continuous action segments, the tightness of temporal continuity, and the amplitude of response fluctuations in continuous response segments, and normalization is performed using the range normalization method; based on the traction direction at the corresponding position in the traction direction sequence, the cross-domain action intensity is calculated, and the expression is: ; in, For cross-domain interaction strength, The degree of continuity of the changes after normalization To determine the tightness of time continuity after normalization, This represents the normalized response fluctuation amplitude.
[0069] It should be noted that the degree of change, the tightness of time continuity, and the amplitude of response fluctuations have all been normalized to dimensionless values. All terms in the numerator and denominator are dimensionless quantities. Therefore, the expression for calculating the intensity of cross-domain interaction has unified dimensions and is dimensionless.
[0070] The range normalization method refers to a processing method that uses the maximum and minimum values in the same type of data to compress the current value into a range, and transforms data with different dimensions or different value ranges into the same dimensionless range, which facilitates subsequent intensity calculation, comparative analysis and comprehensive processing.
[0071] The cross-domain action intensity is arranged sequentially according to the time progression in the traction direction sequence, and the correspondence between each cross-domain action intensity and the corresponding traction direction, state action trajectory and forming response trajectory is maintained to generate a traction intensity distribution chain.
[0072] S3.3: Based on the traction direction sequence and traction intensity distribution chain, the cross-domain action direction and cross-domain action intensity are reconstructed by the cross-domain potential field aggregation algorithm to construct the cross-domain transfer potential field.
[0073] Specifically, after generating the traction direction sequence and the traction intensity distribution chain, the traction direction corresponding to each position is read sequentially according to the arrangement order in the traction direction sequence; the cross-domain action intensity corresponding to each position in the traction intensity distribution chain is read in the same arrangement order, and each traction direction is combined with the cross-domain action intensity at the same position.
[0074] According to the time progression order in the cross-domain corresponding trajectory group, corresponding combinations with the same traction direction and continuously changing cross-domain action intensity at adjacent positions are aggregated in the same direction to form continuous traction segments; the sequential connection relationship between adjacent continuous traction segments is sorted out, and continuous traction segments that are connected in time and jointly represent the same continuous processing and propulsion process are connected in sequence to form a continuous distribution structure of cross-domain action direction and cross-domain action intensity.
[0075] The algorithm for constructing cross-domain potential fields is used to reconstruct the overall distribution of all continuous traction segments. Each traction direction is continuously unfolded along the time progression under the support of the corresponding cross-domain action strength, and the one-to-one correspondence between the traction direction, the cross-domain action strength and the corresponding cross-domain trajectory group is maintained to construct the cross-domain transfer potential field.
[0076] It should be noted that the cross-domain transfer potential field is a continuous action representation formed by reconstructing the cross-domain action direction and intensity between the state action trajectory and the forming response trajectory. The role of the cross-domain transfer potential field is to improve the transfer relationship from the previous state change to the subsequent response change from discrete correspondence to continuous distribution expression, providing a basis for subsequent identification of the dominant traction region, extraction of cross-domain distribution features, and determination of the current dominant control target and adjustment position.
[0077] It should be noted that the cross-domain potential field aggregation construction algorithm refers to a processing method that jointly organizes and reconstructs the traction direction sequence and traction intensity distribution chain. It aggregates cross-domain interaction relationships with consistent direction and continuous intensity changes during the time progression, and sequentially connects the continuous traction segments to form a cross-domain transfer potential field that can continuously characterize the process of the transmission of changes in the previous state to changes in the subsequent response.
[0078] S4: Based on the cross-domain transfer potential field, perform traction main domain aggregation, obtain cross-domain distribution characteristics, and perform sub-domain aggregation to generate the target traction distribution.
[0079] S4.1: Perform traction principal domain condensation on the cross-domain transfer potential field, and identify the dominant convergence of the cross-domain action direction and cross-domain action intensity to generate a principal domain traction region sequence.
[0080] Specifically, after the construction of the cross-domain transfer potential field is completed, the traction direction and cross-domain action intensity corresponding to each continuous traction segment are read sequentially according to the time progression order in the cross-domain transfer potential field, and adjacent continuous traction segments are arranged in order according to their front-to-back positional relationship.
[0081] Adjacent continuous traction sections with consistent traction directions are divided into the same candidate cohesion section, and the continuity of the cross-domain action intensity corresponding to each continuous traction section within the same candidate cohesion section is compared. Continuous traction sections with continuous aggregation of cross-domain action intensity are retained within the same candidate cohesion section, and the position where the cross-domain action intensity changes discontinuously is taken as the starting point of the new candidate cohesion section. For example, when the cross-domain action intensities corresponding to adjacent continuous traction sections in the time progression are 0.72, 0.75, 0.78, and 0.31 respectively, since 0.72, 0.75, and 0.78 maintain a continuous aggregation state while 0.31 shows a discontinuous change relative to the previous position, the continuous traction sections corresponding to 0.72, 0.75, and 0.78 are retained within the same candidate cohesion section, and the position corresponding to 0.31 is taken as the starting point of the new candidate cohesion section.
[0082] After completing the comparison of the consistency of traction direction and the comparison of the continuity of cross-domain action intensity, traction main domain agglomeration is performed on each candidate agglomeration segment. Continuous traction segments with consistent traction direction and continuous agglomeration of cross-domain action intensity are merged and organized into dominant traction regions. After completing the identification of all dominant traction regions, each dominant traction region is arranged in sequence according to the time progression order in the cross-domain transfer potential field to generate the main domain traction region sequence.
[0083] It should be noted that the traction domain aggregation refers to the process of identifying and merging the dominant convergence of continuous traction segments in the cross-domain transfer potential field where the traction direction is consistent and the cross-domain interaction intensity is continuously concentrated. The role of traction domain aggregation is to extract the dominant traction region from the continuously distributed cross-domain interaction relationship, thereby providing a basis for subsequent acquisition of cross-domain distribution characteristics and generation of target traction distribution.
[0084] S4.2: Extract cross-domain distribution features from the main domain traction region sequence and summarize its spatial distribution to obtain cross-domain distribution features.
[0085] Specifically, the starting and ending positions of each dominant traction region in the cross-domain transfer potential field are read sequentially according to the arrangement order in the main domain traction region sequence, and the dominant traction regions are arranged in front and behind according to the time progression order in the cross-domain transfer potential field.
[0086] The interval positions, continuous extension positions, and concentrated distribution positions between adjacent main traction zones are extracted and identified segment by segment; the distribution state where the main traction zones are continuously connected is recorded as a continuous distribution relationship; the distribution state where there are intervals between the main traction zones is recorded as an interval distribution relationship; and the distribution state where multiple main traction zones appear in concentrated positions at adjacent locations is recorded as a concentrated distribution relationship.
[0087] The distribution location, length, and concentration of each dominant traction zone are statistically analyzed to obtain the sequential arrangement characteristics of the dominant traction zones in the cross-domain transfer potential field. For example, in the sequence of dominant traction zones, the dominant traction zone located at the pre-processing control domain appears first, followed by the dominant traction zone located at the forming control domain, representing the sequential arrangement of the dominant traction zones along the processing direction; continuous extension characteristics, for example, the same dominant traction zone extends continuously from one position to the next in the time progression process without any interruption, representing the continuous extension state of the dominant traction zone in the cross-domain transfer potential field; and concentrated aggregation characteristics, for example, multiple dominant traction zones appear consecutively at adjacent positions and are concentrated in the same local area, representing the concentrated aggregation state of the dominant traction zones in the cross-domain transfer potential field, forming a cross-domain distribution characteristic.
[0088] S4.3: The cross-domain distribution features are classified and merged using the domain convergence algorithm to generate the domain traction set, and the target traction reconstruction is performed to generate the target traction distribution.
[0089] Specifically, the distribution location, distribution length, and distribution concentration of each dominant traction zone are read sequentially according to the order of the cross-domain distribution characteristics, and the dominant traction zones that are adjacent to each other in distribution location, have similar distribution length, and maintain the same kind of change relationship in distribution concentration are classified into the same sub-domain.
[0090] A domain-based aggregation algorithm is used to classify and merge the dominant traction regions within the same domain. The dominant traction regions within the same domain are merged and organized to generate a domain-based traction set. The distribution position and traction aggregation state of each domain-based traction set are read sequentially according to the time progression order in the cross-domain transfer potential field. The domain-based traction sets are then rearranged according to their previous and next positional relationships and continuously expanded. The traction aggregation states corresponding to each domain-based traction set are then integrated as a whole according to the time progression order to generate the target traction distribution.
[0091] It should be noted that the domain convergence algorithm refers to a processing method that classifies and merges dominant traction areas with similar distribution locations, similar distribution lengths, and the same traction aggregation trends in cross-domain distribution characteristics. By organizing dominant traction areas with clear connections and similar distribution patterns into the same domain, a domain traction set is formed.
[0092] S5: Based on the target traction distribution and combined with the cross-domain transfer potential field, determine the current dominant control target and corresponding adjustment position, and generate a process control instruction set.
[0093] S5.1: Based on the target traction distribution and cross-domain transfer potential field, perform dominant target competition determination, calculate the degree of dominance for competition comparison, and determine the current dominant control target.
[0094] Specifically, based on the time progression order in the target traction distribution, the traction aggregation state and distribution position corresponding to the sub-domain traction set are read sequentially, and the traction direction, cross-domain action intensity and continuous traction section consistent with the corresponding position of each sub-domain traction set are read from the cross-domain transfer potential field.
[0095] The traction aggregation state corresponding to each sub-domain traction set is associated with the traction direction, cross-domain action intensity and continuous traction section at the same location in groups, and the traction aggregation duration, traction direction concentration and cross-domain action intensity concentration of each sub-domain traction set at continuous time location are statistically analyzed.
[0096] The duration of traction aggregation is quantified as the duration of traction aggregation, the concentration of traction direction is quantified as the concentration ratio of traction direction, and the concentration of cross-domain action intensity is quantified as the average concentration of cross-domain action intensity. The duration of traction aggregation, the concentration ratio of traction direction, and the average concentration of cross-domain action intensity are used as the basis for calculating the degree of dominance, and the degree of dominance of each sub-domain traction set is calculated. According to the order of the target traction distribution, the sub-domain traction set with the highest degree of dominance and the priority aggregation state in continuous time position is determined as the current dominant control target.
[0097] The expression for calculating the dominance of the domain-specific traction set is: ; in, The degree of dominance of the domain-driven set, For the normalized first The traction aggregation duration value corresponding to each domain traction set. For the normalized first The proportion of traction directions corresponding to each domain traction set. For the normalized first Cluster mean of cross-domain action intensity corresponding to each domain traction set The total number of the domain-dependent traction sets. This is the index of the domain-based traction set.
[0098] like Figure 5The comparison results between the threshold-based adjustment method and the method of this invention in terms of actuator shaft load feedback are shown. The threshold-based adjustment method only continuously reads the actuator shaft load feedback within the forming control domain and uses whether the actuator shaft load feedback exceeds a threshold as the adjustment trigger. When it exceeds the threshold, a single load suppression adjustment is performed, and the adjustment stops when it falls back to the threshold range. In contrast, the method of this invention first binds the thermodynamic state variables, pressure state variables, actuator shaft load feedback, and finished product quality feedback signals in the preprocessing control domain and the forming control domain with unified batch identification, generating a cross-domain process evolution sequence. Then, through control domain decoupling mapping, a pre-stage state sequence and a post-stage response sequence are formed. The state action trajectory and forming response trajectory are generated by the cooperative folding algorithm and the fluctuation folding algorithm, respectively. Subsequently, a cross-domain transmission potential field is constructed to complete the dominant target competition determination, adjustment position anchoring, and gating instruction arrangement, outputting a process control instruction set. As can be seen from the figure, the curve corresponding to the method of this invention falls back faster and has a lower peak value in the key fluctuation range, indicating that the method of this invention can more accurately identify the cross-section transmission influence and more effectively determine the adjustment position, thereby reducing actuator shaft load feedback fluctuations and improving control coherence and overall operational stability.
[0099] S5.2: Extract the concentrated region of cross-domain action corresponding to the current dominant control target in the cross-domain transfer potential field, and perform adjustment position anchoring to determine the adjustment position of the preprocessing control domain and the adjustment position of the shaping control domain.
[0100] Specifically, the sub-domain traction set and distribution position corresponding to the current dominant control target are read according to the arrangement order in the target traction distribution, and the continuous traction segment corresponding to the sub-domain traction set position is found from the cross-domain transfer potential field; the continuous traction segments with consistent traction direction and continuous aggregation of cross-domain action intensity are organized into cross-domain action concentration area.
[0101] Based on the front-to-back positional relationship in the transdomain potential field, the transdomain interaction concentration region located at the front level and the transdomain interaction concentration region located at the back level are segmented and distinguished. The transdomain interaction concentration region located at the front level is mapped to the preprocessing control domain, and the transdomain interaction concentration region located at the back level is mapped to the forming control domain.
[0102] Adjustment position anchoring is performed according to the specific distribution location of the cross-domain action concentration area in the pretreatment control domain and the forming control domain. The position in the pretreatment control domain corresponding to the cross-domain action concentration area is determined as the pretreatment control domain adjustment position, and the position in the forming control domain corresponding to the cross-domain action concentration area is determined as the forming control domain adjustment position.
[0103] S5.3: Based on the adjustment positions of the preprocessing control domain and the shaping control domain, and combined with the target traction distribution, perform cross-domain adjustment channel allocation to generate an adjustment channel configuration sequence.
[0104] Specifically, the thermodynamic state variables and pressure state variables corresponding to the pretreatment control domain adjustment positions are read sequentially, with the pretreatment control domain adjustment position preceding the molding control domain adjustment position, as well as the actuator load feedback and finished product quality feedback signals corresponding to the molding control domain adjustment positions.
[0105] Based on the order of arrangement in the target traction distribution, the domain traction set, traction aggregation state and distribution position corresponding to the current dominant control target are read, and the domain traction set, traction aggregation state and distribution position are associated with the thermodynamic state variables, pressure state variables, actuator load feedback and finished product quality feedback signals.
[0106] Based on the traction aggregation state and distribution location of each sub-domain traction set in the target traction distribution, and according to the thermodynamic state variables, pressure state variables, actuator load feedback, and finished product quality feedback signals and the current dominant control target, the regulation channel that corresponds to the current dominant control target and has a concentrated traction aggregation state is determined as the priority execution regulation channel, the regulation channel that is adjacent to the current dominant control target and has a secondary concentrated traction aggregation state is determined as the subsequent execution regulation channel, and the regulation channel that does not fall within the range of the current dominant control target is determined as the temporarily suspended execution regulation channel.
[0107] Following the cross-domain advancement order of preprocessing control domain first and shaping control domain last, the priority adjustment channels and subsequent adjustment channels are arranged sequentially to generate an adjustment channel configuration sequence.
[0108] S5.4: Perform gated instruction arrangement on the regulation channel configuration sequence, complete the regulation gate configuration according to the current dominant control objective, and generate process control instruction set.
[0109] Specifically, according to the order of the adjustment channel configuration sequence, the first priority adjustment instruction is written to the priority adjustment channel, the second priority adjustment instruction is written to the subsequent adjustment channel, and the adjustment instruction to remain in a waiting state is written to the deferred adjustment channel.
[0110] After completing the writing of the first-order adjustment instruction, the second-order adjustment instruction, and the adjustment instruction that remains in a waiting state, all adjustment instructions are arranged in a cross-domain progression order, with the preprocessing control domain preceding the shaping control domain. The adjustment actions corresponding to the preceding adjustment channel are entered into the instruction arrangement position first, and the adjustment actions corresponding to the following adjustment channel are entered into the instruction arrangement position later.
[0111] After completing the coordination and integration, all adjustment instructions are arranged in the order of execution corresponding to the current dominant control objective to form a process control instruction set.
[0112] likeFigure 6 The comparison results between the threshold-based adjustment method and the method of this invention on the finished product quality feedback signal are shown. The threshold-based adjustment method independently monitors the finished product quality feedback signal in the molding control domain. When the finished product quality feedback signal is lower than the threshold, it triggers a single quality compensation adjustment. It does not read the thermodynamic state variables and pressure state variables in the preprocessing control domain, nor does it perform cross-domain correlation analysis. Therefore, it can only make local corrections based on the current quality results. The method of this invention, on the other hand, binds the multi-stage industrial operation data in a time sequence to form a cross-domain process evolution sequence. It uses the thermodynamic state variables and pressure state variables in the preprocessing control domain as the operating information of the previous stage, and the actuator load feedback and finished product quality feedback signal in the molding control domain as the operating information of the next stage. It establishes the interaction relationship between the previous and next stages through the state action trajectory, molding response trajectory, cross-domain corresponding trajectory group, and cross-domain transfer potential field, and completes the determination of the dominant target competition and the generation of the process control instruction set based on the target traction distribution. Figure 6 As can be seen, the finished product quality feedback signal corresponding to the method of the present invention is more stable and less discrete overall. This indicates that the method of the present invention does not passively correct after quality deviation occurs, but integrates the cross-section transmission of influence and multi-objective competition into the control decision-making process. Therefore, it can better maintain the quality stability and process adaptability of the molding stage.
[0113] This embodiment also provides a computer device applicable to the automated control method in the livestock feed processing process, comprising: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the automated control method in the livestock feed processing process proposed in the above embodiment.
[0114] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.
[0115] This embodiment also provides a storage medium storing a computer program, which, when executed by a processor, implements the automated control method for livestock feed processing as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.
[0116] In summary, this invention constructs a cross-domain process evolution sequence by temporally binding multi-segment industrial operation data with processing segment identifiers, and forms state action trajectories and forming response trajectories through multi-dimensional state folding mapping; it reconstructs the cross-domain interaction relationship between the preceding state and the subsequent response by utilizing a cross-domain transfer potential field construction algorithm, and achieves process control decision output by combining target traction distribution, dominant target competition determination, and gating instruction arrangement; it can more accurately identify cross-segment transfer effects, more effectively determine the dominant control target and adjustment position, thereby improving the control coherence, process adaptability, and overall operational stability in livestock feed processing.
[0117] It should be noted that 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 preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.
Claims
1. An automated control method for livestock feed processing, characterized in that, include: Collect multi-stage industrial operation data and processing stage identifiers, and bind the data streams to the time sequence to generate a cross-domain process evolution sequence; Perform multidimensional state folding mapping on the cross-domain process evolution sequence, extract the state action trajectory and forming response trajectory, and generate cross-domain corresponding trajectory groups; A cross-domain transfer potential field construction algorithm is adopted to perform traction correlation calculation on the cross-domain corresponding trajectory group to construct the cross-domain transfer potential field. Based on the cross-domain transfer potential field, the traction main domain agglomeration is performed to obtain the cross-domain distribution characteristics, and then the domain convergence is performed to generate the target traction distribution. Based on the target traction distribution and combined with the cross-domain transfer potential field, the current dominant control target and corresponding adjustment position are determined, and a process control instruction set is generated.
2. The automated control method for livestock feed processing as described in claim 1, characterized in that, The multi-section industrial operation data includes thermodynamic state variables, pressure state variables, actuator load feedback, and finished product quality feedback signals. The processing section identifier includes a pre-processing control field and a forming control field.
3. The automated control method for livestock feed processing as described in claim 2, characterized in that, The specific steps for generating the cross-domain process evolution sequence are as follows: Record the start and end times of the preprocessing control domain and the forming control domain respectively, obtain the data acquisition window, and establish a unified batch identifier; Based on the data acquisition window, industrial operation data from multiple work sections are bound across domains according to a unified batch identifier to generate a cross-domain process evolution sequence.
4. The automated control method for livestock feed processing as described in claim 3, characterized in that, The process of performing multidimensional state folding mapping on the cross-domain process evolution sequence, extracting the state action trajectory and forming response trajectory, and generating cross-domain corresponding trajectory groups are as follows: Based on the cross-domain process evolution sequence, the industrial operation data of multiple sections are decoupled and mapped into the control domain according to the processing section identifier to generate the front-end state sequence and the back-end response sequence. Based on the preceding state sequence, a collaborative folding algorithm is used to collaboratively compress and merge the thermodynamic and pressure state variables to generate state action trajectories. Based on the subsequent response sequence, the response compression and fluctuation merging of the execution axis load feedback and finished product quality feedback signals are performed through the fluctuation folding algorithm to generate the molding response trajectory. Perform cross-domain flow grouping on the state action trajectory and the formed response trajectory to generate cross-domain corresponding trajectory groups.
5. The automated control method for livestock feed processing as described in claim 1 or 4, characterized in that, The method employs a cross-domain transfer potential field construction algorithm to perform traction correlation calculations on cross-domain corresponding trajectory groups and construct a cross-domain transfer potential field. The specific steps are as follows: Based on the cross-domain corresponding trajectory group, the cross-domain action direction of the state action trajectory and the forming response trajectory is determined by the traction direction identification algorithm, and a traction direction sequence is generated. Based on the traction direction sequence, the cross-domain action intensity of the corresponding trajectory group is continuously quantified to generate a traction intensity distribution chain. Based on the traction direction sequence and traction intensity distribution chain, a cross-domain potential field is constructed by reconstructing the distribution of cross-domain action direction and cross-domain action intensity through a cross-domain potential field aggregation algorithm.
6. The automated control method for livestock feed processing as described in claim 5, characterized in that, The specific steps for obtaining cross-domain distribution characteristics by performing traction-based main domain condensation based on the cross-domain transfer potential field are as follows: The main domain condensation is traction-driven on the cross-domain transfer potential field, and the dominant convergence is identified by the cross-domain action direction and cross-domain action intensity to generate the main domain traction region sequence. The cross-domain distribution features of the main domain traction region sequence are extracted and the spatial distribution is summarized to obtain the cross-domain distribution features.
7. The automated control method for livestock feed processing as described in claim 6, characterized in that, The target traction distribution is generated by classifying and merging cross-domain distribution features using a domain convergence algorithm to generate a domain traction set, and then performing target traction reconstruction.
8. The automated control method for livestock feed processing as described in claim 6, characterized in that, The process control instruction set is generated by determining the current dominant control target and corresponding adjustment position based on the target traction distribution and combined with the cross-domain transfer potential field, and the specific steps are as follows: Based on the target traction distribution and cross-domain transfer potential field, the dominant target competition judgment is performed, and the degree of dominance is calculated for competition comparison to determine the current dominant control target; Extract the current dominant control target in the cross-domain transfer potential field corresponding to the cross-domain action concentration region, and perform adjustment position anchoring to determine the adjustment position of the preprocessing control domain and the adjustment position of the shaping control domain; Based on the adjustment positions of the preprocessed control domain and the shaping control domain, and combined with the target traction distribution, cross-domain adjustment channel allocation is performed to generate an adjustment channel configuration sequence; The gating instruction sequence of the regulation channel is arranged, and the regulation gating configuration is completed according to the current dominant control objective to generate the process control instruction set.
9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the automated control method in the livestock feed processing process according to any one of claims 1 to 8.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the automated control method in the livestock feed processing process according to any one of claims 1 to 8.