Rice mill pressure self-adaptive control method, system and equipment based on artificial intelligence and medium

Abstract

The invention discloses a rice mill pressure self-adaptive control method, system and equipment based on artificial intelligence and a medium, and relates to the technical field of industrial control systems.The method comprises the steps that the opening degree of a first rice outlet is set; setting a first feeding speed according to the opening degree of the first rice outlet; defining a control set; obtaining a plurality of first control sets; obtaining a preset first pressure, and screening the plurality of first control sets to obtain a plurality of second control sets conforming to the first pressure; controlling the rice husking machine to mill rice according to the second control set, performing feature extraction on the rice husking process of the rice husking machine, and generating a first rice husking evaluation value; a plurality of second control sets under the first pressure and the corresponding first rice milling evaluation values are mapped to a preset control set-rice milling evaluation display model, a third control set is generated, the third control set is applied to rice milling, and the problem that under the same target rice milling pressure, different control target combinations have multi-source characteristic coupling influences, and rice milling efficiency is improved is solved. And the rice milling quality is not stable.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of industrial control systems, and more particularly, to a pressure self-adaptive control method, system, device and medium for a rice mill based on artificial intelligence. BACKGROUND

[0002] In the modern rice processing industry, the rice mill as the core equipment, its running performance is directly related to the key indicators such as the yield, broken rice rate, milling degree and energy consumption. In the existing rice milling process, pressure control as an important adjusting means in the milling process, its regulation effect directly determines the quality and production efficiency of the processed products.

[0003] At present, the industrial rice milling equipment generally realizes indirect control of the internal pressure of the milling chamber by controlling the adjustable structural parameters such as the opening degree of the rice outlet and the feeding speed. Smaller opening degree of the rice outlet helps to improve the internal pressure and enhance the milling effect, but it is easy to cause rice blocking, increase the broken rice rate and increase the energy consumption; although larger opening degree can improve the rice discharge rate and system smoothness, it may lead to insufficient milling. Similarly, increasing the feeding speed can enhance the milling pressure per unit time and improve the yield, but it may also cause system load fluctuation, heat accumulation and structural wear problems; while the feeding speed is too low, the efficiency is reduced, and the milling stability is poor.

[0004] The pressure control target of the rice mill is essentially to precisely control the internal pressure in a reasonable range under the requirements of different rice varieties, moisture content, target milling degree, etc., in order to balance the rice quality, energy efficiency and mechanical life. However, in the current technology, the pressure control process has the following deficiencies: Both of them have a profound impact on the broken rice rate, milling degree, response speed and energy consumption while regulating the pressure: when the feeding speed is set too high, although it can improve the response ability and processing rate, it is easy to cause instantaneous pressure to be too large, the broken rice rate to rise and the energy consumption to fluctuate; when the opening degree of the rice outlet is set too small, the system anti-blocking ability decreases, the energy load increases, and the equipment wear is intensified; on the contrary, when both of them are set too low or too wide, the pressure forming ability is weakened, which leads to insufficient milling or reduced system running efficiency. Therefore, how to reasonably select the control combination to optimize the milling quality, system energy consumption and running stability under the premise of meeting the target pressure has not been considered in the existing technology.

[0005] In view of the above problems, the present application provides a solution. SUMMARY

[0006] In order to overcome the above-mentioned defects of the prior art, embodiments of the present application provide an artificial intelligence-based rice mill pressure adaptive control method, system, device and medium, which constructs a mapping model between control parameters and rice milling results, and introduces a multi-feature evaluation mechanism to achieve optimal screening of different control combinations under the same pressure target, thereby solving the problem of unstable rice milling quality caused by multi-source characteristic coupling of different control target combinations under the same target rice milling pressure.

[0007] To achieve the above object, the present application provides the following technical solutions: An artificial intelligence-based rice mill pressure adaptive control method, comprising: constructing a rice variety experience curve to set a first rice outlet opening degree; setting a first feeding speed according to the first rice outlet opening degree to meet a preset reference impedance as a target; defining a control set, the control set comprising one rice outlet opening degree and one feeding speed; performing Latin hypercube sampling based on a low-dimensional control space on the first rice outlet opening degree and the first feeding speed, and randomly combining to obtain a plurality of first control sets; obtaining a preset first pressure, and screening the plurality of first control sets according to a pre-obtained mapping relationship between the control set and the pressure to obtain a plurality of second control sets meeting the first pressure; controlling the rice mill to mill rice according to the second control set, and extracting features of the rice mill milling process to generate a first rice milling evaluation value; mapping the plurality of second control sets under the first pressure and the corresponding first rice milling evaluation value to a preset control set-rice milling evaluation display model, and combining a preset rice milling mode to perform data analysis to generate a third control set for application in rice milling; and the mapping relationship between the control set and the pressure is fitted by polynomial regression modeling of the control set and the pressure.

[0008] In a preferred embodiment, the rice variety experience curve is used to set the first rice outlet opening degree, specifically: the first rice outlet opening degree required is back calculated according to the rice variety experience curve and a preset system reference minimum stable rice outlet speed; the rice variety experience curve is pre-constructed, specifically: the characteristic parameters of the rice sample are obtained, a pre-trained lightweight clustering model is used to determine the type of the rice variety, and the corresponding rice variety label is output; based on the identified rice variety label, the initial linear response relationship curve of the rice outlet opening degree-rice outlet speed is constructed by combining the real-time feedback of the current initial processing rice outlet pressure sensor trend and the rice outlet quantity; and the initial linear response relationship curve is locally fitted to form the experiential response curve corresponding to the specific rice variety.

[0009] In a preferred implementation, the first feeding speed is set according to the first outlet opening degree, aiming at a preset reference impedance, specifically: the reference impedance is the impedance under standard grain measurement, and the impedance is the ratio of motor load change and discharge pressure feedback change; a reference feeding rate under standard grain measurement is obtained, and the product of the reference feeding rate and the reference impedance is the same as the product of the first outlet opening degree and the first feeding speed, so as to obtain the first feeding speed.

[0010] In a preferred implementation, the first outlet opening degree and the first feeding speed are subjected to Latin hypercube sampling based on a low-dimensional control space, and are randomly combined to obtain a plurality of first control sets, specifically: the first outlet opening degree and the first feeding speed are respectively pushed in a preset step along the lifting direction and the weakening direction for a plurality of times to construct a variable variation sequence, specifically: taking the first outlet opening degree as a starting point, sequentially lifting a plurality of times according to a preset opening degree step to form an opening degree lifting sequence; at the same time, sequentially reducing a plurality of times based on the starting point to form an opening degree weakening sequence; similarly, taking the first feeding speed as a starting point, sequentially pushing a plurality of times according to a preset feeding step to form a feeding lifting sequence; and sequentially pushing a plurality of times to form a feeding weakening sequence; the four sequences are combined to form an outlet opening degree variable set and a feeding speed variable set.

[0011] In a preferred implementation, the first pressure is a demand pressure, and a pressure that the system needs to maintain; according to a mapping relationship between the control set and the pressure obtained in advance, specifically: under a plurality of control combinations, the pressure value when the system is running is collected to construct a training sample set; each sample contains a pressure response value corresponding to a control set; the dimension of the control set is set, the order of the polynomial regression is selected, and a polynomial model is generated; a feature matrix and a target vector are constructed, the feature matrix is a matrix expressed by the control set, and the target vector is a corresponding pressure vector, the feature matrix and the target vector are mapped to the polynomial model; the least square method is used for training to solve the weight of the best polynomial model, and the mapping relationship between the control set and the pressure is obtained.

[0012] In a preferred implementation, the feature extraction includes a broken rice rate feature, a system working fluency feature, and a rice milling stability feature; the specific acquisition method of the system working fluency feature is as follows: after each control set runs for a preset first time, the coefficient of variation of the speed error in a unit time is obtained to represent the discharge stability; the mean square error of the adjacent control instruction response time is obtained to represent the response consistency; the two are combined into a two-dimensional stability-consistency feature point, and the Euclidean distance of the feature point to the origin is taken as the system working fluency feature value.

[0013] In a preferred embodiment, the second control set under the first pressure and the corresponding first rice milling evaluation value are mapped to a preset control set-rice milling evaluation display model, specifically: each control set is bound with its corresponding first rice milling evaluation value, so that each point has a matching relationship between control parameter combination and rice milling effect; the control set-rice milling evaluation display model is an empty distribution space with relative positional relationship; the second control set under the first pressure and the corresponding first rice milling evaluation value are mapped to the preset control set-rice milling evaluation display model, and the discrete points are converted into a visual continuous area through local trend modeling; the preset rice milling mode includes a mark of control target preference and a first range of preset first rice milling evaluation value, preference opening degree and preference feeding speed; under the condition that the first range is met in the continuous area, the control set corresponding to the point of the peak of the region with the largest preference is selected as the third control set, which is applied to rice milling.

[0014] A system of a rice mill pressure adaptive control method based on artificial intelligence, comprising an initialization module, a control set generation module, a pressure mapping and screening module, a control evaluation module and a control application module; the initialization module is used to construct a rice variety experience curve to set a first rice outlet opening degree; according to the first rice outlet opening degree, a first feeding speed is set to meet a preset reference impedance; the control set generation module is used to perform Latin hypercube sampling based on a low-dimensional control space on the first rice outlet opening degree and the first feeding speed, and randomly combine to obtain a plurality of first control sets; the pressure mapping and screening module is used to obtain a preset first pressure, and screen the plurality of first control sets according to the pre-obtained mapping relationship between the control set and the pressure to obtain a plurality of second control sets meeting the first pressure; the control evaluation module is used to control the rice mill to mill rice according to the second control set, extract features from the rice milling process of the rice mill, and generate a first rice milling evaluation value; the control application module is used to map the plurality of second control sets under the first pressure and the corresponding first rice milling evaluation value to a preset control set-rice milling evaluation display model, and combine a preset rice milling mode to perform data analysis and generate a third control set applied to rice milling.

[0015] An electronic device, comprising: at least one processor; and a memory connected in communication with the at least one processor; wherein the memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor to enable the at least one processor to execute a rice mill pressure adaptive control method based on artificial intelligence.

[0016] A computer readable storage medium stores a computer program, and the computer program is executed by a processor to implement a rice mill pressure adaptive control method based on artificial intelligence.

[0017] The technical effects and advantages of the rice mill pressure self-adaptive control method, system, device and medium based on artificial intelligence of the present application are as follows: 1. The present application constructs a precise empirical rice outlet opening response curve based on a lightweight clustering model of rice characteristics, dynamically sets the feeding speed in combination with the reference impedance matching, realizes intelligent self-adaptive adjustment of the rice outlet opening and the feeding speed, effectively covers the control parameter space by using Latin hypercube sampling, accurately fits the relationship between the control set and the pressure by using polynomial regression, constructs a neural network evaluation model in combination with multi-dimensional feature extraction, realizes comprehensive prediction and optimization of the rice milling quality and system stability, finally intuitively assists the selection of the control strategy through the mapping display model, improves the parameter space exploration and model fitting efficiency, the multi-dimensional feature evaluation enhances the accurate control of the rice milling quality and system performance, the mapping and visual analysis improve the intelligent decision level of the control strategy, and effectively promotes the stability, efficiency and product quality improvement of the rice milling process. BRIEF DESCRIPTION OF DRAWINGS

[0018] Figure 1 The flowchart of the rice mill pressure self-adaptive control method based on artificial intelligence of the present application is shown. Figure 2 The structure diagram of the rice mill pressure self-adaptive control system based on artificial intelligence of the present application is shown. DETAILED DESCRIPTION

[0019] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0020] Embodiment 1, Figure 1 The rice mill pressure self-adaptive control method based on artificial intelligence of the present application is given, including the following steps: S1, constructing a rice experience curve to set the first rice outlet opening.

[0021] In this embodiment, the rice experience curve is constructed to set the first rice outlet opening, specifically: According to the rice experience curve and the preset system reference minimum stable rice rate, the required opening degree of the first rice outlet is backstepped; The rice experience curve is pre-constructed, specifically: Obtain the characteristic parameters of the rice sample, use the pre-trained lightweight clustering model to determine the type of the rice, and output the corresponding rice label; Based on the identified rice variety label, combined with the current processing initial stage of the rice outlet pressure sensor change trend and the real-time feedback of the rice outlet quantity, the initial linear response relationship curve of the rice outlet opening degree-rice outlet rate is constructed. The initial linear response relationship curve is locally fitted to form an empirical response curve corresponding to a specific rice variety.

[0022] It should be noted that the lightweight clustering model is used to quickly cluster and classify the input rice sample images or sensor parameters to generate a rice variety label. This model can use improved K-means, Mini-Batch K-means, or a lightweight model structure with sparse constraints Gaussian mixture model, and perform clustering operations on embedded representations of key features such as particle size, color, surface texture, and infrared reflectivity, to achieve rice variety classification without a large amount of labeled data. The model has short inference time and low resource consumption, and is suitable for deployment on edge side or mill end control chips.

[0023] It should be noted that the construction of the rice variety empirical curve belongs to the existing mature technical field, and its essence is a function fitting process of multiple input openings and corresponding rice outlet rates. Local regression, spline interpolation, or nonlinear fitting based on least squares can be used to complete it. Its application in the rice milling control process has been verified in many grain processing intelligent systems, and the specific implementation method is well known to those skilled in the art, which will not be described here in this embodiment.

[0024] It should be noted that the main purpose of local fitting processing is to solve the non-uniform response problem between opening degree and rice outlet rate. In some intervals, such as the small opening degree stage, the flow of rice grains is affected by friction and accumulation, showing a nonlinear growth trend. If a global linear or polynomial model is used, it is easy to mask the local variation law, resulting in a decrease in control accuracy. Through local fitting, the slope, turning point, and saturation interval can be more accurately described, providing more accurate mapping basis for subsequent initial opening degree backstepping.

[0025] S2, according to the first rice outlet opening degree, set the first feeding speed to meet the preset reference impedance as the target.

[0026] In this embodiment, according to the first rice outlet opening degree, set the first feeding speed to meet the preset reference impedance as the target, specifically: The reference impedance is the impedance under standard grain measurement, and the impedance is the ratio of motor load change and discharge pressure feedback change; Obtain the reference feeding rate under standard grain measurement, so that the product of the reference feeding rate and the reference impedance is the same as the product of the first rice outlet opening degree and the first feeding speed, to obtain the first feeding speed.

[0027] It should be noted that the product of the reference feeding rate and the reference impedance represents the energy flux of the system at a unit opening degree or the feeding efficiency at a unit resistance; more specifically, the product embodies the "feeding control strategy under the same load impedance level per unit opening area"; by ensuring that the product remains consistent between the standard working condition (reference feeding rate and reference impedance) and the actual working condition (first discharge opening degree and first feeding speed), the system load can be stabilized, and motor overload or idling caused by excessive or insufficient feeding can be avoided; the balance of the discharge pressure is maintained, reducing the risk of blockage or poor feeding; the impedance disturbance caused by different types of grain or changes in particle size is adapted, and the feeding speed is dynamically adjusted when the structural opening degree is constant. This is actually a feeding speed adjustment model based on impedance matching and force-energy conversion conservation.

[0028] It should be noted that the motor load change signal is obtained through a current sensor, and the discharge pressure is fed back in real time through a pressure sensor. After filtering and fitting, the impedance ratio is obtained.

[0029] In this embodiment, the significance of setting the first discharge opening degree and the first feeding speed is: The first discharge opening degree and the first feeding speed constitute the initial control state of the entire rice milling control process, directly affecting the running stability and load characteristics of the rice mill in the initial stage, and also determining the center point and sampling boundary of the subsequent control space construction. Specifically: On the one hand, the initial control combination serves as the reference center for generating the control set (such as Latin hypercube sampling), and its physical reasonableness determines whether the sampling space covers the effective area. If the initial value is set too far from the actual load characteristics of the system or the processing characteristics of the rice, it may result in a large number of invalid combinations in the subsequent sampling control set (such as abnormal pressure, excessive or insufficient feeding), reducing the sampling efficiency and evaluation value.

[0030] On the other hand, the mapping relationship between the control set and the pressure (such as model training based on polynomial regression) essentially relies on the combination samples derived from the initial control point to construct the function space. Therefore, the higher the physical credibility of the initial point, the better the fitting quality of the subsequent mapping relationship, and the easier it is to form a response model with small fitting error and strong generalization ability.

[0031] In addition, the initial control combination also provides a stable reference point for subsequent multi-dimensional evaluation (such as broken rice rate, blocked rice rate, load fluctuation, etc.) under a fixed pressure target, facilitating the comparison of different control combinations on the same physical benchmark.

[0032] In summary, reasonable setting of the first discharge opening degree and the first feeding speed not only ensures the safety of the initial state of the system, but also improves the representativeness and coverage of the control set construction, enhances the convergence and accuracy of the pressure mapping modeling, and ultimately improves the efficiency and quality of the subsequent optimal combination selection.

[0033] S3, defining a control set comprising a discharge opening and a feeding speed.

[0034] S4, Latin hypercube sampling based on a low-dimensional control space for the first discharge opening and the first feeding speed, and randomly combining to obtain a plurality of first control sets.

[0035] In this embodiment, the first control set is used to construct a control space boundary to limit the sampling range.

[0036] In this embodiment, Latin hypercube sampling based on a low-dimensional control space is performed on the first discharge opening and the first feeding speed, and randomly combined to obtain a plurality of first control sets, specifically: The first discharge opening and the first feeding speed are respectively pushed in the lifting direction and the weakening direction for a plurality of times with a preset step size to construct a variable variation sequence, specifically: Taking the first discharge opening as the starting point, the opening is sequentially lifted for a plurality of times according to a preset opening step size to form an opening lifting sequence; at the same time, the opening is sequentially reduced for a plurality of times based on the starting point to form an opening weakening sequence; Similarly, taking the first feeding speed as the starting point, the feeding speed is sequentially pushed up for a plurality of times according to a preset feeding step size to form a feeding lifting sequence; at the same time, the feeding speed is sequentially pushed down for a plurality of times to form a feeding weakening sequence; The four sequences are combined to form a discharge opening variable set and a feeding speed variable set, respectively.

[0037] The obtained data is randomly combined to generate a plurality of first control sets, specifically: The discharge opening variable set and the feeding speed variable set are respectively subjected to full permutation combination, and a plurality of combinations are randomly selected to form a first control set, each first control set comprising a group of discharge openings and a group of feeding speeds, for representing different pressure input control states.

[0038] It should be noted that the application of Latin hypercube sampling in a low-dimensional control space can ensure uniform coverage while avoiding excessive redundancy and invalid combinations, significantly improving the efficiency of parameter space exploration. Especially in the initial stage, smaller but efficient combined samples can provide stable data support for subsequent quality evaluation models, avoiding the deviation of the overall optimization direction caused by improper initial value selection.

[0039] It should be noted that the random combination operation is not completely disordered, but under the premise of ensuring the coverage characteristics of Latin hypercube sampling, by setting the disturbance threshold between control variables, it can avoid generating too concentrated or boundary extreme control combinations, thereby improving the representativeness and diversity of initial sampling.

[0040] S5, obtaining a preset first pressure, screening a plurality of first control sets according to a mapping relationship between the control sets and the pressure obtained in advance, to obtain a plurality of second control sets meeting the first pressure.

[0041] In the embodiment, the mapping relationship between the control sets and the pressure is modeled by polynomial regression to fit the control sets and the pressure.

[0042] In the embodiment, the preset first pressure is a demand pressure and a pressure required to be maintained by the system. In the embodiment, the mapping relationship between the control sets and the pressure obtained in advance is specifically: Under a plurality of control combinations, pressure values when the system runs are collected to construct a training sample set. Each sample contains a pressure response value corresponding to a control set. The dimension of the control set is set, the order of polynomial regression is selected, and a polynomial model is generated. A feature matrix and a target vector are constructed, the feature matrix is a matrix expressed by the control set, the target vector is a corresponding pressure vector, the feature matrix and the target vector are mapped to the polynomial model. The least square method is used for training to solve the weight of the optimal polynomial model, and the mapping relationship between the control set and the pressure is obtained.

[0043] The form of the polynomial model is as follows: ; Wherein, , and the like are to be trained weights, x is the opening degree of the rice outlet, y is the feeding speed, and P is the pressure.

[0044] It should be noted that the order of polynomial regression controls the length of the polynomial.

[0045] S6, controlling the rice mill to mill rice according to the second control set, and performing feature extraction on the rice milling process of the rice mill and generating a first rice milling evaluation value.

[0046] In the embodiment, the feature extraction includes broken rice rate feature, system working smoothness feature and rice milling stability feature. The broken rice rate feature is one of the key indicators for measuring the rice integrity in the rice milling process, and is used to reflect the change trend of the degree of rice breakage under the current rice milling control parameters. In the embodiment, the broken rice rate feature can be quantitatively extracted by image recognition, rice particle size statistics or weight ratio detection, etc. The feature is not only affected by factors such as milling pressure, flow rate, roller speed, etc., but also related to the quality of the raw rice, and has strong sensitivity and discrimination. Therefore, the broken rice rate feature can be used as an important reference for evaluating the advantages and disadvantages of the control set, and is used to support the subsequent polynomial regression model construction and rice milling performance feedback optimization.

[0047] The specific acquisition method of the broken rice rate feature is as follows: A high-speed industrial camera is arranged at the rice outlet of the rice mill to collect image sequences of the rice outlet process at a fixed frame rate; The collected images are subjected to grayscale, binarization and morphological filtering operations to remove background and noise and highlight the contours of the rice particles; The rice particle regions are extracted based on Canny edge detection and connected region labeling; The length-to-width ratio of each particle is calculated, and it is judged whether it is less than a preset broken rice threshold to make a broken rice judgment, and the broken rice rate feature is calculated by comparing the number of broken rice and the total number of rice.

[0048] It should be noted that analyzing the broken rice rate feature has the following advantages for evaluating the combination of the size of the rice outlet opening and the discharge speed and for the final optimization: Reveal the sensitivity of different parameters to the broken rice rate: through the polynomial regression modeling of the broken rice rate feature, the change trend of the broken rice rate under different rice outlet opening and discharge speed settings can be quantitatively analyzed, and it can be judged which parameter has a more significant impact on the broken rice rate. For example, even if the discharge speed is increased under a large opening, the broken rice rate tends to be stable, while the speed change under a small opening will cause the broken rice rate to rise significantly.

[0049] Realize the priority sorting of parameter regulation: through the model regression coefficient and residual analysis, the contribution degree of the broken rice rate control can be established, so as to judge whether the opening or the discharge speed should be adjusted first under different working conditions to realize a more optimal control strategy.

[0050] Enhance the pertinence and accuracy of the optimization control: taking the broken rice rate feature as the "broken rice risk evaluation index", combined with the opening-speed combination generation mechanism, the high broken rice rate combination can be quickly excluded, and the production efficiency under the guarantee of rice quality can be improved.

[0051] Strong portability, easy to adapt to different devices: the feature evaluation method does not depend on specific hardware structure, and can be adapted to different types of discharge mechanisms. Only the initial parameters need to be calibrated, and the rapid migration of control strategies for multiple models can be realized.

[0052] The system working fluency feature is used to represent whether the material transmission in the rice milling process is smooth, and the core reflects the flow state of the raw rice in the whole channel from the feeding port to the discharging port. The feature can be quantitatively evaluated by real-time monitoring of the frequency, duration and distribution of the blocked rice, and usually takes the "blocked rice rate" as the key indicator. A higher blocked rice rate often means that there are problems such as excessive feeding amount, improper rice opening adjustment or poor discharging in the rice milling process, resulting in uneven internal pressure of the rice milling cavity and fluctuation of the milling load, thereby affecting the efficiency and stability of the overall system. Therefore, the system working fluency feature can not only be used to identify the running bottleneck under the current working condition, but also provide quantitative reference basis for subsequent parameter adjustment.

[0053] The specific acquisition method of the system working fluency feature is as follows: After each control set runs for a preset first time, the coefficient of variation of the speed error in unit time is obtained, representing the discharging stability; The mean square error of the response time of adjacent control instructions is obtained, representing the response consistency; The two are combined into a two-dimensional stability-consistency feature point, and the Euclidean distance from the feature point to the origin is taken as the system working fluency feature value, which is specifically: ; It should be noted that the two-dimensional stability-consistency feature point is , is the coefficient of variation of the speed error in unit time, is the mean square error of the response time of adjacent control instructions, is the system working fluency feature value.

[0054] It should be noted that the discharging stability (coefficient of variation of speed error in unit time) and the response consistency (mean square error of control instruction response time) are combined into a two-dimensional feature point, and the Euclidean distance from the feature point to the origin is taken as the feature value of the system working fluency. The meaning is that the origin represents the target state of "the strongest stability and the highest response consistency" of the system under the ideal working condition (i.e. the variation coefficient and the mean square error are both 0), and the smaller the distance between the feature point and the origin, the closer the system running to the ideal state, and the higher the overall fluency. The Euclidean distance has good similarity ability in the measurement space, can naturally reflect the joint influence of the two indicators on the system performance without weighting, avoids introducing bias due to subjective weight setting, and maintains the simplicity of the model and the stability of the calculation.

[0055] The rice milling stability feature is used to reflect the mechanical output consistency and load state fluctuation law of the rice mill during operation, and mainly measures whether the rice milling system can maintain stable processing pressure and power output when processing rice with different moisture content, particle size or hardness. This feature is usually achieved by analyzing the torque change of the rice milling motor, the load current fluctuation or the standard deviation of the pressure curve in the whitening cavity, which can effectively identify the unstable state caused by rice blockage, intermittent feeding or abnormal rice particles during processing. Specifically, the dynamic consistency of the rice milling process can be quantitatively modeled by extracting the instantaneous load change amount per unit time, frequency response characteristics or periodic disturbance indicators, providing a stability score basis for subsequent quality evaluation and working condition adjustment.

[0056] The specific method for obtaining the rice milling stability feature is as follows: Collect the load torque signal of the rice mill motor within a predetermined time; Denoising the load torque signal to remove high-frequency interference; Slide with a fixed window length on the load signal, calculate the first standard deviation in each window; Calculate the first average value in each window, and take the ratio of the first standard deviation and the first average value as the rice milling stability feature value.

[0057] In this embodiment, analyzing the rice milling stability feature for evaluating the combination of the rice outlet opening size and the discharge speed size and for the final optimization has the following advantages: By sliding window method to obtain the ratio of standard deviation and average value of load torque signal, the degree of load fluctuation in the rice milling process can be effectively reflected, that is, the stability of the system operation. When the ratio is small, the motor load fluctuation is small, the rice milling process is stable, which indicates that the current rice outlet opening size and discharge speed match better, and the rice milling quality is more controllable and the broken rice rate is lower. When the ratio is large, there is obvious fluctuation, which may be caused by uneven feeding or unreasonable opening setting, which has adverse effects on the final product quality. Therefore, by using the stability feature value, a better control parameter combination can be selected to realize intelligent control and improve the overall performance of the rice milling system and the product consistency.

[0058] Further, a first rice milling evaluation value is generated, specifically: Based on the extracted features, an initial evaluation model is constructed, and the initial evaluation model is trained and solved based on a neural network to obtain a solved evaluation model. The input of the solved evaluation model is two parameters of the control set, and the output is the first rice milling evaluation value.

[0059] S7, mapping the second control sets under the first pressure and the corresponding first rice milling evaluation values to a preset control set-rice milling evaluation display model, and combining a preset rice milling mode to perform data analysis to generate a third control set, which is applied to rice milling.

[0060] In this embodiment, the second control sets under the first pressure and the corresponding first rice milling evaluation values are mapped to a preset control set-rice milling evaluation display model, specifically: Each control set is bound to its corresponding first rice milling evaluation value, so that each point has a matching relationship between the control parameter combination and the rice milling effect; The control set-rice milling evaluation display model is an empty distribution space with relative positional relationship; The second control sets under the first pressure and the corresponding first rice milling evaluation values are mapped to a preset control set-rice milling evaluation display model, and the discrete points are converted into a visual continuous region through local trend modeling; The preset rice milling mode includes a mark of control target preference and a first range of preset first rice milling evaluation values, preference opening degree and preference feeding speed; In the case where the first range is met in the continuous region, the control set corresponding to the point of the region peak with the largest preference is selected as the third control set, which is applied to rice milling.

[0061] Embodiment 2, Figure 2 An artificial intelligence-based rice mill pressure adaptive control system is given, which includes an initialization module, a control set generation module, a pressure mapping and screening module, a control evaluation module, and a control application module; The initialization module is used to construct a rice variety experience curve to set the first rice outlet opening degree; and according to the first rice outlet opening degree, a first feeding speed is set to meet a preset reference impedance as a target; The control set generation module is used for Latin hypercube sampling based on a low-dimensional control space for the first rice outlet opening degree and the first feeding speed, and randomly combining to obtain a plurality of first control sets; The pressure mapping and screening module is used to obtain a preset first pressure, and according to the mapping relationship between the control set and the pressure obtained in advance, the plurality of first control sets are screened to obtain a plurality of second control sets meeting the first pressure; The control evaluation module is used to control the rice mill according to the second control set, and to extract features from the rice milling process of the rice mill and generate a first rice milling evaluation value; The control application module is used to map the second control sets under the first pressure and the corresponding first rice milling evaluation values to a preset control set-rice milling evaluation display model, and combine a preset rice milling mode to perform data analysis to generate a third control set, which is applied to rice milling.

[0062] The application also includes an electronic device, comprising: at least one processor; and, a memory connected with the at least one processor in communication; wherein, the memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor to enable the at least one processor to execute the artificial intelligence-based pressure adaptive control method of a rice mill.

[0063] The application also includes a computer-readable storage medium storing a computer program, and the computer program is executed by a processor to implement the artificial intelligence-based pressure adaptive control method of a rice mill.

Claims

1. An artificial intelligence-based pressure self-adaptive control method for a rice mill, characterized in that, The application relates to a rice mill control method, comprising the following steps: constructing a rice variety experience curve to set a first rice outlet opening degree; setting a first feeding speed according to the first rice outlet opening degree, aiming at meeting a preset reference impedance; defining a control set comprising a rice outlet opening degree and a feeding speed; performing Latin hypercube sampling on the first rice outlet opening degree and the first feeding speed based on a low-dimensional control space, and randomly combining to obtain a plurality of first control sets; obtaining a preset first pressure, screening the plurality of first control sets according to a previously obtained mapping relationship between the control set and the pressure, and obtaining a plurality of second control sets meeting the first pressure; controlling rice milling of a rice mill according to the second control set, and performing feature extraction on the rice milling process of the rice mill to generate a first rice milling evaluation value; mapping the plurality of second control sets under the first pressure and the corresponding first rice milling evaluation value to a preset control set-rice milling evaluation display model, and combining a preset rice milling mode to perform data analysis and generate a third control set applied to rice milling. The mapping relationship between the control set and the pressure is fitted by polynomial regression modeling.

2. The artificial intelligence-based rice mill pressure self-adaptive control method according to claim 1, characterized in that, The method for constructing the rice variety experience curve to set the first rice outlet opening degree comprises the following steps: backstepping the required opening degree of the first rice outlet according to the rice variety experience curve and a preset system reference minimum stable rice outlet speed; The rice variety experience curve is pre-constructed, and the method comprises the following steps: obtaining characteristic parameters of a rice sample, using a pre-trained lightweight clustering model to determine the type of the rice variety, and outputting a corresponding rice variety label; based on the identified rice variety label, combining the real-time feedback of the change trend of the initial-stage rice outlet pressure sensor and the rice outlet quantity to construct an initial linear response relationship curve of the rice outlet opening degree-rice outlet speed; performing local fitting processing on the initial linear response relationship curve to form an experiential response curve corresponding to the specific rice variety.

3. The method according to claim 2, wherein, The method for setting the first feeding speed according to the first rice outlet opening degree, aiming at meeting the preset reference impedance, comprises the following steps: The reference impedance is the impedance under standard grain measurement, and the impedance is the ratio of the motor load change and the discharge pressure feedback change; obtaining a reference feeding speed under standard grain measurement, meeting the condition that the product of the reference feeding speed and the reference impedance is the same as the product of the first rice outlet opening degree and the first feeding speed, and obtaining the first feeding speed.

4. The method according to claim 3, wherein, The method for performing Latin hypercube sampling on the first rice outlet opening degree and the first feeding speed based on a low-dimensional control space, and randomly combining to obtain a plurality of first control sets, comprises the following steps: pushing the first rice outlet opening degree and the first feeding speed respectively along the lifting direction and the weakening direction for several times at a preset step length to construct a variable change sequence, which comprises the following steps: taking the first rice outlet opening degree as a starting point, sequentially lifting for several times at a preset opening degree step length to form an opening degree lifting sequence, and simultaneously sequentially reducing for several times based on the starting point to form an opening degree weakening sequence; similarly, taking the first feeding speed as a starting point, sequentially pushing for several times at a preset feeding step length to form a feeding lifting sequence, and simultaneously sequentially pushing for several times to form a feeding weakening sequence; combining the four sequences to form a rice outlet opening degree variable set and a feeding speed variable set.

5. The pressure self-adaptive control method of a rice huller based on artificial intelligence according to claim 4, characterized in that, The first pressure is a demand pressure, i.e. a pressure required to be maintained by the system; According to a pre-acquired mapping relationship between the control set and the pressure, specifically: Under multiple control combinations, the pressure value when the system is running is collected to construct a training sample set; Each sample contains a pressure response value corresponding to a control set; The dimension of the control set is set, the order of the polynomial regression is selected, and a polynomial model is generated; A feature matrix and a target vector are constructed, the feature matrix is a matrix expressed by the control set, the target vector is a corresponding pressure vector, the feature matrix and the target vector are mapped to the polynomial model; The least square method is used for training to solve the weight of the optimal polynomial model, and the mapping relationship between the control set and the pressure is obtained.

6. The method according to claim 5, wherein, The feature extraction includes broken rice rate feature, system operation fluency feature and rice milling stability feature; The specific acquisition method of the system operation fluency feature is as follows: After each group of control sets runs for a preset first time, the coefficient of variation of speed error per unit time is obtained to represent the discharge stability; The mean square error of the response time of adjacent control instructions is obtained to represent the response consistency; The two are combined into a two-dimensional stability-consistency feature point, and the Euclidean distance from the feature point to the origin is taken as the system operation fluency feature value.

7. The method according to claim 6, wherein, A plurality of second control sets under the first pressure and corresponding first rice milling evaluation values are mapped to a preset control set-rice milling evaluation display model, specifically: Each control set is bound with its corresponding first rice milling evaluation value, so that each point has a matching relationship between control parameter combination and rice milling effect; The control set-rice milling evaluation display model is an empty distribution space with relative position relationship; A plurality of second control sets under the first pressure and corresponding first rice milling evaluation values are mapped to a preset control set-rice milling evaluation display model, and through local trend modeling, discrete points are converted into a visual continuous region; The preset rice milling mode includes a label of control target preference and a first range of preset first rice milling evaluation value, preference opening degree and preferred feeding speed; Under the condition that the first range is met in the continuous region, the control set corresponding to the peak of the region with the largest preference is selected as the third control set and applied to rice milling.

8. A system using the pressure self-adaptive control method of a rice mill based on artificial intelligence according to any one of claims 1-7, characterized in that, It includes an initialization module, a control set generation module, a pressure mapping and screening module, a control evaluation module, and a control application module; The initialization module is used to construct a rice seed experience curve to set the first rice outlet opening degree; According to the first rice outlet opening degree, the first feeding speed is set to meet the preset reference impedance; The control set generation module is used for Latin hypercube sampling based on low-dimensional control space for the first rice outlet opening degree and the first feeding speed, and random combination to obtain a plurality of first control sets; The pressure mapping and screening module is used to acquire a preset first pressure, and screen a plurality of first control sets according to the pre-acquired mapping relationship between the control set and the pressure to obtain a plurality of second control sets meeting the first pressure; The control evaluation module is used to control the rice mill according to the second control set, and to extract features from the rice mill milling process and generate a first rice milling evaluation value; The control application module is used for mapping a plurality of second control sets under a first pressure and corresponding first rice milling evaluation values to a preset control set-rice milling evaluation display model, and combining a preset rice milling mode to perform data analysis, generate a third control set, and apply to rice milling.

9. An electronic device, comprising: The electronic device comprises: at least one processor; and a memory connected to the at least one processor in communication; wherein The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor to enable the at least one processor to execute the artificial intelligence-based pressure adaptive control method for a rice mill as claimed in any one of claims 1 to 7.

10. A computer readable storage medium storing a computer program, characterized in that, The computer program is executed by the processor to implement the artificial intelligence-based pressure adaptive control method for a rice mill as claimed in any one of claims 1 to 7.

Images