Material drying method, device and medium based on scene recognition and parameter self-calibration

Abstract

The application discloses a material drying method and device based on scene recognition and parameter self-calibration and a medium, and relates to the technical field of drying materials. The method comprises the following steps: acquiring material information; searching for corresponding reference process parameters and historical calibration trajectories; taking the reference process parameters as initial parameters, stretching the historical calibration trajectories according to the target moisture content of the current material on the time axis, taking the initial calibration trajectory curve as a reference path of the current drying process, and controlling the drying equipment to start; collecting the temperature and humidity in the drying cavity according to a preset sampling frequency, and comparing the temperature and humidity with reference temperature and humidity at corresponding time points on the initial calibration trajectory curve; inputting a temperature and humidity deviation vector into a deviation regressor to obtain an offset correction amount of the initial calibration trajectory curve from the current time to a subsequent preset time length section. The application shortens the parameter debugging cycle of different material drying processes and improves the drying product qualification rate.

Description

Technical Field

[0001] This application relates to the field of drying material technology, and in particular to a material drying method, equipment and medium based on scene recognition and parameter self-calibration. Background Technology

[0002] Existing material drying control methods typically rely on operators' experience to manually set process parameters or use fixed drying curves for control. In recent years, some adaptive adjustment methods for drying parameters based on predictive models or rule-based reasoning have emerged. For example, drying parameters are predicted using a pre-trained neural network model, and then parameters such as the temperature, flow rate, and residence time of the drying medium are adjusted sequentially according to preset priority rules based on real-time data such as material moisture content and temperature.

[0003] However, these methods have the following drawbacks: the predictive model requires a large amount of labeled data of different types of materials for offline training, and its applicability to new material scenarios lacking historical data is poor, resulting in limited model generalization ability; rule-based adjustments can only respond to deviations at the current moment and cannot predict future trends in the drying process, leading to lag in parameter adjustments and a tendency for over-adjustment or oscillation; the experience accumulated in each drying process exists in the form of discrete parameter adjustment records and has not been transformed into structured knowledge that can be directly reused in subsequent drying processes, resulting in the need for re-adjustment for the same or similar materials each time, which is inefficient.

[0004] Based on the above analysis, the problems and shortcomings of the existing technology are as follows: Existing material drying control methods suffer from unstable drying efficiency and quality due to the difficulty in reusing experiential knowledge and the lag in parameter adjustment. Summary of the Invention

[0005] This application provides a material drying method, equipment, and medium based on scene recognition and parameter self-calibration, which can solve the problems of unstable drying efficiency and quality caused by the difficulty in reusing experience knowledge and the lag in parameter adjustment in existing material drying control methods.

[0006] In a first aspect, embodiments of this application provide a material drying method based on scene recognition and parameter self-calibration. The method includes: acquiring material information, including material type, initial moisture content, and target moisture content; retrieving corresponding benchmark process parameters and historical calibration trajectories from a pre-stored knowledge base based on the material information; using the benchmark process parameters as initial parameters, stretching the historical calibration trajectory according to the target moisture content of the current material along the time axis to generate an initial calibration trajectory curve, and using the initial calibration trajectory curve as the reference path for the current drying process to control the start of the drying equipment; when the drying equipment starts, collecting the temperature and humidity inside the drying chamber at a preset sampling frequency, and comparing it with the reference temperature and humidity at the corresponding time point on the initial calibration trajectory curve to obtain a temperature and humidity deviation vector; inputting the temperature and humidity deviation vector into a deviation regressor to obtain the offset correction amount for the initial calibration trajectory curve from the current time to the subsequent preset time period segment; and transforming the remaining segment of the initial calibration trajectory curve from the current time based on the offset correction amount to obtain the corrected calibration trajectory curve.

[0007] In one implementation of this application, the reference process parameters are used as initial parameters, and the historical calibration trajectory is stretched along the time axis according to the target moisture content of the current material to generate an initial calibration trajectory curve. Specifically, this includes: obtaining the historical material corresponding to the historical calibration trajectory, extracting the initial moisture content and target moisture content of the historical material, and calculating the historical total dehydration rate; calculating the time axis scaling factor based on the ratio of the current material's total dehydration rate to the historical total dehydration rate; and linearly scaling the time axis of the historical calibration trajectory according to the time axis scaling factor to generate the initial calibration trajectory curve.

[0008] In one implementation of this application, the temperature and humidity deviation vector is input into a deviation regressor to obtain the offset correction amount of the initial calibration trajectory curve from the current time to the subsequent preset time period. Specifically, this includes: taking the difference between the temperature and humidity at the current time and the reference temperature and humidity at the corresponding point on the initial calibration trajectory curve as the single-time temperature and humidity deviation; forming a deviation sequence from the single-time temperature and humidity deviations of a consecutive preset number of sampling times, inputting it into the deviation regressor, and performing time-series fitting on the deviation sequence to output a continuous curve characterizing the change of the offset correction amount over time; and using the continuous curve as the offset correction amount.

[0009] In one implementation of this application, the remaining segment of the initial calibration trajectory curve from the current moment is transformed according to the offset correction amount to obtain the corrected calibration trajectory curve. Specifically, this includes: obtaining the trajectory curve to be calibrated based on the reference value at each time point of the remaining segment of the initial calibration trajectory curve from the current moment and the function value of the continuous curve at the corresponding time point; determining whether the function value in the trajectory curve to be calibrated exceeds the safe operating boundary of the equipment; if it exceeds the safe operating boundary of the equipment, truncating it according to the boundary value, and using the truncated trajectory curve as the corrected calibration trajectory curve.

[0010] In one implementation of this application, after obtaining the corrected calibration trajectory curve, the method further includes: drying according to the calibration trajectory curve, measuring the actual material moisture content after drying, and calculating the deviation from the target moisture content input by the user; when the deviation is less than a preset accuracy threshold, marking the actual parameter sequence of this drying process as a successful drying trajectory and storing it in the knowledge base; when the deviation is greater than or equal to the preset accuracy threshold, marking it as a failed drying trajectory, storing it in the knowledge base, and assigning it a low matching weight.

[0011] In one implementation of this application, the method further includes: for the drying failure trajectory, backtracking by a preset time at the time of failure, and extracting the time segment before failure as a partial valid trajectory; associating and storing the partial valid trajectory with the measured temperature and humidity deviation vector at the time of failure, as a negative sample under abnormal working conditions, and using it to reduce the similarity weight during matching.

[0012] In one implementation of this application, the method further includes: when the number of successful drying trajectories of the same material type in the knowledge base, whose initial moisture content and target moisture content are both within a preset range, exceeds a preset storage limit, the successful drying trajectories are clustered to obtain multiple trajectory clusters; the mean curve and standard deviation curve of the successful drying trajectories within the trajectory clusters at each time point are calculated.

[0013] In one implementation of this application, based on material information, the corresponding baseline process parameters and historical calibration trajectories are retrieved from a pre-stored knowledge base. Specifically, this includes: extracting the drying rate curve from the historical calibration trajectory, the drying rate curve being composed of the derivative of moisture content with time; generating a theoretical drying rate curve based on the initial and target moisture contents of the current material; calculating the time warp distance between the drying rate curve of the historical calibration trajectory and the theoretical drying rate curve; and weighting and summing the reciprocal of the time warp distance and the discrete similarity based on material type to obtain a similarity value.

[0014] Secondly, embodiments of this application also provide a material drying device based on scene recognition and parameter self-calibration. The device includes at least one processor and a memory communicatively connected to the at least one processor. The memory stores instructions executable by the at least one processor, which are executed by the at least one processor to enable the at least one processor to perform any step of the material drying method based on scene recognition and parameter self-calibration.

[0015] Thirdly, embodiments of this application also provide a non-volatile computer storage medium for material drying based on scene recognition and parameter self-calibration, which stores computer-executable instructions, configured to execute any step of a material drying method based on scene recognition and parameter self-calibration.

[0016] The material drying method, equipment, and medium based on scene recognition and parameter self-calibration provided in this application achieve rapid transfer of drying experience between different materials by pre-storing historical calibration trajectories and stretching the trajectory time axis according to the current material's dehydration requirements. This eliminates the need to retrain the model or manually adjust it for new materials. By collecting real-time temperature and humidity deviations and inputting them into a deviation regressor to generate a continuous curve-like offset correction, the reference path for the remaining segments is transformed as a whole rather than corrected point by point. This makes parameter adjustment forward-looking and avoids the lag and oscillation of traditional point-by-point feedback. By managing successful and failed drying trajectories separately and introducing clustering compression and drying rate curve similarity matching, the knowledge base can adaptively optimize, and the matching accuracy and calibration efficiency for new materials continuously improve after long-term operation. In summary, this invention significantly shortens the parameter debugging cycle of drying processes for different materials, improves the qualified rate of dried products, and reduces the system's dependence on operator experience. Attached Figure Description

[0017] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 A flowchart illustrating the material drying method based on scene recognition and parameter self-calibration provided in this application embodiment; Figure 2 This is a schematic diagram of the internal structure of a material drying device based on scene recognition and parameter self-calibration provided in an embodiment of this application. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0019] This application provides a material drying method, equipment, and medium based on scene recognition and parameter self-calibration, which solves the problems of unstable drying efficiency and quality caused by the difficulty in reusing empirical knowledge and the lag in parameter adjustment in the material drying control methods of the prior art.

[0020] The technical solutions proposed in the embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0021] Figure 1 This is a flowchart illustrating a material drying method based on scene recognition and parameter self-calibration provided in an embodiment of this application. Figure 1 As shown in the embodiments of this application, the material drying method based on scene recognition and parameter self-calibration specifically includes the following steps: Step 10: Obtain material information, including material type, initial moisture content, and target moisture content.

[0022] In this step, the type of material is selected by the user from a preset list of material types, including agricultural products, industrial solid waste, textiles, etc. The initial moisture content is the percentage value of the moisture content of the material before drying, and the target moisture content is the percentage value of the moisture content to be achieved after drying. Both are input by the user through the human-computer interaction interface.

[0023] Step 20: Based on the material information, retrieve the corresponding baseline process parameters and historical calibration trajectory from the pre-stored knowledge base.

[0024] In this step, the baseline process parameters are the setpoint sequence of target temperature and target humidity at each time stage during the drying process, which are discretized and stored at fixed time intervals; the historical calibration trajectory is the actual process parameter sequence executed from the start time to the end time during the historical drying process, with each time point recording a tuple containing the temperature setpoint and humidity setpoint.

[0025] As an optional embodiment, based on the material information, the corresponding benchmark process parameters and historical calibration trajectory are retrieved from the pre-stored knowledge base, specifically including: Step 201: Extract the drying rate curve from the historical calibration trajectory, the drying rate curve is composed of the derivative of moisture content with time.

[0026] In this step, the moisture content change record corresponding to the historical calibration trajectory is first read from the knowledge base, and the measured moisture content at each moment during the drying process is stored with the same time axis sampling points. Then, the change of moisture content with time is numerically differentiated, and the derivative value of moisture content at each sampling point is calculated. After smoothing and filtering, the drying rate curve is formed. The drying rate curve is composed of the derivative of moisture content with time.

[0027] Step 202: Generate a theoretical drying rate curve based on the initial moisture content and target moisture content of the current material.

[0028] In this step, based on the preset drying rate model, which assumes that the drying rate remains constant during the constant-rate drying stage and that the drying rate is linearly related to the moisture content during the falling-rate drying stage, the critical moisture content of the material is read from the database according to the current material type, and the duration of the constant-rate stage and the falling-rate stage are calculated by combining the initial moisture content and the target moisture content, and the theoretical drying rate curve is constructed in segments.

[0029] Step 203: Calculate the time warping distance between the drying rate curve of the historical calibration trajectory and the theoretical drying rate curve.

[0030] In this step, the time warping distance between the drying rate curve of the historical calibration trajectory and the theoretical drying rate curve is calculated. The time warping distance is calculated using a dynamic time warping algorithm: Let the historical drying rate curve be a sequence Q=[q1,q2,…,qm] and the theoretical drying rate curve be a sequence T=[t1,t2,…,tn]. Construct an m×n distance matrix, where each element d(i,j) is the absolute value of the difference between qi and tj. Find the path with the minimum cumulative distance from (1,1) to (m,n) through dynamic programming. The cumulative distance on the path is the time warping distance.

[0031] Step 204: Take the reciprocal of the time-normalized distance and the discrete similarity based on material type, respectively, and sum them by weight to obtain the similarity value.

[0032] In this step, the reciprocal of the time-warped distance and the discrete similarity based on material type are weighted and summed to obtain the similarity value. The weight coefficients used for the weighted summation are preset values, where the weight coefficient of the reciprocal of the time-warped distance is denoted as α, and the weight coefficient of the discrete similarity based on material type is denoted as β, and α+β=1, with the value of α ranging from 0.3 to 0.7. The discrete similarity based on material type is set to 1 when the historical material is exactly the same as the current material type, and 0 when they are different types.

[0033] The similarity value is used to sort multiple historical calibration trajectories in the knowledge base, and the one with the highest similarity value is selected as the retrieved historical calibration trajectory. At the same time, the benchmark process parameters stored in association with the historical calibration trajectory are also retrieved and extracted.

[0034] Step 30: Using the baseline process parameters as initial parameters, stretch the historical calibration trajectory along the time axis according to the target moisture content of the current material to generate the initial calibration trajectory curve, and use the initial calibration trajectory curve as the reference path for the current drying process to control the start of the drying equipment.

[0035] In this step, the baseline process parameters are a two-dimensional array containing the setpoints for the drying medium temperature and humidity. The array length corresponds to the number of points after the total drying time is discretized at fixed time intervals. The historical calibration trajectory is a sequence of process parameters actually executed during historical drying processes, with the time axis length corresponding to the total drying time of that historical process. Since the target moisture content of the current material may differ from that of the historical material, the time axis of the historical calibration trajectory needs to be stretched or compressed to align the endpoint with the expected completion time of the current drying process, thereby generating an initial reference path suitable for the current material.

[0036] As an optional embodiment, the baseline process parameters are used as initial parameters, and the historical calibration trajectory is stretched along the time axis according to the target moisture content of the current material to generate an initial calibration trajectory curve. Specifically, it may include: Step 301: Obtain the historical material corresponding to the historical calibration trajectory, extract the initial moisture content and target moisture content of the historical material, and calculate the historical total dehydration rate.

[0037] In this step, the initial moisture content M hist,init and target moisture content M hist,target The historical total dehydration rate is calculated using the following formula: .

[0038] Step 302: Calculate the time axis scaling factor based on the ratio of the current total dehydration rate of the material to the historical total dehydration rate.

[0039] In this step, based on the obtained initial moisture content M of the current material... curr,init and target moisture content M curr,target Calculate the total dehydration rate of the current material: Then calculate the time axis scaling factor k: The time axis scaling factor k represents the ratio of the current material's dehydration requirement to the historical material's dehydration requirement. When k > 1, it means that the current material needs more dehydration time, and the historical calibration trajectory needs to be stretched on the time axis. When k < 1, it means that the current material needs less dehydration, and the time axis needs to be compressed. When k = 1, the original time axis remains unchanged.

[0040] To ensure that the scaled trajectory is physically feasible, a lower and upper limit can be set for the value of k, for example, k∈[0.5,2.0]. If the calculated k exceeds this range, the boundary value will be used instead, and a prompt message will be sent to the user, suggesting that the input of the material moisture content be checked.

[0041] Step 303: Linearly scale the time axis of the historical calibration trajectory according to the time axis scaling factor to generate the initial calibration trajectory curve.

[0042] In this step, the time axis of the retrieved historical calibration trajectory is linearly scaled according to the time axis scaling factor k: Let the original length of the time axis of the historical calibration trajectory be T. hist The process parameter sequence is P hist (t), where t∈[0,T] hist The new timeline length after scaling is T. curr =k×T hist For any time τ∈[0,T] on the new time axis curr] The corresponding process parameter values ​​are determined by the following formula: ; The scaled parameter values ​​are obtained by linear interpolating the original trajectory. The starting point of the new trajectory corresponds to the drying start time τ=0, and the ending point corresponds to the drying completion time τ=T predicted based on the current total dehydration rate of the material. curr This generates the initial calibration trajectory curve.

[0043] In practice, if the sampling time interval of the historical calibration trajectory is Δt hist The scaled target sampling time interval is Δt curr =k×Δt hist To match the control cycle of the drying equipment, the scaled trajectory can be resampled, and the process parameter values ​​can be interpolated to the fixed control cycle time points of the equipment to form a discrete initial calibration trajectory curve.

[0044] Step 40: When the drying equipment is started, the temperature and humidity inside the drying chamber are collected according to the preset sampling frequency, and compared with the reference temperature and humidity at the corresponding time point on the initial calibration trajectory curve to obtain the temperature and humidity deviation vector.

[0045] In this step, the specific values ​​are determined based on the response speed and control accuracy requirements of the drying equipment; the collected temperature and humidity include the measured temperature and relative humidity values ​​inside the drying chamber; let the current sampling time be t. i Timing begins from the start of drying, and the corresponding time point t on the initial calibration trajectory curve. i The reference temperature is T ref (ti), with reference humidity as H ref (ti); the measured temperature is T. meas(ti), the measured humidity is H meas (ti), then the temperature and humidity deviation vector Δ(ti) is defined as: It characterizes the degree of deviation of the current drying process from the reference path. When the measured temperature is higher than the reference temperature, the first component of the deviation vector is positive, indicating that the heating power needs to be reduced or the dehumidification needs to be increased; otherwise, it is negative, indicating that the heating power needs to be increased; the same applies to humidity deviation.

[0046] In practice, since temperature and humidity sensors may have noise, a moving average filter can be applied to the deviation values ​​at multiple consecutive sampling times, and the filtered deviation vector can be used as the input for subsequent steps.

[0047] Step 50: Input the temperature and humidity deviation vector into the deviation regressor to obtain the offset correction amount of the initial calibration trajectory curve from the current time to the subsequent preset time period.

[0048] As an optional embodiment, the temperature and humidity deviation vector is input into the deviation regressor to obtain the offset correction amount of the initial calibration trajectory curve from the current time to the subsequent preset time period. Specifically, it may include: Step 501: The difference between the temperature and humidity at the current time and the reference temperature and humidity at the corresponding point on the initial calibration trajectory curve is taken as the temperature and humidity deviation at a single time.

[0049] In this step, the current sampling time t i The difference between the measured temperature and humidity and the reference temperature and humidity at the corresponding point on the initial calibration trajectory curve is taken as the temperature and humidity deviation at a single moment. It reflects the instantaneous deviation at the current moment. Due to sensor noise and equipment inertia, using it alone can easily lead to parameter jitter. Therefore, it is necessary to combine the deviations of multiple historical moments for comprehensive judgment.

[0050] Step 502: The temperature and humidity deviations at a single moment of a continuously preset number of sampling times are combined into a deviation sequence, which is then input into the deviation regressor. The deviation regressor performs time-series fitting on the deviation sequence and outputs a continuous curve representing the change of the offset correction amount over time.

[0051] In this step, the temperature and humidity deviations at a single moment from a predetermined number of sampling times (N) are combined into a deviation sequence, where N ranges from 5 to 20, preferably 10. Let the current moment be ti, then the deviation sequence is: Input the deviation sequence S(ti) into the deviation regressor, perform time-series fitting on the deviation sequence, and output a continuous curve representing the change of the offset correction amount with time, defined on the time interval from the current time ti to ti+L, denoted as δ(τ), where τ∈[0,L] is the time offset relative to the current time.

[0052] Specifically, the bias regressor incorporates a fitting function that assumes the offset correction varies with time in a second-order polynomial within the prediction window L: Where a, b, and c are the coefficients to be fitted. The goal of the fitting is to make the second-order polynomial curve best approximate the trend extracted from the historical deviation sequence. In the actual fitting process, the sampling points of the deviation sequence S(ti) are first weighted or smoothed to extract a feature vector representing the recent deviation change trend. Then, the coefficients a, b, and c are solved by the least squares method. Specifically, an overdetermined system of equations is constructed, using the deviation values ​​at several recent moments in the historical deviation sequence as observations and time as the independent variable, to solve for the polynomial coefficients that minimize the sum of squared errors.

[0053] To simplify the calculation, a=0 can be fixed, and only linear fitting can be used. In this case, the offset correction is a linear function δ(τ)=b. τ+c is suitable for operating conditions where the deviation changes relatively slowly; for scenarios where the deviation changes rapidly, the second-order polynomial can better capture acceleration information and achieve advance correction.

[0054] Step 503: Use the continuous curve as the offset correction amount.

[0055] In this step, the first-order or second-order polynomial curve δ(τ) is used as the offset correction amount, and a correction value is given at each time point τ within the prediction window. For the two dimensions of temperature and humidity, two independent offset correction amount curves can be established separately, or the temperature and humidity deviation vectors can be merged into a scalar deviation and fitted, and then proportionally allocated to the temperature and humidity corrections. In actual control, this correction amount will be superimposed on the remaining segment of the initial calibration trajectory curve to form the corrected trajectory.

[0056] To avoid excessive corrections that could cause equipment to exceed limits, threshold constraints can be set for the polynomial coefficients a, b, and c. For example, the absolute value of the temperature correction should not exceed 2°C every 10 seconds, and the humidity correction should not exceed 5%RH every 10 seconds. If the calculated correction exceeds this range, it should be limited to the boundary values.

[0057] Step 60: Based on the offset correction amount, transform the remaining segment of the initial calibration trajectory curve from the current moment to obtain the corrected calibration trajectory curve.

[0058] In this step, the remaining segment of the initial calibration trajectory curve starting from the current time ti is transformed according to the offset correction amount δ(τ) to obtain the corrected calibration trajectory curve. The remaining segment refers to the time interval from the current sampling time ti to the end of the curve of the initial calibration trajectory curve. The total length of the segment gradually shortens as the drying process progresses.

[0059] The specific transformation method is as follows: the offset correction curve δ(τ) is superimposed onto each time point of the remaining segment of the initial calibration trajectory curve, that is, for any time point t∈[ti,T] in the remaining segment... curr Let τ = t t i The corrected reference process parameter value (relative to the current time offset) is: Wherein, Pinit(t) Pinit ( t The initial calibration trajectory curve at time point t t The original process parameter values ​​(including temperature and humidity setpoints; for two-dimensional parameters, these can be corrected independently or proportionally using a uniform correction amount), Pcorr(t). Pcorr ( t ) represents the corrected process parameter value.

[0060] Since the offset correction curve δ(τ) is only defined within the prediction window [0,L], for the portion of the remaining segment that exceeds the prediction window, any of the following methods can be used: keep the correction amount δ(L) at the end of the prediction window constant and extrapolate it to the entire remaining segment; only correct the segment within the prediction window, and keep the initial trajectory unchanged for the portion that exceeds the prediction window; use linear decay to gradually reduce the correction amount back to zero in order to avoid abrupt changes in the trajectory.

[0061] Preferably, method one is adopted, that is, the correction amount at the end of the prediction window remains unchanged until the drying is finished, so as to maintain the continuity of parameter adjustment.

[0062] As an optional embodiment, the remaining segment of the initial calibration trajectory curve from the current moment is transformed according to the offset correction amount to obtain the corrected calibration trajectory curve. Specifically, it includes: Step 601: Based on the reference value at each time point of the remaining segment of the initial calibration trajectory curve from the current moment and the function value of the continuous curve at the corresponding time point, the trajectory curve to be calibrated is obtained.

[0063] In this step, the reference process parameter values ​​at each time point within the remaining segment of the initial calibration trajectory curve from the current moment, and the function value of the continuous curve δ(τ) at the corresponding time point, are added together to obtain the trajectory curve to be calibrated. That is, for any time point t within the remaining segment, the following calculation is performed. P obtained from this temp (t) is the trajectory curve to be calibrated without boundary constraints. It deviates from the initial trajectory within the prediction window. The shape of the deviation is determined by the polynomial form of the offset correction.

[0064] Step 602: Determine whether the function value in the trajectory curve to be calibrated exceeds the safe operating boundary of the equipment.

[0065] In this step, the boundary values ​​can be provided by the equipment manual or preset by the user according to the material's tolerance; the setting of the boundary values ​​should ensure the safe operation of the equipment and prevent damage to the material.

[0066] Step 603: If the equipment exceeds the safe operating boundary, truncate it according to the boundary value, and use the truncated trajectory curve as the corrected calibration trajectory curve.

[0067] In this step, the truncation operation is performed independently for each discrete time point that exceeds the boundary. To ensure the smoothness of the truncated trajectory, local smoothing processing can be applied to the parameter values ​​near the truncation point, such as using moving average or first-order low-pass filtering. The truncated trajectory curve is used as the corrected calibration trajectory curve, denoted as P. final (t), used for subsequent drying process control.

[0068] As an improvement, if the cutoff is large and the boundary is reached at multiple consecutive time points, a warning message can be generated to remind the user that the current drying process may be facing equipment limits or material abnormalities, and manual intervention or adjustment of the baseline process parameters is recommended.

[0069] As an optional embodiment, after obtaining the corrected calibration trajectory curve, the method may further include: Step 701: Drying according to the calibration trajectory curve, measuring the actual material moisture content after drying, and calculating the deviation from the target moisture content input by the user; Step 702: When the deviation is less than a preset accuracy threshold, marking the actual parameter sequence of this drying process as a successful drying trajectory and storing it in the knowledge base; Step 703: When the deviation is greater than or equal to the preset accuracy threshold, marking it as a failed drying trajectory, storing it in the knowledge base, and assigning it a low matching weight.

[0070] Specifically, in this step, the default weight for the failed trajectory can be set to 0.3. wfail =0.3 to 0.5, with a weight of 1.0 for successful trajectories. During subsequent retrieval and matching, the calculated similarity value needs to be multiplied by the weight of this trajectory to reduce the probability of failed trajectories being selected. Furthermore, the storage record of failed trajectories should also include a failure reason code, such as high final moisture content, low final moisture content, or interruption in the drying process, to facilitate post-event analysis by the user.

[0071] As an optional embodiment, the method may further include: step 704: for the drying failure trajectory, backtrack by a preset time at the time of failure, and extract the time segment before failure as a part of the valid trajectory.

[0072] In this step, for the drying failure trajectory, although the overall standard is not met, it may still follow a reasonable drying pattern for a certain period of time before the failure occurs. In order to avoid completely discarding useful information, a preset time is reversed at the moment of failure, and that time segment is extracted as a part of the valid trajectory. The moment of failure refers to the moment when irreversible anomalies first occur during the drying process, such as the sudden slowdown in the rate of decrease in moisture content or the continuous exceedance of temperature, or the moment when failure is determined at the end of drying.

[0073] Step 705: Associate and store some valid trajectories with the measured temperature and humidity deviation vector when the trigger fails, as negative samples under abnormal working conditions, to reduce the similarity weight during matching.

[0074] In this step, some valid trajectories and their associated failure time deviation vectors are stored as negative samples in a separate area of ​​the knowledge base, or mixed with normal trajectories but with an additional negative sample flag. When performing historical calibration trajectory retrieval and matching in the future, for the matched candidate trajectory, if it contains negative sample information, the matching weight of the candidate trajectory is further reduced based on the similarity between the temperature and humidity deviation vector at the current moment and the failure time deviation vector stored in the negative samples. For example, it is multiplied by a penalty factor, so that it moves backward in the ranking to avoid repeating the same failure path.

[0075] Through the aforementioned negative sample management mechanism, the knowledge base can learn from failures, gradually reduce its reliance on undesirable trajectories, and improve the robustness and self-healing ability of the drying control system.

[0076] As an optional embodiment, the method may further include: when the number of successful drying trajectories of the same material type in the knowledge base, whose initial moisture content and target moisture content are both within a preset range, exceeds a preset storage limit, clustering the successful drying trajectories to obtain multiple trajectory clusters; and calculating the mean curve and standard deviation curve of the successful drying trajectories within the trajectory clusters at each time point.

[0077] In this step, when the number of successful drying trajectories in the knowledge base for the same material type, with both the initial moisture content and the target moisture content falling within their respective preset ranges, exceeds the preset storage limit, the clustering and compression process is triggered.

[0078] Specifically, the preset storage limit can be set according to the system storage capacity and retrieval efficiency. For example, a maximum of 100 successful trajectories can be retained under the same material type. The preset range is used to define similar working conditions: the range width of the initial moisture content can be set to ±5%, and the range width of the target moisture content can be set to ±3%. Only when the initial and target moisture contents of the historical trajectory and the current material both fall within this range are they considered as candidate objects in the same cluster pool.

[0079] First, the successful trajectories are clustered. All successful drying trajectories in the candidate pool are used as samples, and a clustering algorithm based on curve shape similarity is employed to divide them into several trajectory clusters. Then, the mean curve and standard deviation curve are calculated. For each trajectory cluster, the temperature and humidity setpoints of all successful trajectories within the cluster at the same time axis sampling points are statistically aggregated. The time axes of each trajectory are first normalized to the same length using linear interpolation, and then the mean curve at each normalized time point is calculated. Where n is the number of trajectories within the cluster, P j (t) represents the process parameter value of the j-th trajectory at time point t; standard deviation curve: Finally, all successful individual trajectories within the cluster are deleted, leaving only one representative trajectory, the mean curve μ(t). Simultaneously, the standard deviation curve σ(t) is associated with and stored with this representative trajectory for subsequent matching to evaluate its reliability.

[0080] The above are embodiments of the method proposed in this application. Based on the same inventive concept, embodiments of this application also provide a material drying device based on scene recognition and parameter self-calibration, the structure of which is as follows: Figure 2 As shown.

[0081] Figure 2 This is a schematic diagram of the internal structure of a material drying device based on scene recognition and parameter self-calibration, provided as an embodiment of this application. Figure 2 As shown, the device includes: At least one processor 201; And a memory 202 that is communicatively connected to at least one processor; The memory 202 stores instructions that can be executed by at least one processor. The instructions are executed by at least one processor 201 to enable at least one processor 201 to: perform any step of a material drying method based on scene recognition and parameter self-calibration.

[0082] Some embodiments of this application provide corresponding to Figure 1 A non-volatile computer storage medium for material drying based on scene recognition and parameter self-calibration stores computer-executable instructions, which are configured to execute any step of the material drying method based on scene recognition and parameter self-calibration.

[0083] The various embodiments in this application are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the embodiments for IoT devices and media are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0084] The systems, media, and methods provided in this application are one-to-one correspondences. Therefore, the systems and media also have similar beneficial technical effects as their corresponding methods. Since the beneficial technical effects of the methods have been described in detail above, the beneficial technical effects of the systems and media will not be repeated here.

[0085] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0086] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0087] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0088] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0089] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0090] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0091] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0092] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0093] The above description is merely an embodiment of this application and is not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A material drying method based on scene recognition and parameter self-calibration, characterized in that, The method includes: Obtain material information, including material type, initial moisture content, and target moisture content; Based on the material information, retrieve the corresponding benchmark process parameters and historical calibration trajectories from the pre-stored knowledge base; Using the benchmark process parameters as initial parameters, the historical calibration trajectory is stretched along the time axis according to the target moisture content of the current material to generate an initial calibration trajectory curve. The initial calibration trajectory curve is then used as the reference path for the current drying process to control the start of the drying equipment. When the drying equipment is started, the temperature and humidity inside the drying chamber are collected according to a preset sampling frequency and compared with the reference temperature and humidity at the corresponding time point on the initial calibration trajectory curve to obtain the temperature and humidity deviation vector. The temperature and humidity deviation vector is input into the deviation regressor to obtain the offset correction amount of the initial calibration trajectory curve from the current time to the subsequent preset time period. Based on the offset correction amount, the remaining segment of the initial calibration trajectory curve from the current moment is transformed to obtain the corrected calibration trajectory curve.

2. The material drying method based on scene recognition and parameter self-calibration according to claim 1, characterized in that, The step of using the benchmark process parameters as initial parameters and stretching the historical calibration trajectory along the time axis according to the target moisture content of the current material to generate an initial calibration trajectory curve specifically includes: Obtain the historical materials corresponding to the historical calibration trajectory, extract the initial moisture content and target moisture content of the historical materials, and calculate the historical total dehydration rate; Calculate the time axis scaling factor based on the ratio of the current total dehydration rate of the material to the historical total dehydration rate; The time axis of the historical calibration trajectory is linearly scaled according to the time axis scaling factor to generate the initial calibration trajectory curve.

3. The material drying method based on scene recognition and parameter self-calibration according to claim 1, characterized in that, The step of inputting the temperature and humidity deviation vector into the deviation regressor to obtain the offset correction amount of the initial calibration trajectory curve from the current time to the subsequent preset time period specifically includes: The difference between the current temperature and humidity and the reference temperature and humidity at the corresponding point on the initial calibration trajectory curve is taken as the temperature and humidity deviation at a single moment. The temperature and humidity deviations at a single moment of a continuously preset number of sampling times are combined into a deviation sequence, which is then input into the deviation regressor. The deviation regressor performs time-series fitting on the deviation sequence and outputs a continuous curve representing the change of the offset correction amount over time. The continuous curve is used as the offset correction amount.

4. The material drying method based on scene recognition and parameter self-calibration according to claim 3, characterized in that, The step of transforming the remaining segment of the initial calibration trajectory curve from the current moment according to the offset correction amount to obtain the corrected calibration trajectory curve specifically includes: The trajectory curve to be calibrated is obtained based on the reference values ​​at each time point of the remaining segment from the current time in the initial calibration trajectory curve, and the function values ​​of the continuous curve at the corresponding time points. Determine whether the function value in the trajectory curve to be calibrated exceeds the safe operating boundary of the equipment; If the trajectory exceeds the safe operating boundary of the device, it is truncated according to the boundary value, and the truncated trajectory curve is used as the corrected calibration trajectory curve.

5. The material drying method based on scene recognition and parameter self-calibration according to claim 1, characterized in that, After obtaining the corrected calibration trajectory curve, the method further includes: Drying is performed according to the calibration trajectory curve, and the actual moisture content of the material after drying is measured to calculate the deviation from the target moisture content input by the user. When the deviation is less than the preset accuracy threshold, the actual parameter sequence of this drying process is marked as a successful drying trajectory and stored in the knowledge base; When the deviation is greater than or equal to a preset accuracy threshold, it is marked as a drying failure trajectory, stored in the knowledge base, and assigned a low matching weight.

6. The material drying method based on scene recognition and parameter self-calibration according to claim 5, characterized in that, The method further includes: For the drying failure trajectory, a preset time is rewound at the moment of failure, and the time segment before the failure is extracted as a part of the valid trajectory; The effective trajectories are associated with and stored with the measured temperature and humidity deviation vectors when the trigger fails, serving as negative samples under abnormal operating conditions to reduce similarity weights during matching.

7. The material drying method based on scene recognition and parameter self-calibration according to claim 5, characterized in that, The method further includes: When the number of successful drying trajectories of the same material type in the knowledge base, whose initial moisture content and target moisture content are both within a preset range, exceeds the preset storage limit, the successful drying trajectories are clustered to obtain multiple trajectory clusters. Calculate the mean curve and standard deviation curve of the successful drying trajectory within the trajectory cluster at each time point.

8. The material drying method based on scene recognition and parameter self-calibration according to claim 1, characterized in that, The step of retrieving corresponding baseline process parameters and historical calibration trajectories from a pre-stored knowledge base based on the material information specifically includes: The drying rate curve is extracted from the historical calibration trajectory, and the drying rate curve is composed of the derivative of moisture content with respect to time. Based on the initial moisture content and target moisture content of the current material, a theoretical drying rate curve is generated; Calculate the time warping distance between the drying rate curve of the historical calibration trajectory and the theoretical drying rate curve; The reciprocal of the time-normalized distance and the discrete similarity based on material type are weighted and summed to obtain the similarity value.

9. A material drying device based on scene recognition and parameter self-calibration, characterized in that, The device includes: At least one processor; And, a memory communicatively connected to the at least one processor; The memory stores instructions executable by the at least one processor, which, when executed by the at least one processor, enable the at least one processor to: Perform the steps of the material drying method based on scene recognition and parameter self-calibration as described in any one of claims 1-8.

10. A non-volatile computer storage medium for material drying based on scene recognition and parameter self-calibration, storing computer-executable instructions, characterized in that: The computer-executable instructions are set as follows: Perform the steps of the material drying method based on scene recognition and parameter self-calibration as described in any one of claims 1-8.

Images