A work order feeding metering checking method based on linkage of BOM and process path

Abstract

The present application relates to the field of weighing instrument metrology, and discloses a work order feeding metering checking method based on linkage of BOM and process path. Theoretical consumption and limit difference are obtained according to the work order based on the bill of materials, and a verification scale value, a pre-weighing container and a scale interval are jointly selected, so that the indication before and after feeding is in the same target scale interval. Each weighing instrument is zeroed before the first work step and keeps the zero point compensation amount; zero load observation nodes are formed during the zero load period before and after the work step, and adjacent work steps reuse intermediate nodes, zero point drift is estimated from the front and rear nodes, and the compensation difference is subtracted. The leave-one-out fitting is performed on the central node of the rolling four-node group; the same node is confirmed as disturbed by two adjacent four-node groups, and the zero point innovation amount is corrected in the reverse direction according to the respective time ratio of adjacent work steps. The limit difference containing the expanded uncertainty is used to determine the control of subsequent work steps, and weighing and checking data are recorded.

Description

Technical Field

[0001] This invention belongs to the field of weighing and measurement technology, specifically relating to a work order material feeding measurement verification method based on the linkage between BOM and process path. Background Technology

[0002] The work order material input measurement verification is used to check whether the actual amount of material input into the process equipment is consistent with the theoretical amount specified in the BOM. The current practice is usually to install electronic scales at the weighing station, and the operator weighs each item according to the work order prompts and manually confirms it, or the manufacturing execution system reads the reading and compares it with the theoretical amount by a fixed percentage.

[0003] Among the publicly disclosed techniques, one approach is to measure the rate of change of the zero point over time when the weighing instrument is unloaded and extrapolate compensation based on the stored zero-point drift rate during loading; another approach is to obtain zero-load values ​​once before and once after the sample weighing to correct the weighing result; yet another approach is to obtain readings from the same device before and after the operation and calculate the difference to determine the net change. These approaches improve zero-point drift bias on the scale of a single weighing.

[0004] Current practices still have several shortcomings. The automatic zeroing and automatic zero-point tracking functions of the weighing instrument may include the continuously increasing load in the zero-point correction during slow material feeding, and long-term disabling of this function can lead to accumulated zero-point drift. Multi-gradient weighing instruments automatically switch graduation intervals based on the current load; when the two readings of the difference method are in different intervals, the difference can only be obtained using the coarser calibration graduation value. Residual material on the weighing platform or incomplete removal of pre-weighed containers can disturb the zero-load value; directly using this zero-load value will introduce the disturbance into the weighing result. Fixed percentage tolerances are not correlated with the weighing instrument's calibration graduation value and measurement uncertainty, and the risk of incorrect acceptance and rejection of boundary batches lacks quantitative basis. Summary of the Invention

[0005] This invention aims to solve the problems in existing work order material feeding measurement verification where the zero-point correction function and the feeding process interfere with each other in terms of timing, the two readings of the difference method are located in different scale intervals, there is no way to detect the disturbance of the zero load value itself, and there is no correlation between the verification tolerance and the measurement capability. It provides a work order material feeding measurement verification method based on the linkage between BOM and process path.

[0006] To solve the above problems, the technical solution of the present invention is implemented as follows: A method for verifying the metering of material feed in a work order based on the linkage between BOM and process path is implemented at a weighing station equipped with a multi-gradient weighing instrument. The multi-gradient weighing instrument has at least two graduation intervals, each with a verification graduation value. For each multi-gradient weighing instrument participating in the work order, it is zeroed once before the first feeding step of that work order, and a zero-point compensation is maintained during the period the instrument is handling that work order, ensuring that the reading is taken from a fixed zero-point reference. The method includes the following steps: Step 1: Expand the work order according to the hierarchical structure of the Bill of Materials (BOM) to obtain the theoretical usage and tolerance rate, and use the product of the tolerance rate and the theoretical usage as the limit of error; obtain the material feeding step sequence according to the process sequence of the process path, and arrange the material feeding steps assigned to the same multi-gradient weighing instrument in the order of their sequence to obtain the instrument step sub-sequence. Step 2: The target quantity to be used is initially taken as the theoretical quantity. The scale interval in which the verification scale value is within one-third of the limit error is selected as the target scale interval. A pre-weighed container is selected so that the tare weight and the sum of the tare weight and the target quantity to be used are both within the target scale interval. Step 3: Collect one set of readings during the zero-load period of the weighing platform before and after the feeding step. Take the arithmetic mean of each set of sampled values ​​as the original zero-load value and the arithmetic mean of the sampling time as the effective observation time to obtain the zero-load observation node. The same zero-load observation node is reused between two adjacent feeding steps in the tool step sub-sequence. The zero-point drift rate is obtained by dividing the difference between the original zero-load values ​​of the two zero-load observation nodes before and after the feeding step by the difference between the effective observation times of the two zero-load observation nodes. Step 4: Within the target graduation range, read the reading before feeding and the reading after feeding. The actual feeding amount is obtained by adding the difference between the reading before feeding and the reading after feeding to the product of the zero drift rate and the difference between the two reading times. Step 5: For each of the two middle zero-load observation nodes within a consecutive sequence of four zero-load observation nodes in the tooling step sub-sequence, perform a straight-line fitting to obtain the standardized one-left-off residual; confirm the consistency of the verification results of the same zero-load observation node for two four-node groups with three adjacent and overlapping zero-load observation nodes; when the absolute value of the standardized one-left-off residual of the same zero-load observation node in two adjacent four-node groups is within 3, it is recorded as a verified zero-load observation node; when both the preceding and following zero-load observation nodes of the feeding step become verified zero-load observation nodes, the actual feeding amount obtained in Step 4 is taken as the maximum. Final actual input; Remove two adjacent four-node groups from the same zero-load observation node and perform the aforementioned straight-line fitting, and record the fitted values ​​obtained at the effective observation time of the zero-load observation node as the first fitted value and the second fitted value, respectively; If the same zero-load observation node is unique in both adjacent four-node groups with a standardized left-one residual absolute value exceeding 3, and the absolute value of the difference between the first fitted value and the second fitted value is within the expanded uncertainty of that difference, then the zero-load observation node is identified as a disturbed node, where the expanded uncertainty is the standard deviation of the difference between the first fitted value and the second fitted value. Multiply the standard uncertainty by a coverage factor of 2, and take the fitted value with the smaller prediction standard uncertainty between the first and second fitted values ​​as the confirmed fitted value. The difference between the original zero-load value and the confirmed fitted value at the zero-load observation node is taken as the zero-point innovation. The feeding step with the disturbed node as the subsequent zero-load observation node is recorded as the previous feeding step, and the feeding step with the disturbed node as the preceding zero-load observation node is recorded as the next feeding step. The difference between the reading time of the indication after feeding and the reading time of the indication before feeding in the previous feeding step is divided by the effective observation time of the disturbed node and the reading time of the previous feeding step. The difference between the effective observation times of the zero-load observation node is used to obtain the previous time ratio; the difference between the reading time of the indication after feeding and the reading time of the indication before feeding in the next feeding step is divided by the difference between the effective observation time of the subsequent zero-load observation node and the effective observation time of the disturbed node in the next feeding step to obtain the next time ratio; the product of the zero-point innovation amount and the previous time ratio is subtracted from the actual feeding amount of the previous feeding step, and the product of the zero-point innovation amount and the next time ratio is added to the actual feeding amount of the next feeding step to obtain the final actual feeding amount of each of the two adjacent feeding steps; other cases are transferred to pending review; Step 6: The absolute value of the difference between the final actual amount and the theoretical amount is within the limit of error as the qualification condition, and the execution permission of subsequent feeding steps is controlled according to the judgment result.

[0007] Furthermore, each scale interval also has a lower load limit and an upper load limit; the method for selecting the target scale interval and the pre-nominated container in step 2 is as follows: set a boundary margin, which is the sum of the upper limit of load fluctuation, the upper limit of indication stability fluctuation, and the maximum allowable cumulative zero drift of the multi-scale weighing instrument in the work order; select a combination of multi-scale weighing instrument, scale interval, and pre-nominated container that simultaneously meets the following three conditions, and take the scale interval in the combination as the target scale interval: Condition 1, the verification scale interval value of the scale interval is within one-third of the limit error; Condition 2, the lower limit of the actual tare weight of the pre-nominated container is greater than or equal to the sum of the lower load limit and the boundary margin of the scale interval; Condition 3, the sum of the upper limit of the actual tare weight of the pre-nominated container, the target quantity to be delivered, the upper limit of the pre-weighing overshoot, and the boundary margin is less than or equal to the upper load limit of the scale interval.

[0008] Furthermore, if there are two or more combinations that simultaneously satisfy conditions 1, 2, and 3, the combination with the smallest verification scale value is selected first. If there are two or more combinations with the smallest verification scale value, the lower limit of the actual tare weight of the pre-weighed container minus the boundary margin is taken as the conservative load lower limit, and the upper limit of the conservative load is taken as the sum of the upper limit of the actual tare weight, the target quantity to be fed, the upper limit of the pre-weighing overshoot, and the boundary margin. The distance from the conservative load lower limit to the lower limit of the scale interval load and the distance from the upper limit of the scale interval load to the conservative load upper limit are divided by the width of the scale interval, and the smaller of the two quotients is taken as the normalized boundary margin. The combination with the largest normalized boundary margin is selected. If the combination that simultaneously satisfies conditions 1, 2, and 3 is an empty set, the information of insufficient measurement capacity is output, and the process is switched to the re-judgment of multi-scale weighing instruments with smaller verification scale values, multiple feedings, or manual weighing.

[0009] Furthermore, the weighing station is equipped with a occupancy detection device to confirm zero load on the weighing platform, and the process equipment is equipped with a complete transfer confirmation device. The method for collecting one set of indications in step 3 is as follows: when the occupancy detection device outputs a zero load signal on the weighing platform and the complete transfer confirmation device outputs a complete transfer confirmation signal of the previous feeding step within the step interval between two adjacent feeding steps, the zero load observation window is opened, and the indications are collected according to the decorrelation interval. The decorrelation interval is the time interval corresponding to the first time that the autocorrelation function of the zero load value sequence of the multi-gradient weighing instrument falls into the 95% confidence band of the autocorrelation function and remains within the confidence band for three consecutive hysteresis points thereafter.

[0010] Furthermore, the Type A component is obtained by dividing the standard deviation of the sampled values ​​within the zero-load observation window by the square root of the number of sampling points, and the resolution component is obtained by dividing the minimum verification scale value applicable at zero load by the square root of 12. The root sum of the squares of the Type A component and the resolution component is taken as the standard uncertainty of the zero-load observation node. The readings are continuously collected until the following condition is met: the root sum of the squares of the standard uncertainties of the two zero-load observation nodes before and after the feeding step is multiplied by the ratio of the upper bound of the difference between the reading time registered in the feeding step and the lower bound of the difference between the effective observation time, and then multiplied by 3. The resulting product is within one-quarter of the tolerance limit. When a zero-load observation node is reused by two adjacent feeding steps, it is calculated according to the two adjacent feeding steps respectively, and the one with more required sampling points is taken as the condition for stopping the collection. The minimum number of sampling points and the maximum window duration are set. When the duration of the zero-load observation window reaches the maximum window duration, the feeding step is transferred to the pending verification stage.

[0011] Furthermore, the method for reading the pre-feeding and post-feeding readings in step 4 is as follows: Place the pre-weighed container containing the material on the weighing platform, and read the pre-feeding reading after the reading range is within 1 verification scale value for 2 consecutive seconds; transfer the material in the pre-weighed container to the process equipment, reset the pre-weighed container to the weighing platform, and read the post-feeding reading after the reading again meets the requirement that the reading range is within 1 verification scale value for 2 consecutive seconds; the material remaining in the pre-weighed container is included in the post-feeding reading, and the actual feeding amount is the net feeding amount of the pre-weighed container; when the complete transfer confirmation device outputs a complete transfer confirmation signal, the actual feeding amount is the receiving amount of the process equipment.

[0012] Further, the standardized leave-one residual obtained in step 5 is as follows: The central zero-load observation node is removed. Using the reciprocal of the square of the standard uncertainty of the remaining three zero-load observation nodes as weights, a weighted least squares straight line is fitted between the effective observation time and the original zero-load value. The fitted value of the line at the effective observation time of the removed node is obtained. The difference between the original zero-load value and the fitted value of the removed node is divided by the standard uncertainty of the predicted residual, which is the sum of the standard uncertainty of the removed node and the predicted standard uncertainty of the fitted value, calculated using the root of the square. The standardized leave-one residual is obtained. When the standard uncertainties of the four nodes are equal, the standard uncertainty of the predicted residual is equal to the product of this standard uncertainty and the predicted residual coefficient. The predicted residual coefficient is the square root of the sum of 1, one-third, and the deviation quotient. The deviation quotient is the sum of the effective observation time of the removed node and the remaining... The square of the difference between the mean effective observation times of the three nodes is divided by the sum of the squares of the deviations of the effective observation times of the other three nodes from that mean. Two adjacent four-node groups are each moved out of the same node to perform the straight-line fitting, and the fitted values ​​obtained at the effective observation time of that node are recorded as the first fitted value and the second fitted value, respectively. When the absolute value of the standardized left-one residual of the same node in two adjacent four-node groups is within 3, it is recorded as a verified zero-load observation node. When the same node is the only node in two adjacent four-node groups that exceeds 3, and the absolute value of the difference between the first fitted value and the second fitted value is within the expanded uncertainty of that difference, it is identified as a disturbed node. The expanded uncertainty is the standard uncertainty of the difference between the first fitted value and the second fitted value multiplied by a coverage factor of 2. All other cases are transferred to a pending verification process, and additional verification nodes are collected.

[0013] Furthermore, after obtaining the post-zero load observation node of the feeding step, the preliminary actual feeding quantity of the feeding step is obtained. When the preliminary actual feeding quantity is used to replace the final actual feeding quantity to perform the qualification judgment condition and is judged to be qualified, the next feeding step is allowed to perform weighing. When both the pre-zero load observation node and the post-zero load observation node of the feeding step become verified zero load observation nodes or are corrected after being confirmed to be disturbed, the final actual feeding quantity of the feeding step is obtained. When the final actual feeding quantity is used to perform the qualification judgment condition and is judged to be qualified, the next feeding step is allowed to perform weighing. When the final judgment is unqualified, the feeding steps that have not yet been executed are frozen, and the materials involved in the next feeding step that has been executed according to the preliminary judgment are transferred to batch isolation and manual verification. Two initial verification nodes are continuously collected at the beginning of the instrument step subsequence. When the number of zero load observation nodes obtained is less than four or the test results are pending verification, the weighing platform is kept at zero load and additional verification nodes are collected. Two terminal verification nodes are continuously collected at the end of the instrument step subsequence.

[0014] Furthermore, the qualification criterion is replaced by the absolute value of the difference between the final actual input quantity and the theoretical input quantity, plus the expanded uncertainty of the final actual input quantity being within the tolerance limit. The expanded uncertainty of the final actual input quantity is obtained by multiplying the fractional quantification component of the pre-input and post-input values, the repeatability component of the target scale interval, and the zero-point drift compensation component by the root of the square and the inclusion factor 2. The fractional quantification component is the square root of the calibration scale value divided by 12 and then multiplied by the square root of 2. The zero-point drift compensation component is the larger of the standard uncertainty calculated based on the node standard uncertainty, the standard uncertainty predicted by the fitted value, and the ratio of the actual time to the standard uncertainty and the tolerance of one-twelfth.

[0015] Further, the sum of the final actual quantities weighed in each step of the same feeding process is recorded as the cumulative actual quantity, and the absolute value of the difference between the cumulative actual quantity and the theoretical quantity is recorded as the deviation. The expanded uncertainty of the cumulative actual quantity is obtained by multiplying the standard uncertainty of each final actual quantity and the covariance generated by the shared zero-load observation node by a coverage factor of 2. The condition for acceptance is that the sum of the expanded uncertainty of the deviation and the cumulative actual quantity is within the tolerance limit. When the acceptance condition is met, the process is deemed acceptable and subsequent feeding steps are permitted to be weighed. When the acceptance condition is not met and the cumulative actual quantity is less than the theoretical quantity, the process is deemed under-feeding, and the difference between the theoretical quantity and the cumulative actual quantity is taken as the new target quantity to be fed. The process returns to step 2 and repeats the steps within the supplementary feeding time limit set in the process path. Steps 3 to 5 are performed, and the cumulative actual amount and its expanded uncertainty are recalculated after each supplementary feeding. If the qualification judgment condition is not met and the cumulative actual amount is greater than the theoretical amount, it is judged as overfeeding, and subsequent feeding steps are frozen and transferred to rework, batch isolation or manual handling. If the qualification judgment condition is not met and the cumulative actual amount is equal to the theoretical amount, it is transferred to the insufficient measurement capability review. When the subsequent feeding step is changed to another multi-division weighing instrument, the zero-load observation node of the original multi-division weighing instrument is terminated, and the other multi-division weighing instrument is set to zero separately before its first feeding step and an independent starting verification node and zero-load observation node chain are established. If the absolute value of any zero-load original value exceeds the maximum allowable cumulative zero drift, the subsequent feeding steps of the instrument are frozen and transferred to measurement verification.

[0016] By adopting the above technical solution, the present invention has the following beneficial effects: the absolute tolerance given by the BOM is converted into the admission conditions of the indexing interval; the actual tare weight range of the pre-weighed container, the target quantity to be fed, the pre-weighed quantity overshoot and the boundary margin jointly limit the conservative load range, so that the two indications of the same feeding step are kept in the same target indexing interval; the zero-point compensation amount remains unchanged during the work order undertaken by the instrument; the zero-load observation window is jointly triggered by the step boundary event, the weighing platform occupancy signal and the complete transfer confirmation signal; the feeding process is not absorbed by the automatic zero-point tracking; the sampling amount of the zero-load observation node is determined by the limit and the time range registered by the process path, and the indication resolution is taken into account; the same node is confirmed to be consistent in two adjacent four-node groups; the single-group entanglement caused by the disturbance of the end reference node will not directly trigger automatic correction; after the disturbed node is confirmed, corrections with different amplitudes and opposite directions are made according to the time ratio of each of the adjacent steps; the judgment rule containing expanded uncertainty and the underfeed, qualified and overfeed processing together form an auditable work step permission record. Attached Figure Description

[0017] Figure 1 This is a schematic diagram illustrating the timing coupling principle between the zero-load observation node and the material feeding step under a fixed zero-point reference in an embodiment of the present invention. Figure 2 This is a schematic diagram illustrating the principle of the combined selection of weighing instruments, graduation ranges, and pre-weighing containers in an embodiment of the present invention. Figure 3 This is a schematic diagram of the difference subtraction weighing and linear zero-point drift compensation principle in an embodiment of the present invention, where (a) is the indication trajectory of one feeding step, and (b) is the change of the original zero-load value over time in the same period. Figure 4 This is a schematic diagram of the four-node leave-one test principle in an embodiment of the present invention, where (a) is the straight-line fitting and zero-point innovation quantity of a four-node group, and (b) is the standardized leave-one residual of the two zero-load observation nodes under test in the four-node group; Figure 5 This is a schematic diagram illustrating the principle of how zero-point innovation amount makes reverse corrections to two adjacent feeding steps at different ratios in an embodiment of the present invention. Figure 6 This is a schematic diagram illustrating the variation of the standard uncertainty of the zero-load observation node with the number of effective sampling points and the determination of the lower limit of the number of effective sampling points in an embodiment of the present invention. Figure 7 This is a schematic diagram of the weighing station, the closed transfer channel, and the complete transfer confirmation device in an embodiment of the present invention; Figure 8 This is a schematic diagram illustrating the principle of consistency confirmation between the zero-load observation node chain and the adjacent four-node group in an embodiment of the present invention. Detailed Implementation

[0018] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. This embodiment takes a work order from a resin intermediate batching workshop as an example. All weighing instruments involved in the work order are within the validity period of verification or calibration. The verification record or calibration certificate specifies each graduation interval, verification graduation value, and repeatability data. Before the work order is completed, a standard weight with a valid traceability certificate is used for performance verification. The monotonic clock of the control terminal is synchronized with the workshop clock. The workshop has several weighing stations, each equipped with one tabletop multi-gradient weighing instrument with a maximum weighing capacity of 30,000 grams and three graduation intervals: the first graduation interval has a lower load limit of 0 grams, an upper load limit of 6,000 grams, and a verification graduation value of 1 gram; the second graduation interval has a lower load limit of 6,000 grams, an upper load limit of 15,000 grams, and a verification graduation value of 2 grams; and the third graduation interval has a lower load limit of 15,000 grams, an upper load limit of 30,000 grams, and a verification graduation value of 5 grams. Multi-gradient weighing instruments automatically determine the graduation range based on the current load. The external control terminal does not directly specify the graduation range; this embodiment limits the load range to ensure the weighing instrument operates within the target graduation range. A through-beam photoelectric occupancy detection device is installed next to the weighing platform at the weighing station to output a zero-load signal for the platform. The process equipment is equipped with a complete transfer confirmation device.

[0019] Figure 7 This embodiment illustrates the configuration of the weighing station and the material transfer device. The weighing platform of the multi-gradient weighing instrument is supported on a base by a load cell, the output of which is transmitted to the control terminal via a cable. The pre-weighing container is placed entirely on the weighing platform, containing the material. The transmitter and receiver of the occupancy detection device are positioned on opposite sides of the weighing platform. After the pre-weighing container moves away from the platform, it outputs a zero-load signal, which is not contingent on the indicated value being close to zero. The material enters the process equipment through a closed, branchless transfer channel. The complete transfer confirmation device simultaneously receives three signals: source venting confirmation, channel closure confirmation, and process equipment material arrival confirmation. When all three signals are valid for a set holding time, a complete transfer confirmation signal is output. The source venting confirmation is generated when the discharge valve completes its specified stroke and the material detection signal at the container outlet is reset; the channel closure confirmation is generated by the valve position or cover interlock contact; and the material arrival confirmation is generated by the material level at the process equipment inlet or by the material arrival sensor. The actual amount fed into the process equipment is considered as the amount of material received when there is no leakage or branch in the closed channel and the emptying is confirmed. If any of the confirmation conditions are not met, the actual amount fed into the process equipment only represents the net amount of material transferred out of the pre-nominated container.

[0020] After a work order is initiated, each multi-gradient weighing instrument participating in the work order undergoes zeroing once before its first feeding step. Automatic zeroing, automatic zero-point tracking, and other functions that update the zero-point compensation are paused via the instrument's calibrated control interface, without altering the calibration coefficient, calibration graduation value, or other legally mandated metrological parameters. While the zero-point compensation remains constant, the displayed value changes with load and zero-point drift. The reading when the weighing platform is unloaded is the original zero-load value relative to this fixed zero-point reference, with its positive direction aligned with the positive direction of the load reading. If the weighing instrument does not support this operating mode, the original reading channel without dynamic zero-point correction can be read; if this channel is unavailable, the instrument will not be included in the joint selection process. Each weighing instrument maintains its own fixed zero-point reference and zero-load value sequence; zero-load observation nodes are not reused between instruments.

[0021] In step 1, the work order is expanded layer by layer according to the Bill of Materials (BOM) hierarchy to obtain the material code, theoretical usage, and tolerance rate of the material master data for each process node. In this embodiment, the initiator premix with material code M-2041 is used, with a theoretical usage of 4500 grams and a tolerance rate of 0.001. The product of the tolerance rate and the theoretical usage is used as the limit of error. In the formula, L is the tolerance limit, γ is the tolerance rate, and Q0 is the theoretical dosage; in this embodiment, the tolerance limit is 4.5 grams. The material codes are mapped to weighing stations according to the sequence of processes in the process path, resulting in a feeding step sequence; then, the feeding steps assigned to the same multi-gradient weighing instrument are arranged in chronological order, resulting in an instrument step sub-sequence. The process path registers the upper bound Tr of the difference between the reading time and the lower bound Tn of the difference between the effective observation time for each feeding step. In this embodiment, Tr is 600 seconds and Tn is 1200 seconds. Two planned quantities are determined before weighing begins and are used to calculate the number of sampling points required for zero-point drift compensation under the most unfavorable time conditions.

[0022] Step 2 involves joint selection and matching among various multi-gradient weighing instruments, graduation intervals, and pre-weighed containers. The boundary margin is the sum of the upper limit of load fluctuation, the upper limit of indication stability fluctuation, and the maximum permissible cumulative zero drift of the weighing instrument within this work order. Each upper limit is determined by pre-work order performance verification and historical monitoring data at the specified confidence level and recorded in the instrument's capability record. In this embodiment, the three upper limits are 6 grams, 4 grams, and 10 grams, respectively, with a boundary margin M of 20 grams. The lower and upper limits of the actual tare weight of the pre-weighed container are taken as the lower and upper envelopes of the periodic verification values ​​under clean and dry conditions, respectively, and the uncertainty of the tare weight verification is included. Before each use, the empty container's indication is checked again; containers exceeding the envelope are excluded from this selection and matching. The boundary margin is used to cover load fluctuations and zero-point changes in the vertical direction, ensuring that the two indications before and after feeding do not exceed the graduation interval boundary even under the most unfavorable conditions.

[0023] Condition 1 is the quantification admission condition for the scale interval, requiring the calibration scale value of the scale interval to meet the following requirements. In the formula, e is the calibration scale value of the scale interval, and L is the limit of error. When the indicated value is rounded to the nearest scale value, the rounding error limit for a single reading is half of the calibration scale value. The difference subtraction method uses two readings, and the rounding error limit for the most unfavorable difference is one calibration scale value. One-third of the acceptance factor limits the quantization error to a portion of the limit of error, and the remaining portion is used for repeatability and zero-point drift compensation. In this embodiment, L is 4.5 grams, and the upper limit of acceptance is 1.5 grams; the calibration scale value of 1 gram for the first scale interval meets the condition, while 2 grams and 5 grams for the second and third scale intervals do not. The BOM limit of error is thus converted into the acceptance condition for the scale interval.

[0024] Conditions 2 and 3 constrain the conservative load envelope corresponding to the nominal container, satisfying In the formula, Tlo and Thi are the lower and upper limits of the actual tare weight of the pre-weighed container, respectively; Blo and Bhi are the lower and upper limits of the load in the scale interval, respectively; q is the target quantity to be used; O is the upper limit of the pre-weighing overshoot; and M is the boundary margin. The conservative lower limit of the load is Tlo minus M, and the conservative upper limit of the load is the sum of Thi, q, O, and M. For the first weighing, q is taken as the theoretical quantity of 4500 grams, and O is taken as 45 grams. The actual tare weight ranges of containers T-08, T-15, and T-32 are 795 grams to 805 grams, 1190 grams to 1210 grams, and 7480 grams to 7520 grams, respectively. Within the first indexing interval, the conservative load envelope of container T-08 is 775 g to 5370 g, and that of container T-15 is 1170 g to 5775 g, both of which satisfy conditions 2 and 3; the upper limit of the conservative load of container T-32 is 12085 g, which exceeds 6000 g, and its second indexing interval does not satisfy condition 1.

[0025] There are two combinations that satisfy all three conditions, and both have a test scale value of 1 gram. Let Alo represent the conservative lower bound Tlo minus M, and Ahi represent the conservative upper bound Thi, q, O, and M. The normalized boundary margin is calculated as follows: The calculation is performed, where g is the normalized boundary margin, and Blo and Bhi are the lower and upper load limits of the indexing interval, respectively. For container T-08, the lower boundary distance is 775 grams, and the upper boundary distance is 630 grams; g is 630 divided by 6000, which is 0.105. For container T-15, the corresponding distances are 1170 grams and 225 grams; g is 225 divided by 6000, which is 0.0375. The system selects container T-08 with the larger g value and the first indexing interval. The normalized boundary margin uses the remaining distance from the nearest switching boundary, ensuring that material moisture absorption, container adhesion, or zero-point changes can still be absorbed by the reserved distance.

[0026] Figure 2The combined selection results of this embodiment are shown. The vertical axis represents the load, the horizontal dashed line represents the switching boundary between the first and second scale intervals, and the three rectangles represent the conservative load envelope after taking into account the boundary margin M. The lower and upper conservative load limits for container T-08 are 775 grams and 5370 grams, respectively, and the distances from the lower and upper switching boundaries are 775 grams and 630 grams, respectively; the corresponding distances for container T-15 are 1170 grams and 225 grams; the upper conservative load limit for container T-32 is 12085 grams, and the verification scale value of the second scale interval does not meet condition 1, so it is not selected. The boundary margin of 20 grams has already been included in the upper and lower boundaries of the rectangles, and is not enlarged separately according to the vertical axis ratio in the figure. When the combination that satisfies all three conditions is an empty set, the system outputs a measurement capacity insufficient information and switches to an instrument with a smaller verification scale value, multiple feedings, or manual weighing process, while the limit remains unchanged.

[0027] Step 3: Collect zero-load readings within each step interval. The start point of the step interval is the moment the pre-weighed container moves off the weighing platform in the previous feeding step, and the end point is the moment the pre-weighed container is placed on the weighing platform in the next feeding step. The zero-load observation window is opened when the occupancy detection device outputs a zero-load signal for the weighing platform, and the complete transfer confirmation device outputs a complete transfer confirmation signal for the previous feeding step. The no-load reading under a fixed zero-point reference may deviate from 0 grams with accumulated zero-point drift. Using an external status signal to open the window avoids missing the window due to the reading being close to 0 grams.

[0028] Within the window, readings are collected at decorrelation intervals. Digital filtering in the weighing instrument causes correlation between adjacent outputs; directly taking points at the highest output rate will overestimate the amount of independent information. Before the work order, a zero-load value sequence of length N0 is recorded under the unloaded state of the weighing platform, and the sample autocorrelation function is calculated. The 95% confidence band under the white noise assumption is taken as ±1.96 divided by the square root of N0. When the autocorrelation function first enters the confidence band and remains within the confidence band for three consecutive lag points, the time interval corresponding to the first entry point is taken as the decorrelation interval. In this embodiment, the decorrelation interval is 0.5 seconds, and one point is taken every 0.5 seconds within the window. Sampling points are treated as approximately uncorrelated; if performance verification still shows residual correlation, the number of effective sampling points is calculated using the autocorrelation sequence and replaced with the original number of sampling points.

[0029] The standard uncertainty of each zero-load observation node is calculated as follows: The calculation is performed using the formula, where u is the standard uncertainty of the zero-load observation node, s is the standard deviation of the sampled values ​​within the zero-load observation window, n is the number of effective sampling points, and e0 is the minimum calibration scale applicable at zero load. The first term is the square of the Type A standard uncertainty of the arithmetic mean of the original zero-load values, and the second term is the square of the resolution component. When the indicated value is quantized according to the calibration scale, the quantization error is uniformly distributed within one scale width, and its standard uncertainty is the square root of e0 divided by 12. In this embodiment, s is 6 grams, e0 is 1 gram, and the resolution component is 0.2887 grams. The resolution component does not decrease with the increase of the number of sampling points, constituting the lower limit of the standard uncertainty of the zero-load observation node.

[0030] Sampling continues until the indicated conditions are met. In the formula, Tr is the upper bound of the difference between the reading times registered for the feeding step in the process path, Tn is the lower bound of the difference between the effective observation times of the two zero-load observation nodes before and after the feeding step, uk-1 and uk are the standard uncertainties of the preceding and following zero-load observation nodes, respectively, and L is the limit of error. The uncertainties of the two nodes are combined using the root of square, multiplied by the ratio of the difference between the reading times to the difference between the effective observation times, to obtain the standard uncertainty of the zero-point drift compensation; then multiplied by the coverage factor 3, and the result is required to be within one-quarter of the limit of error. In this embodiment, one-quarter of the limit of error is 1.125 grams, and the time ratio is 600 seconds divided by 1200 seconds, i.e., 0.5. Assuming the standard uncertainties of the two nodes are equal, the upper limit of the standard uncertainty of a single node is 0.5303 grams; the resolution component is 0.2887 grams, and the upper limit of the Type A component is 0.4449 grams. When the standard deviation of the sampled value is 6 grams, the number of effective sampling points is calculated to be 181.8947, which is rounded up to 182 points. When sampling at 0.5-second intervals, the time span from point 1 to point 182 is 90.5 seconds, and the engineering setting can be no less than 91 seconds.

[0031] Figure 6 The process of determining the effective number of sampling points is illustrated. The horizontal axis represents the effective number of sampling points, and the vertical axis represents the standard uncertainty: the dashed line represents the Type A component, the dotted line represents the resolution component, the solid line represents the combined standard uncertainty, and the dotted-dash line represents the upper limit of the standard uncertainty of 0.5303 grams; the intersection of the solid line and the dotted-dash line corresponds to point 182, and the value at that point is 0.4449 grams. When only the Type A component is used for calculation and the resolution component is omitted, the resulting number of sampling points is too small, and the zero-point drift compensation may exceed the allocated uncertainty share.

[0032] When a zero-load observation node is reused by two adjacent feeding steps, the required number of sampling points is calculated based on the upper limit of the difference between the reading times of the two feeding steps, the lower limit of the difference between the effective observation times, and the limit error. The larger value is taken as the stopping condition for data acquisition. The minimum number of sampling points and the maximum window duration are set to 30 points and 300 seconds, respectively. If the sampling condition is not met by the end of the maximum window duration, the relevant feeding step is transferred to pending verification. If the absolute value of any zero-load original value exceeds the maximum allowable cumulative zero drift of 10 grams, the issuance of new step permits on that instrument is stopped, and the process is transferred to metrological verification. After the window ends, the arithmetic mean of all sampled values ​​is taken as the zero-load original value, and the arithmetic mean of all sampling times is taken as the effective observation time. The zero-load original value corresponds to the average of the values ​​displayed within the window; therefore, the effective observation time uses the average of the sampling times, not the window end time.

[0033] Figure 1 This diagram illustrates the temporal coupling relationship between the zero-load observation node and the feeding step in this embodiment. The horizontal axis represents time, and the vertical axis represents the readings of the multi-gradient weighing instrument. The reading trajectory is close to zero when the weighing platform is unloaded, jumps to the pre-feeding reading when the pre-weighing container is placed on the weighing platform, falls back to the post-feeding reading after the material is transferred out and the pre-weighing container is reset, and returns to zero again after the pre-weighing container is moved away from the weighing platform. The four narrow bands with profile lines represent four zero-load observation windows, with effective observation times of 200 seconds, 1400 seconds, 2600 seconds, and 3800 seconds, corresponding to nodes N1 to N4. Feeding step S1 is located between nodes N1 and N2, and feeding step S2 is located between nodes N2 and N3; node N2 is both the subsequent zero-load observation node of feeding step S1 and the preceding zero-load observation node of feeding step S2, meaning it is reused by two adjacent feeding steps. Each zero-load observation node in the same instrument's sub-sequence is connected end-to-end to form a unified zero-point baseline. When the next feeding step is reclassified to another multi-gradient weighing instrument via step 2, the reuse of the original multi-gradient weighing instrument's zero-load observation node terminates, and the other multi-gradient weighing instrument is set to zero separately and its initial zero-load observation node is established. The zero-load value sequences of the two instruments are not merged.

[0034] Zero drift rate according to The calculation is performed, where v is the zero-point drift rate, Zk and Zk-1 are the original zero-load values ​​of the subsequent and preceding zero-load observation nodes of the feeding step, respectively, and τk and τk-1 are the effective observation times for the two nodes. In this embodiment, the original zero-load value of node N1 is 2 grams and the effective observation time is 200 seconds, and the original zero-load value of node N2 is 14 grams and the effective observation time is 1400 seconds. Therefore, the zero-point drift rate of the feeding step S1 is 0.01 grams per second.

[0035] Step 4: Perform the differential weighing method. Place the pre-weighed container containing the material on the weighing platform. First, verify that the current reading is within the conservative load envelope of the target graduation range. Then, wait for the range of readings to be within 1 times the verification graduation value for 2 consecutive seconds. At time 400, read the reading before feeding: 5303 grams. Transfer the material to the process equipment and reset the pre-weighed container to the weighing platform. Verify the target graduation range and stability conditions again. At time 1000, read the reading after feeding: 806 grams. If any reading exceeds the conservative load envelope, the reading is invalid and the process returns to step 2. The differential weighing method cancels out the tare weight of the pre-weighed container in the difference between the two readings. The tare weight in the container ledger is only used for joint selection. The residual material on the inner wall of the container is included in the reading after feeding. The resulting difference corresponds to the net amount transferred out of the pre-weighed container.

[0036] Actual investment amount The calculation is as follows: Q represents the actual amount of material fed in this feeding step; W1 and W2 represent the readings before and after feeding, respectively; t1 and t2 represent the two reading times; and v represents the zero-point drift rate. Under a fixed zero-point reference, any reading is equal to the sum of the actual load and the additive zero-point drift at that moment. W1 minus W2 also subtracts the increase in zero-point drift during the period from t1 to t2, so the product of v and t2 minus t1 is added back to the difference. In this embodiment, W1 minus W2 is 4497 ​​grams, the zero-point drift compensation is 6 grams, and the initial actual amount of material fed in feeding step S1 is 4503 grams.

[0037] Figure 3 The above process is illustrated. (a) shows the indication trajectory of feeding step S1, marking the indication before feeding, the indication after feeding, and the difference between the reading times; (b) shows the change of the original zero-load value over time during the same period. The straight line is determined by node N1 and node N2, with values ​​of 4.0 grams and 10.0 grams at the two reading times, respectively. The difference of 6.0 grams is the zero-point drift of this feeding step. Step 4 compensates for the additive zero-point drift that changes approximately linearly with time, but does not compensate for sensitivity drift, load nonlinearity error, creep after loading and unloading, or step offset caused by mechanical impact; Step 5 is used to verify whether the zero-load node sequence conforms to the assumptions of single-point disturbance and local linearity. When the complete transfer confirmation signal is valid, the actual feeding amount is taken as the material received by the process equipment; when the confirmation signal is invalid, the actual feeding amount only represents the net transfer amount of the pre-nominated container, and the material received is determined by independent metering means on the process equipment side.

[0038] Step 5 addresses disturbances at zero-load observation nodes. Residual material on the weighing platform, incomplete removal of the pre-weighing container, or accidental overlap of tooling fixtures can all cause the original zero-load value to deviate from the local linear trend. When an intermediate node is too high, the preceding feeding step with that node as the subsequent node will have a larger zero-point drift rate, while the following feeding step with that node as the preceding node will have a smaller zero-point drift rate. The two actual feeding quantities are affected in opposite directions, and the amplitude is determined by the ratio of the difference between their respective reading times to the node's time span. The deviation between two adjacent actual feeding quantities and the theoretical quantity also includes their respective unknown actual feeding deviations. Node disturbances cannot be identified solely based on two actual feeding quantities; therefore, the disturbed node is verified by the zero-load original value sequence itself.

[0039] From the node chain of the same instrument step subsequence, select four consecutive zero-load observation nodes, and designate the second and third nodes as the checked nodes. Each time, remove one checked node, and use the reciprocal of the square of the standard uncertainty of the remaining three nodes as weights to perform a weighted least squares linear fit based on the effective observation time and the original zero-load value. When the standard uncertainties of the remaining three nodes are equal, the weighted fit degenerates into an ordinary least squares fit, and is then... The fitted value at the time point of the removed node is obtained. In the formula, Z with a cusp and subscript k is the fitted value, Z and τ with horizontal lines are the arithmetic mean of the original zero-load values ​​of the other three nodes and the effective observation time, respectively, b is the slope of the fitted line, and subscript j identifies the other three nodes. Using the two central nodes as the nodes under inspection allows the fitted value to be obtained by interpolation and controls the prediction uncertainty; the end nodes move to the central position after the four-node group rolls.

[0040] The state of each zero-load observation node is determined jointly by two adjacent four-node groups. End nodes within a single four-node group may also be disturbed; end disturbances are passed to the left-one residual of the central node via fitting. Automatically replacing the central node based solely on a single four-node group risks mislocating end disturbances to the central node. Each four-node group moves one node forward along the node chain, with the same node successively occupying the 3rd position in the previous group and the 2nd position in the next group, thus obtaining two centering checks. Only when the two checks are consistent is a verified or confirmed disturbed state established; if the conclusions are inconsistent, if both residuals in either group exceed the limit simultaneously, or if the fitted values ​​are inconsistent, the node remains in a pending verification state. This rule does not change the calculation method of the left-one fitting for four nodes and blocks the path where a single group of end disturbances directly triggers automatic correction.

[0041] The standard uncertainty of the predicted residual of the removed node is calculated as follows: Calculate, standardize, leave one residual as Calculation. In the formula, uR,k is the standard uncertainty of the predicted residual of the k-th examined node, uk is the standard uncertainty of that node itself, xk is a column vector with elements successively 1 and τk, X is the fitting design matrix for the other three nodes, with each row successively containing 1 and the corresponding effective observation time, W is a diagonal weight matrix, whose diagonal elements are the reciprocals of the squares of the standard uncertainties of the corresponding nodes, Zk is the unloaded original value, Zk with a cusp is the fitted value, and rk is the dimensionless standardized left-one residual. When the standard uncertainty of all four nodes is u, the above formula becomes: ,in In the formula, hk is the lever amount, τ with a horizontal line is the arithmetic mean of the effective observation times of the other 3 nodes, and τj is the effective observation time of the j-th node among the other 3 nodes.

[0042] Each inspected node is first identified as a candidate within a single four-node group. A normal candidate is formed when the absolute value of the standardized left-one residual is within 3; a disturbed candidate is formed when the absolute value of the residual exceeds 3 and the other inspected node in the same group does not exceed 3; a candidate awaiting verification is formed when both inspected nodes exceed 3. When the same node is a normal candidate in two adjacent four-node groups, it is recorded as a verified zero-load observation node. The fitted value obtained by removing the node from the previous four-node group is recorded as the first fitted value, and the fitted value obtained by removing the node from the subsequent four-node group is recorded as the second fitted value. If both adjacent groups list the same node as a disturbed candidate, and the absolute value of the difference between the first and second fitted values ​​is within the expanded uncertainty of that difference, it is confirmed as a disturbed node, and the fitted value with the smaller prediction standard uncertainty between the first and second fitted values ​​is selected as the confirmed fitted value. The standard uncertainty of this difference is obtained by propagating the first and second fitted values ​​with respect to the linear weights of the nodes involved in the fitting and the nodal covariance matrix. The expanded uncertainty is the product of this standard uncertainty and the coverage factor 2. The zero-point innovation is calculated as follows: Calculate, where δ is the zero-point innovation, Zk is the original zero-load value of the disturbed node, and Zk with a cusp is the confirmed fit value. After replacing the node with the confirmed fit value, proceed as follows: and Recalculate two adjacent feeding steps; Qk-1 and Qk marked with an asterisk represent the final actual feeding quantities of the preceding and following feeding steps, respectively, while those without an asterisk represent the preliminary actual feeding quantities. For the preceding feeding step, the disturbed node is used as the subsequent zero-load observation node. The numerator of its time ratio is the difference between the reading time after feeding and the reading time before feeding, and the denominator is the difference between the effective observation time of the disturbed node and the effective observation time of the preceding zero-load observation node. For the following feeding step, the disturbed node is used as the preceding zero-load observation node. The numerator of its time ratio is the difference between the reading time after feeding and the reading time before feeding, and the denominator is the difference between the effective observation time of the subsequent zero-load observation node and the effective observation time of the disturbed node. The recalculated zero-point drift compensation standard uncertainty is propagated according to the actual node uncertainty and the time ratio; when its value exceeds the allocated share, the relevant step is transferred to pending review and is not automatically released.

[0043] Figure 4 The diagram shows the second of two adjacent four-node groups. The raw zero-load values ​​for nodes N1 to N4 are 2 g, 14 g, 14 g, and 20 g, respectively, with effective observation times of 200 s, 1400 s, 2600 s, and 3800 s, respectively. After node N2 is removed, the fitted value obtained from nodes N1, N3, and N4 is 8 g. The standardized leave-one residuals for nodes N2 and N3 in this group are 9.47 and -2.70, respectively, with node N2 being the only out-of-limit node. The previous four-node group also includes an initial verification node V0 with an effective observation time of 0 s and a raw zero-load value of 1 g, and consists of V0, nodes N1, N2, and N3. The standardized leave-one residuals for nodes N1 and N2 in this group are -2.69 and 9.61, respectively, with node N2 again being the only out-of-limit node, and the fitted value for node N2 is also 8 g. Both adjacent groups point to node N2, and the two fitted values ​​are consistent. Node N2 is identified as the disturbed node, and the zero-point innovation amount is 14 grams minus 8 grams, which is 6 grams.

[0044] Figure 5This diagram illustrates how the zero-point innovation amount is used to make reverse corrections to two adjacent feeding steps at different ratios. The horizontal axis represents the previous or subsequent time ratio; the previous time ratio is the difference between the reading time after feeding and the reading time before feeding in the previous feeding step, divided by the difference between the effective observation time of the disturbed node and the effective observation time of the preceding zero-load observation node in that step; the subsequent time ratio is the difference between the reading time after feeding and the reading time before feeding in the subsequent feeding step, divided by the difference between the effective observation time of the following zero-load observation node in that step and the effective observation time of the disturbed node; the vertical axis represents the actual feeding amount correction. The absolute value of the slope of both straight lines passing through the origin is the zero-point innovation amount of 6 grams, with a negative sign for feeding steps where the disturbed node is a subsequent node and a positive sign for feeding steps where the disturbed node is a preceding node. The time ratio for feeding step S1 is 600 seconds divided by 1200 seconds, which is 0.50, with a correction of -3.0 grams; the time ratio for feeding step S2 is 300 seconds divided by 1200 seconds, which is 0.25, with a correction of +1.5 grams. The initial actual feeding amount for feeding step S1 is 4503 grams, which is corrected to 4500 grams. The reading for feeding step S2 before feeding is 5308 grams, and the reading after feeding is 810 grams; the zero-point drift rate when using the uncorrected node is 0 grams per second, the initial actual feeding amount is 4498 grams, which is corrected to 4499.5 grams.

[0045] Step 6 employs two states: preliminary verification and final verification. After obtaining the subsequent zero-load observation node for the feeding step, a preliminary actual feeding quantity is generated. If the preliminary judgment is satisfactory, the next feeding step is conditionally permitted to proceed with weighing. After both the preceding and subsequent zero-load observation nodes complete consistency confirmation of two adjacent four-node groups, the final actual feeding quantity is generated. If the final judgment is satisfactory, the next feeding step is permitted to proceed. If the final judgment is unsatisfactory, the feeding steps that have not yet been executed are immediately frozen, and the materials involved in the next feeding step that has already been executed based on the preliminary judgment are marked as isolated, awaiting reweighing or process evaluation. This state setting allows node inspection to obtain data from the next side, while restricting the continued release of results that have not completed final confirmation.

[0046] At the beginning of the node chain, two initial verification nodes are continuously collected before the first feeding step. If fewer than four nodes have been collected, a candidate disturbed node appears in a single test, or the conclusions of two adjacent four-node groups are inconsistent, the weighing platform is kept at zero load and additional verification nodes are collected. These additional verification nodes only participate in the fit test and are not used as preceding or following nodes for the feeding step. After the instrument step subsequence is completed, two terminal verification nodes are continuously collected. The initial verification nodes, additional verification nodes, and terminal verification nodes all use the same windowing, sampling, standard uncertainty calculation, and state confirmation methods as the zero-load observation nodes.

[0047] A node needs to be in the 3rd and 2nd positions respectively in two adjacent four-node groups to complete two centering checks. After adding two verification nodes at the beginning and end of the node chain, the first and last zero-load observation nodes adjacent to the feeding step can both obtain two centering checks. The outermost verification nodes at both ends are only used as fitting reference nodes, and their states are not directly used for zero-point drift calculation of the feeding step; when the outermost node shows abnormal signs, another verification node is added outwards, and the original outermost node then enters the centering check position.

[0048] Figure 8 The zero-load observation node chain and adjacent four-node groups are shown. Feeding step S1 is located between nodes N1 and N2, and the measurement period for feeding step S2 is located between nodes N2 and N3. Node N2 is reused from two adjacent feeding steps. The additional verification node collects data during the permitted waiting period between nodes N2 and N3, without replacing node N2 or serving as the boundary node for feeding step S2. Four-node groups 1 to 5 each consist of four consecutive nodes, shifting one node forward in each group: node N1 is checked in groups 1 and 2, node N2 is checked in groups 2 and 3, the additional verification node is checked in groups 3 and 4, and node N3 is checked in groups 4 and 5. The status of a node is updated after two consistent results. The final actual feeding quantity of a feeding step is formed after both its preceding and following zero-load observation nodes have completed status confirmation.

[0049] When using protective belts, the criteria for determining the qualification of cumulative material input are as follows: In the formula, QΣ represents the sum of the final actual input quantities obtained from m weighings in the same feeding step, Q0 represents the theoretical input quantity, UΣ represents the expanded uncertainty of the cumulative actual input quantity, and L represents the limit of error. The cumulative expanded uncertainty is calculated according to... The formula is as follows: where ui is the standard uncertainty of the final measured quantity in the i-th weighing, and cov(Qi,Qj) is the covariance between the final measured quantities in the i-th and j-th weighings due to sharing a zero-load observation node; when they do not share a node and can be considered independent after verification, this covariance is taken as 0. For weighing only once, the formula is: In the formula, U is the expanded uncertainty of the final actual quantity in a single measurement; ue is the quantitative composite component of the two readings, which is 0.4082 g in this embodiment; the repeatability standard uncertainty of the single reading within the target calibration interval is determined to be 0.5 g by performance verification data; the two readings are synthesized as uncorrelated quantities, resulting in a repeatability component ur of 0.7071 g; ud is the zero-point drift compensation component, which is the larger of the actual propagation calculation value and one-twelfth of the limit error, and is 0.375 g in this embodiment. The coverage factor is set to 2, resulting in U of 1.80 g. The protection zone requires that the deviation and its expanded uncertainty be within the overall limit error; the judgment risk near the boundary is lower than that without the protection zone.

[0050] In this embodiment, the actual final feeding amount of feeding step S1 is 4500 g, with a deviation of 0 g. After adding the expanded uncertainty of 1.80 g, the total value is within the tolerance limit of 4.5 g, so it is determined as qualified; the actual final feeding amount of feeding step S2 is 4499.5 g, with a deviation of 0.5 g. After adding the expanded uncertainty, the total value is 2.30 g, which is also determined as qualified. If the zero drift compensation in step 4 is not performed, the difference between the indicated value before feeding and the indicated value after feeding is 4497 g, and the sum of the 3 g deviation and the expanded uncertainty is 4.80 g, so feeding step S1 is determined as unqualified; if step 4 is performed but the disturbed node confirmation and correction in step 5 is not performed, the preliminary actual feeding amount is 4503 g, and the determination result is also unqualified. Step 4 eliminates local linear zero drift, and step 5 eliminates single-point disturbance confirmed by two-group consistency, and the two correspond to different error sources.

[0051] When determined as qualified, the actual final feeding amount, the original zero-load value of each zero-load observation node, the effective observation time, the zero drift rate, the zero increment, the node status, the code of the pre-weighing container used, and the expanded uncertainty adopted in the determination shall be written into the work order feeding record. When the sum of the deviation and the cumulative expanded uncertainty exceeds the tolerance limit and the cumulative actual feeding amount is less than the theoretical dosage, it is determined as under-feeding. The difference between the theoretical dosage and the cumulative actual feeding amount is taken as a new target amount to be fed, and the process returns to step 2 to re-perform joint matching; condition 2 and condition 3 use the new target amount to be fed, and condition 1 still uses the tolerance limit corresponding to the original theoretical dosage. After each supplementary feeding, the determination is performed based on the updated cumulative actual feeding amount, the original theoretical dosage and the re-synthesized cumulative expanded uncertainty, and the covariance caused by shared zero-load observation nodes is included in the synthesis. When the cumulative actual feeding amount is greater than the theoretical dosage and exceeds the allowable range, it is determined as over-feeding, and the subsequent feeding steps are frozen and the process is transferred to rework, batch isolation or manual disposal. When the absolute value of any original zero-load value exceeds the maximum allowable cumulative zero drift, the subsequent feeding steps of the appliance are also frozen and metrological verification is performed.

[0052] The admission coefficient in condition 1 can be adjusted according to the error budget of quantization, repeatability and zero drift compensation. A decrease in the coefficient will reduce the tolerance share occupied by the quantization term, and reduce the number of selectable division intervals at the same time; an increase in the coefficient will expand the selectable range, but this is on the premise that repeatability and zero drift compensation still have sufficient budgets. When on-site repeatability data meets the requirements, the admission coefficient can be set to one half; when the material is of high value and the proportion is sensitive, the admission coefficient can be set to one fifth. The adopted coefficient and its error budget are written into the process parameter version.

[0053] The boundary margin can be updated according to field conditions. When the weighing station environment is controlled, the upper limit of load fluctuation can be reduced based on new monitoring data; when long-term zero-point monitoring shows improved instrument stability, the maximum permissible cumulative zero drift can be adjusted accordingly. The boundary margin can also be set according to a fixed multiple of the target graduation interval verification value, but this multiple should be determined by measured fluctuation, zero drift, and indication stability data. The 20-gram limit in this embodiment only applies to the instrument and environment of this embodiment and should not be used as a direct setting value for other weighing stations.

[0054] The decorrelation interval can also be a pre-verified fixed value. When the digital filtering parameters of the weighing instrument and the field vibration conditions remain unchanged, the decorrelation interval can be calculated during periodic performance checks and fixed in the control terminal; each performance check should still be performed under the condition that three consecutive hysteresis points are within the 95% confidence band. When the weighing instrument can output unfiltered raw samples, the decorrelation interval can be the original sampling period; if the original sequence still has correlation, the standard uncertainty should be calculated using the number of effective sampling points.

[0055] The standardization residual limit can be set based on the long-term statistical distribution of zero-load observation nodes. Decreasing the limit increases the detection rate of small disturbances and the probability of normal nodes being transferred to pending review, while increasing the limit has the opposite effect. A limit of 2.5 can be used when the zero-load noise is close to normal and the standard uncertainty is stable, and 3.5 can be used when environmental fluctuations are large. After changing the limit, the same version of the limit should be used for two adjacent four-node groups to avoid using different criteria for two tests of the same node.

[0056] If two adjacent four-node groups cannot reach a consistent conclusion, additional verification nodes are collected, and the node groups are expanded to five or six. A weighted leave-one-out fitting is performed on all candidate nodes with nodes on both sides. Automatic correction is only applied if the same candidate node repeatedly becomes the only out-of-limit node in two adjacent windows and the fitted values ​​are consistent. If multiple nodes simultaneously exceed the limit, the residuals exhibit systematic curvature over time, or the fitted values ​​are inconsistent, the local linearity hypothesis is deemed invalid. The material feeding steps for the relevant time period remain pending verification, confirmed by weighing or manual measurement, and nodes are not replaced by a single fitted value.

[0057] Without protective strips, acceptance criteria can be simplified to the deviation being within the tolerance limit, and the shared risk rule and corresponding expanded uncertainty should be declared in the material input record. When the actual deviation is close to the acceptance boundary, the measurement uncertainty will significantly increase the probability of false acceptance or rejection; when the actual deviation is exactly at the boundary and the measurement error is approximately symmetrical, the probability of falling on either side of the boundary is close to one-half. Whether this rule is adopted is agreed upon in advance by the supplier and the buyer and will not be switched temporarily during operation.

[0058] This method can also be used for metering liquid materials. The pre-weighing container is a covered transfer tank, and the liquid is transferred out through a closed pipeline without branches. Source purging, valve closure, and the arrival signal from the process equipment together form a complete transfer confirmation. The amount of liquid remaining on the inner wall of the container is included in the indicated value after feeding; the actual amount fed is still the net amount transferred out of the transfer tank. If there is stagnation in the pipeline and the emptying confirmation is not completed, the actual amount fed is not directly used as the amount received by the process equipment.

[0059] Two or more multi-gradient weighing instruments can be configured at the same weighing station, and each feeding step is allocated according to the joint selection results. Each instrument maintains its own instrument step sub-sequence, fixed zero-point reference, and zero-load observation node chain, and nodes are not reused between instruments. If the zero-point drift rate or original zero-load value of an instrument exceeds the allowable range determined by the verification record, calibration certificate, or pre-work order performance verification, it is marked as pending verification, and subsequent feeding steps are reassigned to other instruments that meet the conditions; new instruments are individually zeroed before the first feeding step for that work order, and the initial verification node is collected.

[0060] This method can be accomplished by a division of labor between the weighing station control terminal and the workshop manufacturing execution system. The control terminal is responsible for acquiring station occupancy and complete transfer signals, managing the zero-load observation window, acquiring indicated values ​​and times, determining stability conditions, calculating standard uncertainty, and monitoring the fixed zero-point reference state. The manufacturing execution system is responsible for BOM expansion, process path analysis, generating tool and step sub-sequences, selecting weighing tools, graduation intervals, and pre-weighed containers, confirming the consistency of adjacent four-node groups, correcting the actual amount fed, synthesizing cumulative uncertainty, and granting step permission. Each feeding step saves the status as not started, initially qualified, pending verification, finally qualified, and finally unqualified, along with the corresponding status transition time. The control terminal uploads the node, two indicated values, reading time, and confirmation signal; the manufacturing execution system issues the target graduation interval, pre-weighed container code, lower limit of sampling points, and step permission; if the status versions are inconsistent, the permission is rejected.

[0061] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Equivalent substitutions, parameter adjustments, and partial improvements made by those skilled in the art based on the technical solutions of the present invention, such as changing the division of the graduation interval of multi-gradient weighing instruments, changing the order of determining the decorrelation interval and the lower limit of the number of sampling points, or replacing least squares linear fitting with other equivalent robust regression methods, should all be included within the scope of protection of the present invention.

Claims

1. A method for verifying the metering of material input in a work order based on the linkage between BOM and process path, executed at a weighing station equipped with a multi-gradient weighing instrument, wherein the multi-gradient weighing instrument has at least two graduation intervals, each graduation interval having a verification graduation value; for each multi-gradient weighing instrument participating in the work order, it is zeroed once before the first material input step for that work order, and a zero-point compensation is maintained during the period that the instrument undertakes the work order, so that the reading is taken from a fixed zero-point reference; characterized in that, Includes the following steps: Step 1: Expand the work order according to the hierarchical structure of the Bill of Materials (BOM) to obtain the theoretical usage and tolerance rate, and use the product of the tolerance rate and the theoretical usage as the limit of error; obtain the material feeding step sequence according to the process sequence of the process path, and arrange the material feeding steps assigned to the same multi-gradient weighing instrument in the order of their sequence to obtain the instrument step sub-sequence. Step 2: The target quantity to be used is initially taken as the theoretical quantity. The scale interval in which the verification scale value is within one-third of the limit error is selected as the target scale interval. A pre-weighed container is selected so that the tare weight and the sum of the tare weight and the target quantity to be used are both within the target scale interval. Step 3: Collect one set of readings during the zero-load period of the weighing platform before and after the feeding step. Take the arithmetic mean of each set of sampled values ​​as the original zero-load value and the arithmetic mean of the sampling time as the effective observation time to obtain the zero-load observation node. The same zero-load observation node is reused between two adjacent feeding steps in the tool step sub-sequence. The zero-point drift rate is obtained by dividing the difference between the original zero-load values ​​of the two zero-load observation nodes before and after the feeding step by the difference between the effective observation times of the two zero-load observation nodes. Step 4: Within the target graduation range, read the reading before feeding and the reading after feeding. The actual feeding amount is obtained by adding the difference between the reading before feeding and the reading after feeding to the product of the zero drift rate and the difference between the two reading times. Step 5: For each of the two middle zero-load observation nodes within a consecutive sequence of four zero-load observation nodes in the tooling step sub-sequence, perform a straight-line fitting to obtain the standardized one-left-off residual; confirm the consistency of the verification results of the same zero-load observation node for two four-node groups with three adjacent and overlapping zero-load observation nodes; when the absolute value of the standardized one-left-off residual of the same zero-load observation node in two adjacent four-node groups is within 3, it is recorded as a verified zero-load observation node; when both the preceding and following zero-load observation nodes of the feeding step become verified zero-load observation nodes, the actual feeding amount obtained in Step 4 is taken as the maximum. Final actual input; Remove two adjacent four-node groups from the same zero-load observation node and perform the aforementioned straight-line fitting, and record the fitted values ​​obtained at the effective observation time of the zero-load observation node as the first fitted value and the second fitted value, respectively; If the same zero-load observation node is unique in both adjacent four-node groups with a standardized left-one residual absolute value exceeding 3, and the absolute value of the difference between the first fitted value and the second fitted value is within the expanded uncertainty of that difference, then the zero-load observation node is identified as a disturbed node, where the expanded uncertainty is the standard deviation of the difference between the first fitted value and the second fitted value. Multiply the standard uncertainty by a coverage factor of 2, and take the fitted value with the smaller prediction standard uncertainty between the first and second fitted values ​​as the confirmed fitted value. The difference between the original zero-load value and the confirmed fitted value at the zero-load observation node is taken as the zero-point innovation. The feeding step with the disturbed node as the subsequent zero-load observation node is recorded as the previous feeding step, and the feeding step with the disturbed node as the preceding zero-load observation node is recorded as the next feeding step. The difference between the reading time of the indication after feeding and the reading time of the indication before feeding in the previous feeding step is divided by the effective observation time of the disturbed node and the reading time of the previous feeding step. The difference between the effective observation times of the zero-load observation node is used to obtain the previous time ratio; the difference between the reading time of the indication after feeding and the reading time of the indication before feeding in the next feeding step is divided by the difference between the effective observation time of the subsequent zero-load observation node and the effective observation time of the disturbed node in the next feeding step to obtain the next time ratio; the product of the zero-point innovation amount and the previous time ratio is subtracted from the actual feeding amount of the previous feeding step, and the product of the zero-point innovation amount and the next time ratio is added to the actual feeding amount of the next feeding step to obtain the final actual feeding amount of each of the two adjacent feeding steps; other cases are transferred to pending review; Step 6: The absolute value of the difference between the final actual amount and the theoretical amount is within the limit of error as the qualification condition, and the execution permission of subsequent feeding steps is controlled according to the judgment result.

2. The method according to claim 1, characterized in that, Each scale interval also has a lower load limit and an upper load limit; the method for selecting the target scale interval and the pre-weighed container in step 2 is as follows: set a boundary margin, which is the sum of the upper limit of load fluctuation, the upper limit of stable fluctuation of indication, and the maximum allowable cumulative zero drift of the multi-scale weighing instrument in the work order; select a combination of multi-scale weighing instrument, scale interval, and pre-weighed container that simultaneously meets the following three conditions, and take the scale interval in the combination as the target scale interval: Condition 1, the verification scale interval value is within one-third of the limit error; Condition 2, the lower limit of the actual tare weight of the pre-weighed container is greater than or equal to the sum of the lower load limit and the boundary margin of the scale interval; Condition 3, the sum of the upper limit of the actual tare weight of the pre-weighed container, the target quantity to be delivered, the upper limit of the pre-weighing overshoot, and the boundary margin is less than or equal to the upper load limit of the scale interval.

3. The method according to claim 2, characterized in that, When there are two or more combinations that simultaneously satisfy conditions 1, 2, and 3, the combination with the smallest verification scale value is selected first. When there are two or more combinations with the smallest verification scale value, the lower limit of the conservative load is the actual tare weight of the pre-weighed container minus the boundary margin. The upper limit of the conservative load is the sum of the upper limit of the actual tare weight, the target quantity to be added, the upper limit of the pre-weighing overshoot, and the boundary margin. The distance from the lower limit of the conservative load to the lower limit of the scale interval load and the distance from the upper limit of the scale interval load to the upper limit of the conservative load are divided by the width of the scale interval, and the smaller of the two quotients is taken as the normalized boundary margin. The combination with the largest normalized boundary margin is selected. When the combination that simultaneously satisfies conditions 1, 2, and 3 is an empty set, the information of insufficient measurement capacity is output, and the process is switched to the re-judgment of multi-scale weighing instruments with smaller verification scale values, multiple feedings, or manual weighing.

4. The method according to claim 1, characterized in that, The weighing station is equipped with a occupancy detection device to confirm zero load on the weighing platform, and the process equipment is equipped with a complete transfer confirmation device. The method for collecting one set of indications in step 3 is as follows: when the occupancy detection device outputs a zero load signal on the weighing platform and the complete transfer confirmation device outputs a complete transfer confirmation signal of the previous feeding step within the step interval between two adjacent feeding steps, the zero load observation window is opened, and the indications are collected according to the decorrelation interval. The decorrelation interval is the time interval corresponding to the first time that the autocorrelation function of the zero load value sequence of the multi-gradient weighing instrument falls into the 95% confidence band of the autocorrelation function and remains within the confidence band for three consecutive hysteresis points thereafter.

5. The method according to claim 4, characterized in that, The Type A component is obtained by dividing the standard deviation of the sampled values ​​within the zero-load observation window by the square root of the number of sampling points. The resolution component is obtained by dividing the minimum applicable verification scale value at zero load by the square root of 12. The root sum of the squares of the Type A component and the resolution component is taken as the standard uncertainty of the zero-load observation node. The readings are continuously collected until the following condition is met: the root sum of the squares of the standard uncertainties of the two zero-load observation nodes before and after the feeding step is multiplied by the ratio of the upper bound of the difference between the reading time registered in the feeding step and the lower bound of the difference between the effective observation time, and then multiplied by 3. The resulting product is within one-quarter of the tolerance limit. When a zero-load observation node is reused by two adjacent feeding steps, it is calculated according to the two adjacent feeding steps respectively, and the one with more required sampling points is taken as the condition for stopping the collection. Set the minimum number of sampling points and the maximum window duration. When the zero-load observation window duration reaches the maximum window duration, the feeding step will be transferred to the pending review stage.

6. The method according to claim 1, characterized in that, The method for reading the pre-feeding and post-feeding readings in step 4 is as follows: Place the pre-weighed container containing the material on the weighing platform. After the reading range is within 1 verification scale value for 2 consecutive seconds, read the pre-feeding reading. Transfer the material in the pre-weighed container to the process equipment. Return the pre-weighed container to the weighing platform. After the reading again meets the requirement that the reading range is within 1 verification scale value for 2 consecutive seconds, read the post-feeding reading. The material remaining in the pre-weighed container is included in the post-feeding reading. The actual feeding amount is the net feeding amount of the pre-weighed container. When the complete transfer confirmation device outputs a complete transfer confirmation signal, the actual feeding amount is the receiving amount of the process equipment.

7. The method according to claim 5, characterized in that, The standardized leave-one residual in step 5 is obtained as follows: The central zero-load observation node is removed. Using the reciprocal of the squared standard uncertainty of the remaining three zero-load observation nodes as weights, a weighted least squares straight line is fitted between the effective observation time and the original zero-load value. This yields the fitted value of the straight line at the effective observation time of the removed node. The difference between the original zero-load value and the fitted value of the removed node is divided by the standard uncertainty of the predicted residual, which is the sum of the standard uncertainty of the removed node and the predicted standard uncertainty of the fitted value, calculated using the root of the square. This yields the standardized leave-one residual. When the standard uncertainties of the four nodes are equal, the standard uncertainty of the predicted residual is equal to the product of this standard uncertainty and the predicted residual coefficient. The predicted residual coefficient is the square root of the sum of 1, one-third, and the deviation quotient. The deviation quotient is the sum of the effective observation time of the removed node and the original zero-load value. The square of the difference between the mean values ​​of the effective observation times of a node is divided by the sum of the squares of the deviations of the effective observation times of the other three nodes from that mean value. Two adjacent four-node groups are each moved out of the same node to perform the straight-line fitting, and the fitted values ​​obtained at the effective observation time of that node are recorded as the first fitted value and the second fitted value, respectively. When the absolute value of the standardized left-one residual of the same node in two adjacent four-node groups is within 3, it is recorded as a verified zero-load observation node. When the same node is the only node in two adjacent four-node groups that exceeds 3, and the absolute value of the difference between the first fitted value and the second fitted value is within the expanded uncertainty of that difference, it is identified as a disturbed node. The expanded uncertainty is the standard uncertainty of the difference between the first fitted value and the second fitted value multiplied by a coverage factor of 2. In other cases, the process is transferred to a pending review and additional verification nodes are collected.

8. The method according to claim 7, characterized in that, After obtaining the post-zero load observation node of the feeding step, the preliminary actual feeding quantity of the feeding step is obtained. When the preliminary actual feeding quantity is used to replace the final actual feeding quantity to perform the qualification judgment condition and is judged to be qualified, the next feeding step is allowed to perform weighing. When both the pre-zero load observation node and the post-zero load observation node of the feeding step become verified zero load observation nodes or are corrected after being confirmed to be disturbed, the final actual feeding quantity of the feeding step is obtained. When the final actual feeding quantity is used to perform the qualification judgment condition and is judged to be qualified, the next feeding step is allowed to perform weighing. When the final judgment is unqualified, the feeding steps that have not yet been executed are frozen, and the materials involved in the next feeding step that has been executed according to the preliminary judgment are transferred to batch isolation and manual verification. Two initial verification nodes are continuously collected at the beginning of the tooling step subsequence. When the number of zero load observation nodes obtained is less than four or the test results are pending verification, the weighing platform is kept at zero load and additional verification nodes are collected. Two terminal verification nodes are continuously collected at the end of the tooling step subsequence.

9. The method according to claim 5, characterized in that, The qualification criterion is replaced by the absolute value of the difference between the final actual quantity and the theoretical quantity, plus the expanded uncertainty of the final actual quantity being within the tolerance limit. The expanded uncertainty of the final actual quantity is obtained by multiplying the fractional quantification component of the pre-feeding and post-feeding values, the repeatability component of the target scale interval, and the zero-point drift compensation component by the root of the square and the inclusion factor 2. The fractional quantification component is the square root of the calibration scale value divided by 12 and then multiplied by the square root of 2. The zero-point drift compensation component is the larger of the standard uncertainty calculated based on the node standard uncertainty, the standard uncertainty predicted by the fitted value, and the ratio of the actual time to the standard uncertainty and the tolerance of one-twelfth.

10. The method according to claim 9, characterized in that, The sum of the final actual input quantities of each weighing in the same feeding step is recorded as the cumulative actual input quantity. The absolute value of the difference between the cumulative actual input quantity and the theoretical input quantity is recorded as the deviation quantity. The expanded uncertainty of the cumulative actual input quantity is obtained by multiplying the standard uncertainty of each final actual input quantity and the covariance generated by the shared zero-load observation node by the coverage factor 2. In step 6, the condition for acceptance is that the sum of the expanded uncertainty of the deviation quantity and the cumulative actual input quantity is within the tolerance limit. When the qualification criteria are met, the process is deemed qualified. The final actual feed quantity, the original zero-load values ​​and effective observation times of each zero-load observation node, the zero-point drift rate, the zero-point innovation quantity, and the pre-weighed container code are written back, and subsequent feeding steps are allowed to perform weighing. When the qualification criteria are not met and the cumulative actual feed quantity is less than the theoretical quantity, the process is deemed underfeeding. The difference between the theoretical quantity and the cumulative actual feed quantity is taken as the new target quantity to be fed, and the process returns to step 2. Steps 3 to 5 are repeated within the supplementary feeding time limit set in the process path, and the cumulative actual feed quantity and its expanded uncertainty are recalculated after each supplementary feeding. When the qualification criteria are not met and the cumulative actual feed quantity is greater than the theoretical quantity, the process is deemed overfeeding. Subsequent feeding steps are frozen and the process is transferred to rework, batch isolation, or manual disposal. When the qualification criteria are not met and the cumulative actual feed quantity is equal to the theoretical quantity, the process is transferred to the insufficient measurement capacity review. When the subsequent feeding step is changed to another multi-gradient weighing instrument, the reuse of the zero-load observation node of the original multi-gradient weighing instrument is terminated. The other multi-gradient weighing instrument is set to zero separately before its first feeding step and an independent starting verification node and zero-load observation node chain are established. When the absolute value of any zero-load original value exceeds the maximum allowable cumulative zero drift, the subsequent feeding steps of the instrument are frozen and transferred to metrological verification.

Images