A method for controlling operating parameters of a non-invasive material homogenizer

Abstract

The present application belongs to the technical field of homogenizer control, and particularly relates to a non-intrusive material homogenizer operation parameter control method, comprising: collecting the load current of a main drive motor and the revolution phase of a revolving arm in real time; obtaining the interpolation current value when the system trajectory crosses a preset Poincare section by monitoring the revolution phase and using a linear interpolation method; normalizing the interpolation current value and constructing Poincare characteristic points; calculating the manifold dispersion based on the Poincare characteristic points in a sliding window; and using an exponential function to adaptively calculate the target revolution speed of the next period according to the manifold dispersion, and performing variable structure control on the homogenizer. The present application can effectively separate the material intrinsic rheological characteristics from strong periodic mechanical noise, solve the phase lag problem caused by filtering in the prior art, and realize the hybrid end point determination and adaptive energy consumption control of the homogenizer.

Description

Technical Field

[0001] This invention relates to the field of homogenizer control technology. More specifically, this invention relates to a method for controlling the operating parameters of a non-invasive material homogenizer. Background Technology

[0002] Non-invasive material homogenizers are widely used in industrial production for mixing and degassing high-viscosity materials. Their working principle involves a rotating arm driving a tank to rotate at high speed, while the tank itself rotates in the same or opposite direction. This dual centrifugal force field generates strong shear force, achieving uniform mixing of the materials. In actual production, accurately sensing the mixing endpoint is crucial for ensuring product quality. Currently, the industry primarily infers the rheological state of the material within the tank by monitoring the load current or torque of the main drive motor, attempting to determine whether the material has reached the homogenization requirements through load changes.

[0003] However, the homogenizer has inherent strong periodic mechanical motion characteristics during operation, which poses a great challenge to the interpretation of load signals. During one rotation of the orbital arm, due to the asymmetry of the mechanical structure or the effect of gravity, the gravity component of the tank and fixture will produce significant periodic sinusoidal modulation on the motor load. At the same time, the rotation of the tank will also introduce high-frequency periodic vibration interference. This large-amplitude mechanical dynamic noise is often extremely strong and can easily mask the weak load fluctuations caused by changes in the viscosity or rheological properties of the material. This makes it difficult to directly extract the effective signal that can characterize the true state of the material, resulting in an extremely low signal-to-noise ratio.

[0004] To suppress the aforementioned strong periodic mechanical noise, existing technologies typically employ signal processing techniques such as low-pass filtering or averaging to smooth the original load data. While these methods reduce signal fluctuations to some extent, they can lead to severe phase lag, preventing the processed signal from reflecting the system's current state in real time. More seriously, simple filtering cannot distinguish between periodic fluctuations caused by changes in mechanical attitude and load fluctuations caused by changes in material rheological properties. This results in the control system being slow to judge the mixing endpoint or making misjudgments, causing the equipment to often be in an overmixed state. This not only wastes energy but also easily leads to thermal damage to the material due to heat accumulation from prolonged shearing, severely impacting production efficiency and yield. Summary of the Invention

[0005] To address the technical problems in existing technologies, such as interference from strong mechanical noise, phase lag caused by filtering, inability to distinguish fluctuations, leading to misjudgment of the mixing endpoint and thermal damage to materials, this invention provides a method for controlling the operating parameters of a non-invasive material homogenizer, comprising:

[0006] The load current of the main drive motor of the homogenizer and the revolution phase of the revolution arm are collected in real time to construct an operating dataset containing sampling time, load current, and revolution phase. The operating dataset is traversed, and when the revolution phase of an adjacent sampling point crosses the trigger phase of a preset Poincaré cross section, the interpolated current value at the point of crossing the Poincaré cross section is calculated based on the load current, revolution phase, and trigger phase of the adjacent sampling point, and Poincaré feature points are constructed based on the interpolated current value. Based on multiple Poincaré feature points at the current and historical times, the manifold dispersion is calculated, which characterizes the degree of dispersion of the Poincaré feature points in phase space. Based on the manifold dispersion and a preset mixed steady-state threshold, the target revolution speed for the next cycle is determined, and the operating state of the main drive motor is adjusted according to the target revolution speed.

[0007] This invention constructs a runtime dataset by real-time acquisition of load current and revolution phase, correlating dynamic load with kinematic attitude. By monitoring the phase crossing the Poincaré cross section and calculating interpolated current values ​​to construct Poincaré feature points, it achieves stroboscopic sampling under a fixed mechanical attitude. This physically filters out periodic gravitational interference caused by revolution and rotation, avoiding the phase lag problem of traditional filtering and ensuring that the sampled data accurately reflects the energy state of the system. This invention measures the dispersion of Poincaré feature points in phase space by calculating manifold dispersion, which can sensitively capture the dynamic convergence characteristics of materials transitioning from non-uniform to uniform, thereby accurately characterizing the mixing state of materials. Based on manifold dispersion, this invention adaptively adjusts the target revolution speed, providing strong shear force to break up agglomerates in the early stage of mixing and automatically reducing the speed in the later stage of mixing to reduce heat accumulation. This ensures mixing quality while preventing thermal damage to materials, achieving non-invasive, high-precision mixing endpoint determination and energy-saving control.

[0008] Preferably, the real-time acquisition of the load current of the main drive motor of the homogenizer and the revolution phase of the revolution arm to construct an operating dataset including sampling time, load current and revolution phase includes: acquiring the load current of the main drive motor at a preset sampling frequency using a current transformer; acquiring the revolution phase of the revolution arm using an absolute encoder; aligning the load current and revolution phase according to timestamps to generate an operating dataset.

[0009] Preferably, the interpolated current value when crossing the Poincaré section satisfies the expression: ;in, This represents the interpolated current value when passing through the Poincaré section; This represents the load current that has passed through the previous moment; This represents the load current at the moment after the crossing. Indicates the orbital phase at the moment after the crossing; This indicates the orbital phase at the point in time preceding the previous moment; This indicates the trigger phase of the preset Poincaré section.

[0010] Preferably, the construction of Poincaré feature points based on interpolated current values ​​includes: normalizing the interpolated current values ​​each time the motor crosses the Poincaré cross section according to the rated current range of the motor, obtaining the normalized current values ​​each time the motor crosses the Poincaré cross section, and defining the two-dimensional vector formed by the normalized current values ​​each time the motor crosses the Poincaré cross section and their difference values ​​as Poincaré feature points.

[0011] Preferably, the Poincaré feature points satisfy the expression, including: ;in, This is the index of the most recent crossing of the Poincaré section at the current time. Indicates the first A Poincaré feature point; Indicates the first Normalized current value when crossing the Poincaré section for the first time; Indicates the first Normalized current value when crossing the Poincaré section for the first time; This represents the transpose symbol.

[0012] This invention constructs a two-dimensional feature vector containing normalized current values ​​and their differences. The current value directly reflects the current viscosity level of the material, and the differential change between adjacent periods captures the rate of destruction or reconstruction of the internal microstructure of the material under continuous shear. At the same time, it reveals the transient resistance characteristics and dynamic evolution trend of the material. Compared with single load monitoring, it can more comprehensively characterize the thixotropic or fluid evolution law of the material, and provide a data foundation containing kinetic information for subsequent accurate assessment of the mixing state.

[0013] Preferably, the manifold discreteness satisfies the expression: ;in, This is the index of the most recent crossing of the Poincaré section at the current time. Indicates the first Manifold dispersion during the first crossing of the Poincaré section; Indicates the length of the sliding window; Indicates the index within the sliding window; Indicates the first A Poincaré feature point; This represents the geometric centroid of all Poincaré feature points within the current sliding window; Indicates Euclidean distance; Indicates the first The time decay weighting coefficients of each Poincaré feature point.

[0014] This invention measures the distribution characteristics of the system trajectory in phase space by calculating the weighted Euclidean distance between the Poincaré feature point and the geometric centroid within a sliding window, reflecting the stability of the material mixing state. It can effectively distinguish between the discrete state of the system caused by uneven component distribution in the early stage of mixing and the highly aggregated state of the system formed by homogenization in the later stage of mixing.

[0015] Preferably, the time decay weighting coefficient satisfies the expression: ;in, Indicates the forgetting factor; This represents an exponential function with the natural constant as its base.

[0016] This invention introduces an exponential function-based decay mechanism to assign nonlinear weights to historical Poincaré feature points with different time delays. This makes the calculation of manifold discreteness more focused on the system state near the current moment, while retaining the smoothing effect of historical data. This ensures that the control system can respond quickly to transient changes in material viscosity, avoid control lag caused by over-reliance on historical averages, and use historical information to suppress random noise interference, thereby achieving a balance between system response speed and stability.

[0017] Preferably, the target revolution speed for the next cycle satisfies the expression: ;in, Indicates the target orbital speed for the next cycle; This indicates the minimum basic rotational speed required to maintain the tank's revolution. Indicates the maximum allowable speed range of the equipment; This is the index of the most recent crossing of the Poincaré section at the current time. Indicates the first Manifold dispersion during the first crossing of the Poincaré section; This represents the preset steady-state threshold for the mixture; This represents an exponential function with the natural constant as its base.

[0018] This invention utilizes an exponential function to construct a nonlinear mapping relationship between manifold dispersion and target rotation speed. This allows the equipment to automatically maintain full speed operation in the early stages of material mixing to provide strong shear force to break up agglomerates. When the material tends to be homogeneous and stable, the rotation speed is smoothly reduced. This not only ensures mixing efficiency but also effectively avoids overheating or property degradation of the material due to continuous high shear after the mixing endpoint. It achieves on-demand energy supply and realizes the dual effects of energy saving and material protection.

[0019] Preferably, the method for obtaining the mixing steady-state threshold is as follows: when the equipment processes a new material for the first time, a full-cycle benchmark mixing experiment is performed, and the Poincaré feature points within the steady-state sampling time after the material reaches a homogeneous state in the benchmark mixing experiment are obtained. The arithmetic mean and standard deviation of the manifold dispersion within the steady-state sampling time are calculated, and the mixing steady-state threshold is calculated based on the arithmetic mean and standard deviation. ,in, For the mixed steady-state threshold, This is the arithmetic mean of the manifold dispersion over the steady-state sampling period; The standard deviation of the manifold dispersion over the steady-state sampling period; This is the confidence tolerance coefficient.

[0020] Preferably, it further includes: in response to the manifold dispersion being less than or equal to the mixing steady-state threshold and the duration exceeding a preset confirmation period, determining that the material has reached the homogenization endpoint, and controlling the homogenizer to stop.

[0021] The beneficial effects of this invention are as follows: By acquiring the load current and revolution phase of the main drive motor in real time and using the Poincaré section technique to perform interpolation sampling at a fixed phase, this invention directly filters out the periodic gravitational interference caused by revolution and rotation, avoiding the phase lag problem caused by traditional low-pass filtering, thereby obtaining characteristic data reflecting the intrinsic rheological properties of the material. Based on this, by calculating the manifold dispersion to measure the dispersion of Poincaré feature points in phase space, it can sensitively identify the evolution process of the material from non-uniform to uniform, overcoming the interference of strong mechanical noise on the determination of the mixing endpoint. Furthermore, this invention constructs an adaptive speed adjustment mechanism based on the manifold dispersion, providing strong shear force in the early stage of mixing to quickly disperse the material, and automatically reducing the speed in the later stage of mixing to reduce heat accumulation, thereby effectively preventing thermal damage to the material while ensuring mixing quality, realizing intelligent control and energy efficiency optimization of a non-intrusive homogenizer. Attached Figure Description

[0022] Figure 1 This is a flowchart illustrating a method for controlling the operating parameters of a non-invasive material homogenizer according to the present invention;

[0023] Figure 2 This is a schematic diagram of the original load current signal and the Poincaré cross-section sampling points collected in an embodiment of the present invention;

[0024] Figure 3 This is a schematic diagram of the evolution trajectory of the Poincaré section feature points in phase space in an embodiment of the present invention;

[0025] Figure 4 This is a schematic diagram of the manifold dispersion changing over time in an embodiment of the present invention. Detailed Implementation

[0026] 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.

[0027] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0028] This invention discloses a method for controlling the operating parameters of a non-invasive material homogenizer, referring to... Figure 1 This includes steps S1 to S4:

[0029] S1. Real-time acquisition of the load current of the main drive motor of the homogenizer and the revolution phase of the revolution arm to construct an operating dataset containing the sampling time, load current and revolution phase.

[0030] It should be noted that the operating state of a non-invasive homogenizer is characterized by both the kinetic variable, load current, and the kinematic variable, revolution phase. Load data alone cannot distinguish between changes in gravitational potential energy caused by mechanical attitude and changes in drag caused by material viscosity, making it difficult to accurately assess the material state based solely on load data. Therefore, this invention simultaneously acquires both load current and revolution phase.

[0031] Specifically, a high-frequency current transformer installed on the main drive motor controller is used to collect the motor's load current in real time at a preset sampling frequency; simultaneously, an absolute encoder installed on the revolution shaft is used to collect the real-time revolution phase of the revolution arm. The collected load current and revolution phase are aligned and encapsulated according to timestamps to construct an operational dataset containing the sampling time, load current, and revolution phase.

[0032] S2. Traverse the running dataset. When it is detected that the revolution phase of adjacent sampling points crosses the trigger phase of the preset Poincaré cross section, calculate the interpolated current value when crossing the Poincaré cross section based on the load current, revolution phase of adjacent sampling points and the trigger phase of the preset Poincaré cross section, and construct Poincaré feature points based on the interpolated current value.

[0033] It should be noted that the mechanical motion of a homogenizer is a typical periodic forced oscillation system, manifesting as a limiting loop orbit in phase space. Changes in the material mixing state alter the local topology of this limiting loop. Furthermore, direct observation is affected by periodic gravitational interference caused by the revolution and rotation. Therefore, this invention uses a Poincaré section set at a fixed revolution phase to perform stroboscopic sampling of the system, physically filtering out the periodic gravitational interference caused by the revolution and rotation. This ensures that the sampling points only reflect the energy state of the system under the same mechanical attitude.

[0034] Specifically, the trigger phase of the Poincaré section is set as follows: Its value range is to In this embodiment, to avoid the orbital phase data in and The critical point experiences a numerical jump, causing abnormal interpolation calculations. (Selection) In other embodiments, the implementer may set the trigger phase as a trigger phase according to the actual implementation situation.

[0035] Traverse the dataset and monitor the revolution phase of two adjacent sampling points. and satisfy At that time, it is determined that the system trajectory crosses the Poincaré section. Since digital sampling is discrete, to improve accuracy, linear interpolation is used to calculate the precise current value when crossing the Poincaré section:

[0036]

[0037] in, This represents the interpolated current value when passing through the Poincaré section; This represents the load current that has passed through the previous moment; This represents the load current at the moment after the crossing. Indicates the orbital phase at the moment after the crossing; This indicates the orbital phase at the point in time preceding the previous moment; This indicates the preset trigger phase.

[0038] For example, Figure 2 This is a schematic diagram of the original load current signal and the Poincaré cross-section sampling points collected in an embodiment of the present invention.

[0039] To ensure dimensional consistency in subsequent calculations, the interpolated current values ​​for each crossing of the Poincaré section are normalized according to the rated current range of the motor, resulting in normalized current values ​​for each crossing. The two-dimensional vector formed by the normalized current values ​​and their differences for each crossing is defined as the Poincaré feature points.

[0040]

[0041] in, This is the index of the most recent crossing of the Poincaré section at the current time. Indicates the first A Poincaré feature point; Indicates the first Normalized current value when crossing the Poincaré section for the first time; Indicates the first Normalized current value when crossing the Poincaré section for the first time; The transpose symbol is used; since the electromagnetic torque output by the motor is mainly used to overcome the viscous resistance torque of the material, and the viscous resistance torque is proportional to the viscosity of the material, therefore... The magnitude of the viscosity directly reflects the current viscosity level of the material, and the two are positively correlated; It represents the differential change in current within two adjacent revolution periods, reflecting the rate of destruction or reconstruction of the internal microstructure of the material under continuous shearing, and characterizing the thixotropic or fluidity evolution trend of the material.

[0042] S3. Based on multiple Poincaré feature points at the current and historical times, calculate the manifold dispersion, which is used to characterize the degree of dispersion of the Poincaré feature points in phase space.

[0043] It should be noted that in the initial stage of mixing, due to the uneven distribution of material components and the existence of complex nonlinear friction and collisions within the system, the projection point of the load current on the Poincaré section is extremely unstable, exhibiting a discrete state. As mixing progresses, the material gradually forms a homogeneous fluid with a uniform viscosity, and the dynamic behavior of the system converges to a stable fixed point or a very small attractive region. The Poincaré characteristic points exhibit high aggregation, and their evolution trajectory in two-dimensional phase space is as follows: Figure 3 As shown, as mixing proceeds, the Poincaré feature points gradually converge from the discrete state to the steady-state region. Therefore, the distribution characteristics of the Poincaré feature points in this invention construct the manifold discreteness, which is used to reflect the mixing quality of the material.

[0044] Specifically, the construction length is The sliding window selects the nearest one. Calculate the manifold discreteness using 1 Poincaré feature point:

[0045]

[0046] in, This is the index of the most recent crossing of the Poincaré section at the current time. Indicates the first Manifold dispersion during the first crossing of the Poincaré section; The length of the sliding window is set by the implementer based on the actual implementation situation. It is usually set to cover the number of points of 5 to 10 revolution cycles in order to avoid control lag while ensuring the significance of statistical characteristics. In this embodiment, the length of the sliding window is set to 8. Indicates the index within the sliding window; Indicates the first A Poincaré feature point; This represents the geometric centroid of all Poincaré feature points within the current sliding window, and signifies the current steady-state operating point. Indicates Euclidean distance; Indicates the first The time decay weighting coefficients of each Poincaré feature point. When the material is in a rapidly changing, non-uniform stage, the distribution of Poincaré feature points diverges, and the Euclidean distance... The larger value leads to a higher manifold dispersion. The Poincaré feature points are relatively large; when the material has reached a stable homogeneous state, the Poincaré feature points highly coincide in phase space, and the Euclidean distance is relatively large. Approaching 0, resulting in manifold dispersion It gradually decreases and approaches 0.

[0047] Preferably, the time decay weighting coefficient satisfies the expression:

[0048]

[0049] in, Indicates the first Time decay weighting coefficients for each Poincaré feature point; Indicates the index within the sliding window; This represents an exponential function with the natural constant as its base, used to assign nonlinear decay weights to historical Poincaré feature points with different time delays; This indicates the length of the sliding window. Let the index be the sequence number of the most recent crossing of the Poincaré section at the current time. The latest state of the corresponding system during the next traversal, when The smaller the hour, the more... The closer a Poincaré feature point is to the current time, the better. The larger the value, the greater the contribution of the Poincaré feature points near the current time to the assessment of manifold discreteness; The forgetting factor controls the time decay weight distribution of historical Poincaré feature points in the manifold discreteness calculation. Its empirical value is between 0.1 and 0.5. In other embodiments, implementers can adjust this value according to the required sensitivity to historical data. When a system needs to respond quickly to transient changes in material viscosity, the viscosity should be increased. To accelerate the forgetting of historical data, a value of 0.5 can be used; when signal noise is high and system stability needs to be improved, the value should be reduced. To utilize more historical data for smoothing, for example, set it to 0.1.

[0050] For example, Figure 4This is a schematic diagram of the manifold dispersion over time in an embodiment of the present invention. In the initial stage of mixing, due to the inhomogeneity of the material, the manifold dispersion value is large and there are violent fluctuations. As the viscosity of the material decreases and the homogeneity increases, the manifold dispersion shows a decreasing trend.

[0051] S4. Based on the manifold dispersion and the preset mixed steady-state threshold, determine the target revolution speed for the next cycle, and adjust the operating state of the main drive motor according to the target revolution speed.

[0052] It should be noted that material agglomeration is severe in the initial stage of mixing, requiring high shear force, while sustained high shear in the later stage of mixing can lead to thermal damage to the material. Therefore, this invention follows a variable structure logic, maintaining a high rotation speed when the manifold dispersion is large to provide strong shear force to break up the agglomerates, and automatically reducing the rotation speed when the manifold dispersion is small to reduce heat accumulation and prevent overmixing.

[0053] Specifically, based on the real-time calculated manifold dispersion, the target orbital speed for the next cycle is calculated:

[0054]

[0055] in, Indicates the target orbital speed for the next cycle; This indicates the minimum basic rotational speed required to maintain the tank's revolution. Indicates the maximum allowable speed range of the equipment; This is the index of the most recent crossing of the Poincaré section at the current time. Indicates the first Manifold dispersion during the first crossing of the Poincaré section; This represents the preset steady-state threshold for the mixture, which is a dimensionless constant. This represents an exponential function with the natural constant as its base.

[0056] This invention utilizes an exponential function to construct an adaptive adjustment mechanism, which adjusts the manifold dispersion when... Much greater than the mixed steady-state threshold When, the ratio Larger, exponential term Approaching 0, making Approaching The equipment runs at full speed to provide strong shear force; when the manifold dispersion Reduce to near the mixing steady-state threshold When, the ratio Decrease, exponent term Increase, making Smooth descent; when manifold dispersion When it approaches 0, the exponential term Approaching 1, making Approaching the minimum base speed This enables energy to be supplied on demand and prevents over-shearing.

[0057] It should be noted that, due to the significant differences in rheological properties among different material systems, the background noise levels of their homogenized fluid dynamics vary. Therefore, this invention uses material properties to determine the mixing steady-state threshold. The calibration.

[0058] Specifically, this embodiment uses a benchmark calibration method to determine the mixed steady-state threshold. When the equipment processes a new material for the first time, a full-cycle benchmark mixing experiment is performed, and offline detection methods are used to confirm the point in time when the material reaches a perfectly homogeneous state. Extracting time points from the baseline mixed experiment The Poincaré feature point data are then collected within a preset steady-state sampling period. This steady-state sampling period is set by the implementers based on the actual implementation situation and covers at least 10 revolution periods. The arithmetic mean and standard deviation of the manifold dispersion within this steady-state sampling period are calculated. Based on the arithmetic mean and standard deviation, the mixing steady-state threshold of the material is calculated.

[0059]

[0060] in, The mixed steady-state threshold; The arithmetic mean of the manifold discreteness over the steady-state sampling time after the material reaches a perfectly homogeneous state is used to characterize the inherent background discrete noise level of the system under ideal homogeneous conditions. The standard deviation of the manifold dispersion during the steady-state sampling time after the material reaches a perfectly homogeneous state characterizes the fluctuation amplitude under steady state. The confidence tolerance coefficient is typically set to 3 to 5 to ensure that the system can reliably identify the steady state under slight disturbances in the industrial environment. In this embodiment, the confidence tolerance coefficient is set to 2. In other embodiments, the implementer can set it according to the noise disturbance level of the field environment or the anti-interference capability requirements for the system's steady state identification.

[0061] When the manifold dispersion is less than or equal to the mixing steady-state threshold and the duration exceeds the preset confirmation period, it is determined that the material has reached the homogenization endpoint, and the control system outputs a shutdown command. In this embodiment, the confirmation period is set to 5 times the time required for the homogenizer's orbital arm to rotate once. In other embodiments, the implementer can set the confirmation period according to the rheological properties of the material or the reliability requirements for determining the mixing endpoint.

Claims

1. A method of operating parameter control for a non-invasive material homogenizer, characterized by, The method comprises the following steps: Real-time acquisition of the load current of the main drive motor of the homogenizer and the revolution phase of the revolving arm, and construction of a running data set containing the sampling time, the load current and the revolution phase; Traverse the running data set, and when it is monitored that the revolution phases of adjacent sampling points cross the trigger phase of the preset Poincare section, calculate the interpolation current value when the Poincare section is crossed according to the load currents and the revolution phases of the adjacent sampling points and the trigger phase of the preset Poincare section, and construct a Poincare characteristic point based on the interpolation current value; the interpolation current value satisfies: ； represents an interpolated current value when crossing the Poincare section; , respectively represent the load current at the previous time, the next time; represents the revolution phase at the next time of crossing; represents the revolution phase at the previous time of crossing; represents the preset trigger phase of the Poincare section; The Poincare characteristic point satisfies: ； This is the index of the most recent crossing of the Poincaré section at the current time. Indicates the first A Poincaré feature point; , They represent the first sequence Normalized current value when crossing the Poincaré section for the first time; Indicates the transpose symbol; Based on the plurality of Poincare characteristic points at the current time and the historical time, calculate the manifold dispersion, which is used to represent the dispersion degree of the Poincare characteristic points in the phase space; the manifold dispersion satisfies: ； represents the manifold dispersion at the first crossing of the Poincare section; represents the length of the sliding window; represents the index within the sliding window; represents the Poincare feature point at the first crossing of the Poincare section; represents the geometric centroid of all Poincare feature points within the current sliding window; represents the Euclidean distance; represents the time decay weight coefficient of the Poincare feature point at the first crossing of the Poincare section; The time decay weight coefficient satisfies: ； denotes a forgetting factor; denotes an exponential function with a natural constant as base According to the manifold dispersion and the preset mixed steady state threshold, determine the target revolution speed of the next period, which satisfies: ； represents a target revolution speed of the next cycle; represents a minimum base revolution speed required to maintain the revolution of the tank; represents a maximum speed regulation range allowed by the device; represents a preset mixing steady state threshold; Adjust the running state of the main drive motor according to the target revolution speed.

2. The method of claim 1, wherein the method further comprises: The real-time acquisition of the load current of the main drive motor of the homogenizer and the revolution phase of the revolving arm, and the construction of the running data set containing the sampling time, the load current and the revolution phase, comprise: The load current of the main drive motor is acquired by a current transformer at a preset sampling frequency; the revolution phase of the revolving arm is acquired by an absolute value encoder; and the load current and the revolution phase are aligned by time stamp to generate the running data set.

3. The method of claim 1, wherein the method further comprises: The acquisition method of the mixed steady state threshold is: When the device processes a new material for the first time, a full-cycle reference mixing experiment is performed to obtain Poincare characteristic points in a steady-state sampling duration after the material reaches a homogeneous state in the reference mixing experiment, to calculate an arithmetic mean and a standard deviation of manifold dispersion in the steady-state sampling duration, and to calculate a mixing steady-state threshold based on the arithmetic mean and the standard deviation: wherein, is the mixing steady-state threshold, is the arithmetic mean of the manifold dispersion in the steady-state sampling duration; is the standard deviation of the manifold dispersion in the steady-state sampling duration; is a confidence tolerance coefficient.

4. The method of claim 1, wherein, Further comprising: In response to the manifold dispersion being less than or equal to the mixed steady state threshold and the duration exceeding the preset confirmation period, it is determined that the material reaches the homogenization endpoint, and the homogenizer is controlled to stop.

Images