A method for accurately predicting the bulk density of a solid powder

Abstract

The present application relates to a method for accurately predicting the bulk density of solid powder, which is a bulk density prediction method based on random noise filtering algorithm, accurate weighing technology and solid powder suction device. On the basis of volume control, the bulk density of the powder is predicted to control the mass of the suction powder, thereby effectively improving the experimental efficiency and success rate. Through actual data testing, the sampling mass error of the present application does not exceed 5%, which can better complete the sampling task and meet the accuracy requirements of solid powder sampling. The present application can real-time fit and adjust parameters based on actual data, so compared with the direct suction method, the present application has great improvement in accuracy. In addition, the present application can handle different work tasks such as single raw material and single sampling target, multiple raw materials and multiple targets, and can adapt to different task scenarios to avoid the burden of manual adjustment process.

Description

Technical Field

[0001] This invention belongs to the field of accurate solid weighing, specifically a method for accurately predicting the bulk density of solid powder based on a random noise filtering algorithm, applicable to automated micro-volume solid powder absorption sampling devices. Background Technology

[0002] In biological, chemical, and other related experiments, various solid raw materials are required as reactants to participate in chemical reactions. For powdered solid reagents, the common sampling method in laboratories is manual sampling using paper troughs, spatulas, or solid powder samplers. In precision chemical experiments where the accuracy of solid powder mass requirements is as high as milligram level, manual operation is time-consuming, labor-intensive, and prone to errors, making it difficult to meet the requirements.

[0003] Most existing solid samplers are stand-alone devices that utilize simple actions such as grinding or sieving, relying on gravity to cause powder to fall and thus collect samples; or they use airflow to convert powder into a dusty state and guide it into a target container to collect a certain amount of solid powder. These technical solutions do not consider the bulk density of the solid powder during sampling, nor can they adjust the sampling strategy based on feedback. They are functionally limited when handling solid powders with various bulk densities, and cannot handle disturbances and interferences caused by objective environmental factors during sampling in real time, potentially leading to significant errors. For continuously operating automated solid sampling systems, real-time acquisition of sampling information, algorithmic fitting of the current drug's bulk density, and feedback adjustment of the sampling strategy will greatly improve the adaptability, reliability, and accuracy of the sampling equipment, effectively increasing the efficiency of chemical experiments and reducing labor burden and experimental costs. Summary of the Invention:

[0004] This invention provides a method for accurately predicting the bulk density of solid powder. This method is based on a random noise filtering algorithm, accurate weighing technology, and a solid powder collection device. By predicting the bulk density of the powder based on volume control, the mass of the powder collected can be controlled, thereby effectively improving experimental efficiency and success rate.

[0005] The specific technical solution of the present invention is as follows:

[0006] 1. The length of the sampling tube inside the suction head of the solid powder suction device is set to l, and the mass of the sample obtained is set to w. The two should satisfy the following linear relationship:

[0007] w = kl + b. (1)

[0008] Where l is the independent variable, w is the dependent variable, k is the bulk density, and b is the linear intercept. In this scenario, the linear intercept represents the actual powder mass obtained when the lumen length l = 0 under the influence of objective factors.

[0009] During the research, the applicant discovered that in actual sampling, due to objective disturbances, the experimental data are affected by random noise, and the specific values ​​of k and b will fluctuate randomly around the ideal values. To achieve more precise sampling control and reduce experimental error, it is necessary to use the actual values ​​of l and w collected during the sampling process to fit the values ​​of k and b at that moment. Furthermore, due to the fluctuations in the k and b parameters, the l used for each sampling also changes accordingly.

[0010] Based on the above research, a discrete state transition equation for the sampling device is established:

[0011] X(t)=X(t-1)+W(t), (2)

[0012] The packing density and the linear intercept [k,b] in equation (1) are used as the state parameters of this device, i.e., X = [k,b]. W is the process noise, and parameter t is the discrete time.

[0013] 2. Establish matrix Z to store the collected and updated state values ​​at each time point, i.e.

[0014] n is a natural number greater than or equal to 0, representing the total number of (k, b) data pairs of bulk density and linear intercept that have been fitted and recorded for this solid powder. If no experiments or records have been made for this solid powder before, then n = 0, i.e., Z is empty.

[0015] Due to random noise, the observed values ​​of the linear relationship parameters fitted by each data acquisition will fluctuate randomly around the true values. Each acquired observation is recorded in the Z matrix, and the random noise is removed using a Kalman filter. The filtered current state [k,b] is then output as the updated linear relationship parameters. Compared to the unfiltered linear parameters, the filtered linear relationship parameters will be closer to the true values.

[0016] 3. When the target mass of the drug powder is w, the sampling strategy is to sample in three steps. After three samplings, the mass of the drug in the target container reaches w.

[0017] 4. The first step of sampling targets w / 3. If the bulk density of this drug has been predicted before, the recorded matrix Z is used, and random disturbances are removed based on the Kalman filter method to calculate (k,b) which is close to the true value; otherwise, the matrix Z is empty, and the default preset (k,b) = (1.5,0) or (1.0,0).

[0018] Through multiple experiments, the applicant discovered that the bulk density of most powdered drugs under normal temperature and pressure conditions is within the range of [0.5, 2] (mg / mm). Therefore, k = 1.5 or 1.0 is chosen as a suitable initial value for the program preset, used for noise filtering and fitting the bulk density function of solid powder. In an ideal state, when the sampling tube length l = 0, there is no space inside the pipette tip to accommodate the drug powder. Therefore, the mass w of the drug obtained should also be 0. The linear relationship between w and l passes through the origin of the coordinate system (w, l) = (0, 0), that is, w and l satisfy a direct proportional relationship, and the linear intercept b = 0. Therefore, (k, b) = (1.5, 0) or (1.0, 0) is selected as the initial value.

[0019] The lumen length l1 in the first sampling step is calculated based on the linear relationship:

[0020]

[0021] After sampling, the actual mass obtained from the first sampling step is measured as w1. The (l1,w1) data pair is added to the data collection record matrix Z, and the updated (k,b) is obtained.

[0022] 5. The sampling target for the second step is 2w / 3-w1. The lumen length l2 in the second sampling step is calculated based on the linear relationship:

[0023]

[0024] After sampling, the actual mass obtained from the second sampling step is measured as w2. The (l2,w2) data pair is added to the data collection record matrix Z, and the updated (k,b) is obtained.

[0025] 6. The sampling target in the third step is w-w1-w2. The lumen length l3 in the third step sampling is calculated based on the linear relationship:

[0026] l3=(w-w1-w2-b) / k (5)

[0027] The (l3,w3) data pair is added to the data collection record matrix Z, and the updated (k,b) is obtained for subsequent sampling of this drug.

[0028] The actual mass obtained from the third sampling step, w3, was obtained by absorption sampling and weighing.

[0029] Before each sampling operation, the lumen length l used for sampling is given. i By measuring the actual mass w i Then, based on the linear relationship given in equation (1), the corresponding (k) can be obtained by fitting using the linear regression method. i b iThis yields the packing density at each sampling point. After several mass measurements and data fitting, a series of packing density values ​​can be obtained. After Kalman filtering, a packing density (k,b) that approximates the true value can be obtained. Similarly, based on the linear relationship of equation (1), the closer the fitted predicted packing density k is to the true value, the closer the correspondence between the sampling tube length l and the actual obtained mass w is to the actual situation. Thus, the actual mass w obtained in the next step... i The closer the sample size is to the target sample size, the smaller the sampling error. Therefore, under this continuously feedback-adjusted sampling strategy, the sum of the three samples, w1+w2+w3, should be very close to the target value w. This demonstrates the optimization of accuracy in the "bulk density prediction" function of this invention for solid sampling tasks.

[0030] 7. For a task involving a single sampling target quality, the workflow ends here; for an integrated task involving multiple sampling target qualities, repeat steps 4, 5, 6, and 7 until all sampling requirements are met.

[0031] Since the mass and bulk density of this solid powder have already been measured and predicted, a series of (k) i b i The value of (k,b) is recorded in matrix Z. The previously obtained (k,b) can be used in the subsequent sampling process. Based on equation (1), the length l of the pipette tube is calculated and adjusted. The steps of measuring quality and predicting fitting are then carried out. The newly calculated (k) i b i This is also added to matrix Z. In this continuous feedback adjustment process, due to the continuous accumulation of (l) i w i This data is used to predict the bulk density of the raw material. As the amount of data increases and the Kalman filter method is used, the predicted bulk density k will get closer and closer to the true value, overcoming the interference of random errors.

[0032] The solid powder aspiration device involved in this invention is a micro-quantitative solid powder sampling device disclosed in Chinese Utility Model Patent CN202222592221.5. This solid powder aspiration device can aspirate powdered medicines with a length of 0 to L mm at a time.

[0033] This device allows adjustment of the pipette plug within the lumen during sampling by moving the pin up and down, thereby changing the length l of the lumen used for sampling. In the technical solution of this invention, the length l of the lumen used for sampling is adjusted before each sampling to achieve more precise sampling control.

[0034] This invention has been tested with actual data, and the sampling quality error does not exceed 5%, which can effectively complete the sampling task and meet the accuracy requirements for solid powder sampling. Furthermore, this invention can fit and adjust parameters in real time based on actual data, thus significantly improving accuracy compared to direct sampling methods. In addition, this invention can handle different task scenarios, including single sampling of a single raw material and multiple sampling of multiple raw materials and multiple targets, and can adaptively adapt to different task scenarios, avoiding the burden of manual process adjustments. Attached Figure Description

[0035] Figure 1 This is a schematic diagram of the suction head of the solid powder suction device disclosed in Chinese Patent CN202222592221.5, which is used in this invention. Figure 1 middle:

[0036] (1) is the entire sampling pipette head, with the wider upper part used to assemble at the end of the gas path and the slender lower part used to pick up solid powder. (101) is the maximum lumen length L that can be achieved in one sampling.

[0037] (2) A movable pipette plug is assembled inside the pipette head cavity. The pipette plug has fine channels inside, which allow airflow to pass through, so as to ensure the smooth realization of the suction and blowing functions. The airflow will not affect the position of the pipette plug, but the solid powder will be confined below the pipette plug. Therefore, the length of the cavity used for sampling depends on the position of the pipette plug in the cavity, which is (102) the length l of the cavity used for sampling.

[0038] (3) is a push pin that can move up and down, and is mounted above the (2) pipette plug in a position that does not affect the operation of the gas path. Before each sampling, the push pin will move according to the required (102) sampling tube length l, pushing the (2) pipette plug to the corresponding position, thereby adjusting the sampling tube length. Detailed Implementation

[0039] Example 1: The sampled raw material is a single type of solid powder, and the target mass value of the sample is a single unit.

[0040] 1. The length of the sampling tube inside the suction head of the solid powder suction device is set to l, and the mass of the sample obtained is set to w. The two should satisfy the following linear relationship:

[0041] w = kl + b. (1)

[0042] Where k is the packing density and b is the intercept of the linear relationship.

[0043] Establish the discrete state transition equations for the sampling device:

[0044] X(t)=X(t-1)+W(t), (2)

[0045] The packing density and the linear intercept [k,b] in equation (1) are used as the state parameters of this device, i.e., X = [k,b]. W is the process noise, and parameter t is the discrete time.

[0046] 2. Establish matrix Z to store the collected and updated state values ​​at each time point, i.e. Each collected observation data is recorded in the Z matrix. After removing random noise through the Kalman filter, the filtered current state [k,b] is output as the updated linear relationship parameter.

[0047] 3. This embodiment involves a single type of solid powder and a single target mass value. The total target mass of the solid powder is w = 25.00 mg. The sampling strategy is to take samples in three steps. After three samplings, the mass of the solid powder in the target container reaches 25.00 mg.

[0048] 4. The initial sample target w / 3 = 8.33 mg. Since the bulk density of this solid powder was not predicted, based on experimental experience, the bulk density parameter k in formula (1) is set to 1.50, and b to 0. According to the relationship:

[0049]

[0050] The calculated value is l1 = 5.55 mm.

[0051] After adjusting the tube length of the sampling section in the solid powder absorption device to 5.55 mm, the sample was sampled using an aspiration method. The actual sample mass w1 = 8.10 mg. The bulk density k = 1.45945946 was obtained by linear regression least squares fitting, and the linear intercept b = 1.79275728 × 10⁻⁶. -16 .

[0052] Add the (l1,w1) data pair to the collected information recording matrix Z, and obtain the updated (k,b) = (1.46, 1.79 × 10^2)^2. -16 ).

[0053] 5. The target for the second sampling is 2w / 3-w1 = 8.57mg. According to the relationship:

[0054]

[0055] The calculated value is l2 = 5.21 mm.

[0056] After adjusting the tube length of the sampling section in the solid powder absorption device to 5.21 mm, the sample was sampled using an aspiration method. The actual sample mass w2 = 8.30 mg. The bulk density k = 1.59309021 was obtained by linear regression least squares fitting, and the linear intercept b = 1.00778818 × 10⁻⁶. -16 .

[0057] Add the (l2,w2) data pair to the collected information recording matrix Z, and obtain the updated (k,b) = (1.55, 1.27 × 10^2)^2. -16 ).

[0058] 6. The target for the third sampling is w-w1-w2 = 8.60 mg. According to the relationship:

[0059] l3=(w-w1-w2-b) / k (5)

[0060] The calculated value is l3 = 6.09 mm.

[0061] After adjusting the length of the sampling section of the solid powder absorption device to 6.09 mm, the sample was sampled using an aspiration method. The actual sample mass w3 = 8.90 mg. The bulk density k = 1.46141215 was obtained by linear regression least squares fitting, and the linear intercept b = 2.33183634 × 10⁻⁶. -16 .

[0062] Three samples yielded a total of 25.30 mg of solid powder, w1 + w2 + w3. Compared with the target w = 25.00 mg, the error was 1.2% and 0.30 mg, respectively.

[0063] 7. This task involves a single target mass of a single type of solid powder. All targets have been achieved, and the workflow is now complete.

[0064] Example 2: Two types of raw materials were sampled, and the target mass of each type of raw material was two.

[0065] 1. The length of the sampling tube inside the suction head of the solid powder suction device is set to l, and the mass of the sample obtained is set to w. The two should satisfy the following linear relationship:

[0066] w = kl + b. (1)

[0067] Where k is the packing density and b is the intercept of the linear relationship.

[0068] Establish the discrete state transition equations for the sampling device:

[0069] X(t)=X(t-1)+W(t), (2)

[0070] The packing density and the linear intercept [k,b] in equation (1) are used as the state parameters of this device, i.e., X = [k,b]. W is the process noise, and parameter t is the discrete time.

[0071] 2. Establish matrix Z to store the collected and updated state values ​​at each time point, i.e. Each collected observation data is recorded in the Z matrix. After removing random noise through the Kalman filter, the filtered current state [k,b] is output as the updated linear relationship parameter.

[0072] 3. The task involves two types of solid powder: powder A and powder B. The first target value for powder A is 15.00 mg, and the second target value is 30.00 mg. The first target value for powder B is 30.00 mg, and the second target value is 20.00 mg. The sampling strategy is to sample in three steps. For each sampling target, three samples are taken respectively, and the total mass of the powder in each of the three steps reaches the corresponding target value.

[0073] 4. For powder A, the first target value w = 15.00 mg, and the first sample target w / 3 = 5.00 mg. Since the bulk density of this powder has not been predicted, based on experimental experience, the bulk density parameter k in formula (1) is set to 1.00, and b is set to 0. According to the relationship:

[0074]

[0075] The calculated value is l1 = 5.00 mm.

[0076] After adjusting the length of the sampling section of the solid powder aspiration device to 5.00 mm, the sample was aspirated. The actual sample mass w1 = 5.80 mg. The bulk density k = 1.16000000 was obtained by linear regression least squares fitting, and the linear intercept b = 1.46471105 × 10⁻⁶. -16 .

[0077] Add the (l1,w1) data pair to the collected information recording matrix Z, and obtain the updated (k,b) = (1.16, 1.46 × 10^2)^2. -16 ).

[0078] 5. The target for the second sampling is 2w / 3 - w1 = 4.2 mg. According to the relationship:

[0079]

[0080] The calculated value is l2 = 3.62 mm.

[0081] After adjusting the tube length of the sampling section in the solid powder absorption device to 3.62 mm, the sample was sampled using an aspiration method. The actual sample mass w2 = 3.60 mg. The bulk density k = 0.994475138 was obtained by linear regression least squares fitting, and the linear intercept b = 3.63211953 × 10⁻⁶. -17 .

[0082] Add the (l2, w2) data pair to the collected information recording matrix Z, and obtain the updated (k, b) = (1.05, 7.30 × 10^2)^2. -17 ).

[0083] 6. The target for the third sampling is w-w1-w2 = 5.60 mg. According to the relationship:

[0084] l3=(w-w1-w2-b) / k (5)

[0085] The calculated value is l3 = 5.34 mm.

[0086] After adjusting the tube length of the sampling section in the solid powder absorption device to 5.34 mm, the sample was sampled using an aspiration method. The actual sample mass w3 = 5.70 mg. The bulk density k = 1.06741573 was obtained by linear regression least squares fitting, and the linear intercept b = 4.07764165 × 10⁻⁶. -16 .

[0087] Three samples yielded a total of 15.10 mg of solid powder, w1 + w2 + w3. Compared with the target w = 15.00 mg, the errors were 0.67% and 0.10 mg, respectively.

[0088] Add the (l3,w3) data pair to the collected information recording matrix Z, and obtain the updated (k,b) = (1.06, 2.82 × 10^3)^2. -16 ), so that solid powder A can be sampled a second time for continued use.

[0089] 7. This task involves two types of solid powder, with two target masses for each type of solid powder. Currently, only the first target sampling of solid powder A has been completed, and not all targets have been completed. Next, repeat steps 4, 5, 6, and 7.

[0090] 4'. For the second target value of powder A, let w = 30.00 mg, and the first sampling target w / 3 = 10.00 mg. This powder has undergone bulk density prediction in the previous process; therefore, based on step 6, the bulk density parameter k in the preset formula (1) is set to 1.06, and b is set to 2.82 × 10⁻⁶. -16 According to the formula (3), l1 = 9.43 mm was calculated.

[0091] After adjusting the tube length of the sampling section in the solid powder absorption device to 9.43 mm, the sample was sampled using an aspiration method. The actual sample mass w1 = 10.20 mg. The bulk density k = 1.08165429 was obtained by linear regression least squares fitting, and the linear intercept b = 2.90577261 × 10⁻⁶. -16 .

[0092] Add the (l1, w1) data pair to the collected information recording matrix Z, and obtain the updated (k, b) = (1.07, 2.87 × 10^2)^2. -16 ).

[0093] 5'. The second sampling target is 2w / 3-w1=9.80mg, and l2=9.16mm is calculated according to the relationship (4).

[0094] After adjusting the tube length of the sampling section in the solid powder absorption device to 9.16 mm, the sample was sampled using an aspiration method. The actual sample mass w2 = 9.78 mg. The bulk density k = 1.06768559 was obtained by linear regression least squares fitting, and the linear intercept b = 7.49195719 × 10⁻⁶. -16 .

[0095] Add the (l2, w2) data pair to the collected information recording matrix Z, and obtain the updated (k, b) = (1.07, 5.73 × 10^2)^2. -16 ).

[0096] 6'. The third sample target w-w1-w2=10.02mg, and l3=9.36mm was calculated according to the relationship (5).

[0097] After adjusting the tube length of the sampling section in the solid powder absorption device to 9.36 mm, the sample was sampled using an aspiration method. The actual sample mass w3 = 9.90 mg. The bulk density k = 1.05769231 was obtained by linear regression least squares fitting, and the linear intercept b = -5.15468071 × 10⁻⁶. -17 .

[0098] The three samplings yielded a total of 29.88 mg of solid powder, w1 + w2 + w3. Compared with the target w = 30.00 mg, the errors were 0.4% and 0.12 mg, respectively.

[0099] 7'. This task involves two types of solid powders, with two target masses for each type of powder. Currently, only the two target samples for powder A have been completed, and not all targets have been completed. Next, repeat steps 4, 5, 6, and 7.

[0100] 4”. For the first target value of solid powder B, let w = 30.00 mg, and the first sampling target w / 3 = 10.00 mg. Since the solid powder has not been predicted for bulk density, based on experimental experience, the bulk density parameter k in the preset formula (1) is set to 1.50 and b is set to 0. According to the relationship (3), l1 = 6.67 mm is calculated.

[0101] After adjusting the length of the sampling section of the solid powder aspiration device to 6.67 mm, the sample was aspirated. The actual sample mass w1 = 10.10 mg. The bulk density k = 1.51424288 was obtained by linear regression least squares fitting, and the linear intercept b = 2.37851830 × 10⁻⁶. -16 .

[0102] Add the (l1, w1) data pair to the collected information recording matrix Z, and obtain the updated (k, b) = (1.51, 2.38 × 10^2)^2. -16 ).

[0103] 5”. The second sampling target is 2w / 3-w1=9.90mg. According to the relationship (4), l2=6.53mm is calculated.

[0104] After adjusting the tube length of the sampling section in the solid powder absorption device to 6.53 mm, the sample was sampled using an aspiration method. The actual sample mass w2 = 6.9 mg. The bulk density k = 1.05666156 was obtained by linear regression least squares fitting, and the linear intercept b = -8.55339043 × 10⁻⁶. -17 .

[0105] Add the (l2, w2) data pair to the collected information recording matrix Z, and obtain the updated (k, b) = (1.21, 2.22 × 10^2)^2. -17 ).

[0106] 6”. The third sampling target w-w1-w2=13.00mg, and l3=10.75mm was calculated according to the relationship (5).

[0107] After adjusting the length of the sampling section of the solid powder aspiration device to 10.75 mm, the sample was aspirated. The actual sample mass w3 = 12.30 mg. The bulk density k = 1.14418605 was obtained by linear regression least squares fitting, and the linear intercept b = 9.22161667 × 10⁻⁶. -16 .

[0108] The total amount of drug obtained from the three samplings was w1+w2+w3=29.20mg, which is 2.67% and 0.80mg less than the target w=30.00mg.

[0109] Add the (l3,w3) data pair to the collected information recording matrix Z, and obtain the updated (k,b) = (1.17, 5.85 × 10^3)^2. -16 ), so that the second target sample of drug B can be used for continued use.

[0110] 7”. This task involves two drugs, each with two target masses. Currently, only the two target samples of drug A and the first target sample of drug B have been completed. Not all targets have been completed. Next, repeat steps 4, 5, 6, and 7.

[0111] 4”'. For the second target value of drug B, let w = 20.00 mg, and the first sampling target w / 3 = 6.67 mg. This drug has undergone bulk density prediction in the previous process, therefore, based on step 6”, the bulk density parameter k in the preset formula (1) is 1.17, and b is 5.85 × 10 -16 According to the formula (3), l1 = 5.70 mm was calculated.

[0112] After adjusting the tube length of the sampling section in the solid powder absorption device to 5.70 mm, the sample was sampled using an aspiration method. The actual sample mass w1 = 8.60 mg. The bulk density k = 1.50877193 was obtained by linear regression least squares fitting, and the linear intercept b = -1.85134097 × 10⁻⁶. -16 .

[0113] Add the (l1,w1) data pair to the collected information recording matrix Z, and obtain the updated (k,b) = (1.38, 1.08 × 10^6)^2. -16 ).

[0114] 5”'. The second sampling target is 2w / 3-w1=4.73mg, and l2=3.43mm is calculated according to the relationship (4).

[0115] After adjusting the tube length of the sampling section in the solid powder absorption device to 3.43 mm, the sample was sampled using an aspiration method. The actual sample mass w2 = 4.9 mg. The bulk density k = 1.42857143 was obtained by linear regression least squares fitting, and the linear intercept b = -5.21361194 × 10⁻⁶. -17 .

[0116] Add the (l2, w2) data pair to the collected information recording matrix Z, and obtain the updated (k, b) = (1.41, 9.05 × 10^2)^2. -18 ).

[0117] 6”'. The third sampling target w-w1-w2=6.50mg, and l3=4.61mm was calculated according to the relationship (5).

[0118] After adjusting the tube length of the sampling section in the solid powder absorption device to 4.61 mm, the sample was sampled using an aspiration method. The actual sample mass w3 = 6.80 mg. The bulk density k = 1.47505423 was obtained by linear regression least squares fitting, and the linear intercept b = -8.19940511 × 10⁻⁶. -17 .

[0119] The total amount of drug obtained from the three samplings was w1+w2+w3=20.30mg. Compared with the target w=20.00mg, the error was 1.50% and 0.30mg respectively.

[0120] 7”'. This task involves two types of drugs, each with two target qualities. All targets have been achieved, and the workflow ends here.

Claims

1. A method for accurately predicting the bulk density of solid powder, characterized in that... Includes the following steps: 1) The length of the sampling tube inside the suction head of the solid powder suction device is set to l, and the mass of the sample obtained is set to w. The two satisfy the following linear relationship: (1) Where l is the independent variable, w is the dependent variable, k is the bulk density, and b is the linear intercept. The linear intercept represents the actual powder mass obtained when the lumen length l=0 under the influence of objective factors. Establish the discrete state transition equations for the sampling device: (2) Using the packing density and linear relationship intercept in equation (1) As a state parameter of this device, i.e. W represents process noise, and t represents discrete time intervals. 2) Establish matrix Z to store the collected and updated state values ​​at each time point, i.e. ; Where n is a natural number greater than or equal to 0, representing the total number of (k, b) data pairs of bulk density and linear intercept that have been fitted and recorded for this solid powder; if no experiments or records have been made for this solid powder before, then n=0, i.e., Z is empty; 3) When the target mass of solid powder is w, the sampling strategy is to sample in three steps. After three samplings, the mass of solid powder in the target container reaches w. 4) The first step of sampling targets w / 3. If the solid powder has been predicted in terms of bulk density, the recorded matrix Z is used, and random disturbances are removed based on the Kalman filter method to calculate (k, b); otherwise, the matrix Z is empty, and the default preset (k, b) = (1.5, 0) or (1.0, 0) is used. The lumen length l1 in the first sampling step is calculated based on the linear relationship: (3) After sampling, the actual mass obtained from the first sampling step is measured as w1. The (l1, w1) data pair is added to the data collection record matrix Z, and the updated (k, b) is obtained. 5) The sampling target for the second step is 2w / 3 - w1; the lumen length l2 in the second step sampling is calculated based on the linear relationship: (4) After sampling, the actual mass obtained from the second sampling step is measured as w2. The (l2, w2) data pair is added to the data collection information recording matrix Z, and the updated (k, b) is obtained. 6) The sampling target in the third step is w - w1 - w2; the lumen length l3 in the third step sampling is calculated based on the linear relationship: (5) The actual mass obtained from the third sampling step, w3, was obtained by absorption sampling and weighing. Before each sampling operation, the lumen length l used for sampling is given. i By measuring the actual mass w i Then, based on the linear relationship given in equation (1), the corresponding (k) can be obtained by fitting using the linear regression method. i b i This yields the bulk density at each sampling point. After several mass measurements and data fitting, a series of bulk density values ​​can be obtained. After Kalman filtering, a (k, b) value can be obtained, which is the bulk density. The (l3, w3) data pair is added to the data acquisition record matrix Z, and the updated (k, b) is obtained for subsequent sampling of solid powder. 7) For a task involving the quality of a single sampling target, the workflow ends here.

2. The method for accurately predicting the bulk density of solid powder as described in claim 1, characterized in that... For an integrated task process involving multiple sampling target qualities, repeat steps 4), 5), 6), and 7) until all sampling requirements are met.

Images