Disclosure of Invention
Aiming at the defects in the prior art, the invention provides a processing method and a processing system for machining precision prediction data of a gear hobbing machine tool.
A processing method for prediction data of machining precision of a gear hobbing machine tool comprises the steps of obtaining a target gear, obtaining a plurality of historical machining parameter sets corresponding to the target gear, obtaining historical precision data of the target gear after machining is completed according to the historical machining parameter sets, obtaining error proportions corresponding to the historical machining parameter sets according to the historical precision data corresponding to the historical machining parameter sets, wherein each historical machining parameter set comprises a plurality of subdivision parameters, sequentially arranging the historical machining parameter sets according to the sequence from small value to large value of each subdivision parameter in the historical machining parameter sets to form a parameter set sequence corresponding to each subdivision parameter, obtaining a current machining parameter set corresponding to the target gear and comprising a plurality of subdivision parameters, obtaining the position of the value of the ith subdivision parameter in the parameter set sequence corresponding to the current machining parameter set, obtaining the historical machining parameter set corresponding to the adjacent two value of the position and serving as two comparison parameter sets of the ith subdivision parameter, obtaining the error proportions corresponding to the current machining parameter set according to the error proportions corresponding to the subdivision parameter set in the current subdivision parameter set, and obtaining the prediction error proportions corresponding to the subdivision parameter set in the current machining parameter set according to the subdivision error proportions corresponding to the subdivision parameter set corresponding to the ith subdivision parameter set.
Optionally, obtaining the error ratio corresponding to each historical processing parameter set according to the historical accuracy data corresponding to each historical processing parameter set includes obtaining a gear design size of the target gear, obtaining the historical processing size corresponding to each historical processing parameter set according to the historical accuracy data corresponding to each historical processing parameter set, and obtaining the error ratio according to the gear design size and the historical processing size.
Optionally, the error ratio obtained according to the gear design size and the historical machining size is expressed as: Wherein, the method comprises the steps of,For the error ratio corresponding to the j-th set of historical processing parameters,The gear is dimensioned so that, for a gear,For the history process size corresponding to the j-th set of history process parameters,Is the maximum error amount.
Optionally, obtaining the subdivision error ratio corresponding to the ith subdivision parameter in the current processing parameter set according to the error ratio corresponding to the two comparison parameter sets of the ith subdivision parameter comprises obtaining the difference value of the error ratio corresponding to the two comparison parameter sets of the ith subdivision parameter and serving as a total error variable, and obtaining the error ratio corresponding to the smaller value of the two comparison parameter sets of the ith subdivision parameter, obtaining the difference value of the ith subdivision parameter in the two comparison parameter sets and serving as a comparison parameter variable, obtaining the difference value of the ith subdivision parameter in the current processing parameter set and the value of the ith subdivision parameter in the smaller value comparison parameter set and serving as a current parameter variable, and obtaining the subdivision error ratio corresponding to the ith subdivision parameter according to the current parameter variable, the comparison parameter variable, the total error variable and the error ratio corresponding to the smaller value of the two comparison parameter sets.
Optionally, obtaining the subdivision error ratio corresponding to the ith subdivision parameter in the current processing parameter set according to the error ratio corresponding to the two comparison parameter sets of the ith subdivision parameter is expressed as: Wherein, the method comprises the steps of,For the subdivision error ratio corresponding to the ith subdivision parameter,Is the error ratio of a reference parameter set,For the error ratio corresponding to the other control parameter set,The larger value of the i-th subdivision parameter of the two comparison parameter sets,The smaller of the i-th subdivision parameters of the two comparison parameter sets is taken,The value of the ith subdivision parameter of the current processing parameter set is obtained.
Optionally, the method further comprises the steps of obtaining the gear design size of the target gear and obtaining prediction accuracy data according to the prediction error proportion.
Optionally, obtaining the prediction error proportion corresponding to the current processing parameter set according to the subdivision error proportion corresponding to each subdivision parameter in the current processing parameter set is expressed as:, Wherein, the method comprises the steps of,For the prediction error ratio corresponding to the current set of processing parameters,For the number of categories of subdivision parameters in the current set of processing parameters,The contribution weight for the ith subdivision parameter,And the subdivision error ratio corresponding to the ith subdivision parameter.
The system comprises a data acquisition module, a sequence generation module, a data processing module and a prediction module, wherein the data acquisition module is used for acquiring a target gear, acquiring a plurality of historical processing parameter groups corresponding to the target gear, acquiring historical precision data after gear cutting processing according to each historical processing parameter group, acquiring error proportions corresponding to each historical processing parameter group according to the historical precision data corresponding to each historical processing parameter group, wherein each historical processing parameter group comprises a plurality of subdivision parameters, the sequence generation module is used for sequentially arranging a plurality of historical processing parameter groups according to the order of the values of each subdivision parameter in the historical processing parameter groups from small to large to form a parameter group sequence corresponding to each subdivision parameter, the data processing module is used for acquiring the current processing parameter group corresponding to the target gear and comprising a plurality of subdivision parameters, acquiring the position of the value of the i subdivision parameter in the parameter group corresponding to the current processing parameter group, acquiring two adjacent historical processing parameter groups corresponding to the value of the position and serving as two reference parameter groups corresponding to the i subdivision parameter, and acquiring the current subdivision parameter groups corresponding to the current subdivision parameter group corresponding to the current subdivision parameter error proportion in the prediction module according to the error proportions of the current processing parameter group corresponding to the i subdivision parameter group.
Optionally, the data acquisition module is further configured to acquire a gear design size of the target gear, acquire historical processing sizes corresponding to the historical processing parameter sets according to the historical precision data corresponding to the historical processing parameter sets, and acquire an error ratio according to the gear design size and the historical processing sizes.
Optionally, the data processing module is further configured to obtain a difference value of error ratios corresponding to the two reference parameter sets of the ith subdivision parameter and serve as a total error variable, and obtain an error ratio corresponding to a smaller value of the two reference parameter sets of the ith subdivision parameter, obtain a difference value of a value of the ith subdivision parameter in the two reference parameter sets and serve as a reference parameter variable, obtain a difference value of a value of the ith subdivision parameter in the current processing parameter set and a value of the ith subdivision parameter in the smaller value reference parameter set and serve as a current parameter variable, and obtain a subdivision error ratio corresponding to the ith subdivision parameter according to the current parameter variable, the reference parameter variable, the total error variable and the error ratio corresponding to the smaller value of the two reference parameter sets corresponding to the ith subdivision parameter.
The beneficial effects of the invention are as follows:
In the whole processing method of the gear hobbing machine tool processing precision prediction data, firstly, an error proportion mechanism converts absolute dimensional errors into relative tolerance occupancy, data incompatibilities caused by difference of tolerance bands are eliminated, the same-dimensional quantitative evaluation of history and current data is achieved, misjudgment caused by confusion of original error dimensions in the traditional method is avoided, secondly, parameter serialization recombination is used for independently constructing ascending historical tracks for each processing parameter, implicit trend rules (such as error nonlinear transition caused by increment of material hardness) are converted into computable topological relations, the defect that a statistical average method ignores the parameter gradual change process is overcome, further, only two adjacent historical parameter sets are selected to serve as reference points based on accurate positioning of current parameters in the sequence, subdivision error proportion independently influenced by single parameters is calculated through interpolation, precision mutation characteristics (such as error steep effect of feed quantity near a critical value) of a microscopic level capture parameter boundary area are enabled to be provided, finally, dynamic weight aggregation is used for differentiating contribution of core parameters and secondary parameters to error differentiation according to process knowledge, and the problem that the sensitivity of the secondary parameters is influenced by the secondary parameters is solved, and the problem of the secondary parameters is influenced on the system is solved.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is apparent that the described embodiments are some embodiments of the present invention, but not all embodiments of the present invention. The components of the embodiments of the present invention generally described and illustrated in the figures herein may be arranged and designed in a wide variety of different configurations.
Thus, the following detailed description of the embodiments of the invention, as presented in the figures, is not intended to limit the scope of the invention, as claimed, but is merely representative of selected embodiments of the invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
It should be noted that like reference numerals and letters refer to like items in the following figures, and thus once an item is defined in one figure, no further definition or explanation thereof is necessary in the following figures. Furthermore, the terms "first," "second," and the like, are used merely to distinguish between descriptions and should not be construed as indicating or implying relative importance.
As shown in fig. 1, there is provided a method for processing machining precision prediction data of a gear hobbing machine, comprising:
s1, acquiring a target gear, acquiring a plurality of historical processing parameter sets corresponding to the target gear, acquiring historical precision data after finishing cutting processing of the target gear according to each historical processing parameter set, and acquiring error proportions corresponding to each historical processing parameter set according to the historical precision data corresponding to each historical processing parameter set, wherein each historical processing parameter set comprises a plurality of subdivision parameters;
S2, sequentially arranging a plurality of historical processing parameter sets according to the order of values of subdivision parameters in the historical processing parameter sets from small to large, and forming a parameter set sequence corresponding to each subdivision parameter;
S3, acquiring a current processing parameter set corresponding to the target gear and comprising a plurality of subdivision parameters, acquiring a position where a value of an ith subdivision parameter of the current processing parameter set in a parameter set sequence corresponding to the ith subdivision parameter is located, acquiring two historical processing parameter sets corresponding to adjacent values of the position and serving as two comparison parameter sets of the ith subdivision parameter, and acquiring subdivision error proportions corresponding to the ith subdivision parameter in the current processing parameter set according to error proportions corresponding to the two comparison parameter sets of the ith subdivision parameter;
s4, obtaining a prediction error proportion corresponding to the current processing parameter set according to the subdivision error proportion corresponding to each subdivision parameter in the current processing parameter set.
In the present embodiment, in step S1, step S1 is a data base construction stage of the entire prediction method, and the core objective is to sort and transform the history processing information related to the target gear. Specifically, it first needs to ascertain the type of "target gear" and collect a "set of historical processing parameters" representing different combinations of processing conditions that the type of gear accumulated over past production. Each of these parameter sets contains a number of key, dynamically controllable process settings (also known as subdivision parameters, such as the speed of feed, the speed of spindle cutting, the measured hardness value of the gear blank material, etc.). Meanwhile, for each time of the actual finished gear subjected to cutting processing by applying the specific parameter combination, key 'historical precision data' of the actual finished gear needs to be acquired, wherein the key 'historical precision data' refers to the absolute value of a dimension error existing between the actual dimension (such as tooth thickness and tooth shape) of the processed gear and the standard dimension specified by the design drawing. To more scientifically and comparably characterize the significance and extent of the effects of these errors, S1 introduces an error ratio, which is calculated for each set of historical process parameters, by dividing the absolute value of the actual dimensional deviation produced in the historical process by the maximum dimensional deviation tolerance (i.e., the tolerance band width) allowed by the design of the target gear type to yield a normalized, dimensionless ratio. This ratio clearly expresses "how much of the allowable deviation range is occupied by the actual deviation of the process", the closer the ratio value is to 1 or more than 1, the closer the error is to or has exceeded the tolerance limit.
Further, by way of example, assuming that the target gear is a certain type of automotive transmission gear, the tolerance requirement for the critical tooth thickness dimension is + -0.01 mm (i.e., maximum allowable error Δs=0.01 mm). S1, retrieving the production record of the past 20 batches of the model gear from a production database. Each record is a set of historical processing parameters (e.g., record: feed = 0.15mm/rev, cutting speed = 90m/min, material hardness = 220 HBW.) and correlates the actual tooth thickness measured for the batch of gear samples to a standard tooth thickness (e.g., deviation of +0.005mm, absolute error of 0.005 mm). Then, for each of the set of history parameters, the corresponding error ratio is calculated by dividing the "absolute value of history size error" (0.005 mm) by the "maximum allowable error" (0.01 mm) to obtain a ratio (0.5). This 0.5 means that machining with a specific combination of parameters is performed, which produces a dimensional error that occupies 50% of the allowable error range, yet is within the safety margin. By S1, all historical process data (different parameter combinations and their results) are systematically converted into a "scaled" version of this characterization error severity, providing for the subsequent step of using these historical experiences to predict the risk of accuracy under the new parameter combinations.
In step S2, the core purpose of step S2 is to reorganize the historical processing data according to the parameter change rule, and construct a structured sequence that can reflect the continuous change trend of the parameter. In the step, all the historical processing parameter groups which are sorted in the step S1 are arranged in a strictly ascending order according to the values of the historical processing parameter groups aiming at each independent subdivision parameter (such as feed, cutting speed, material hardness and the like), so as to generate a parameter group sequence exclusive for the parameter. The specific operation is that a subdivision parameter (such as cutting speed) is selected, all historical parameter groups are traversed, the values of the parameter are extracted and sorted from small to large, and meanwhile, the complete parameter combination corresponding to each parameter group and the associated error proportion are reserved. The method forms a plurality of parallel longitudinal sequences (such as arrangement of parameter groups in a cutting speed sequence from low to high according to speed values, and arrangement of parameter groups in a feeding quantity sequence from small to large according to feeding quantity), and the relative positions of the parameter groups in each sequence directly represent the progressive relation of the subdivision parameters in values.
Further, taking material hardness parameters as an example, it is assumed that 5 sets of parameters exist in the history repository, and the hardness values are respectively 200HBW, 210HBW, 220HBW, 230HBW, and 240HBW. S2 generates an independent sequence for the hardness parameters, and the 5 sets of parameters are arranged in ascending order of hardness value, namely, a parameter set of hardness=200 HBW, a parameter set of hardness=210 HBW, a parameter set of hardness=220 HBW, a parameter set of hardness=230 HBW, and a parameter set of hardness=240 HBW. Each row in the sequence still contains the original set of values of other parameters such as feed rate, cutting speed, etc. and their error ratios in its entirety. By constructing such a sequence for each subdivision parameter, the gradual change rule of the parameter values (e.g. the hardness increases gradually from 200HBW to 240 HBW) is highlighted, and the relative positional relationship between the parameters is quantitatively recorded. This lays a structural foundation for trend interpolation based on parameter proximity (e.g., the effect of locating 215HBW between hardness 210HBW and 220 HBW) in subsequent steps, thereby overcoming the defect of the traditional method that ignores the implicit rule of parameter ordering.
In S3, the objective is to calculate the independent influence of each subdivision parameter in the current processing parameter combination on the accuracy by using the local relation of the historical parameter sequence. This step first obtains a current set of processing parameters comprising all subdivision parameters to be optimized (e.g. feed, cutting speed, etc.), and then for each subdivision parameter, performs the operation of precisely locating the value position where the current parameter value is located (e.g. the current speed value is located between some two adjacent historical speed values in the sequence) in the parameter-specific ascending sequence established in S2 (e.g. cutting speed sequence). Next, two adjacent history parameter sets before and after the position are extracted as comparison parameter sets (for example, a history set a with a slightly lower speed and a history set B with a slightly higher speed; the current speed value may overlap with the history speed value, when overlapping, the overlapped history set is extracted first as one comparison parameter set, and then an adjacent history value with a higher speed or a history set with a lower speed is extracted as another comparison parameter set; it should be noted that two adjacent (overlapping also calculating) values must exist at the value position where the current parameter value is located to obtain two comparison parameter sets), and these two comparison parameter sets are adjacent to the current parameter only in the parameter value, and other parameters may be completely different. Finally, based on the error proportion historical values (namely the normalized precision deviation duty ratio calculated by S1) of the two comparison groups, the precision influence value, namely the subdivision error proportion, possibly generated when the subdivision parameter is considered in isolation by the current parameter value is deduced through nonlinear interpolation calculation reflecting the parameter proximity relation.
Further, taking the feed rate parameter as an example, assuming that the history feed rate sequence contains three sets of parameters in ascending order (feed rate=0.10 mm/rev, 0.15mm/rev, 0.20mm/rev corresponding set), the current machining setting feed rate is 0.18mm/rev. S3 first positions a position of 0.18 mm/rev-between 0.15mm/rev and 0.20mm/rev in the sequence, thereby selecting the two sets as control parameter sets. Assume that the historical error ratio for the 0.15mm/rev group is 0.4 (i.e., the error is 40% tolerance) and the 0.20mm/rev group is 0.8 (80% tolerance). At this time, the system automatically calculates the subdivision error ratio under the independent action of the current feeding amount according to the distance of 0.18mm/rev relative to two control values (closer to 0.20 mm/rev) and the difference of the historical error ratio (0.8-0.4=0.4), wherein the subdivision error ratio is higher than 0.4 but lower than 0.8. This calculation defaults to a core logic that is a single parameter change induced offset in accuracy, with a continuous transition between adjacent parameter intervals. Through the operation, each subdivision parameter obtains independent influence estimation value which is separated from other parameter interference based on the sequence of the subdivision parameter, and a foundation is laid for subsequent comprehensive weighting.
In S4, the core task is to comprehensively consider the differential process influence of each processing parameter on the precision, and generate an overall precision risk prediction value of the current parameter combination. This step derives the prediction error ratio for the entire parameter combination by introducing a process knowledge weighted aggregation mechanism based on all the subdivision error ratios obtained in S3 (i.e. the contribution ratio of each parameter to the dimensional deviation under "isolated ideal conditions"). Specifically, the system distributes contribution weights to different subdivision parameters in advance according to engineering experience or data analysis (for example, the cutting speed is given a weight of 0.6 due to the sensitivity to thermal deformation, the feeding amount is given a weight of 0.3 due to the small influence on the force deformation, the material hardness is given a weight of 0.1), and the total weight is constant to 1. A weighted sum calculation is then performed by multiplying the proportion of the sub-division error for each parameter by its weight and then accumulating the weighted results for all parameters. The final output value prediction error proportion represents the proportion of the expected size error in the tolerance range under the current parameter combination, and the machining out-of-tolerance risk is directly reflected.
Further, taking the processing of the automobile gearbox gear as an example, the subdivision error proportion of the current parameter calculated in the step S3 is set to be that the cutting speed corresponds to 0.7, the feeding amount corresponds to 0.4 and the material hardness corresponds to 0.5. According to the technological characteristics of the model gear, the speed weight is 60% (main factor), the feeding weight is 30% (secondary factor) and the hardness weight is 10% (weak factor) are set. S4 is calculated as (cutting speed influence 0.7×60% =0.42) + (feed amount influence 0.4×30% =0.12) + (hardness influence 0.5×10% =0.05) =prediction error ratio 0.59. This value indicates that at the present combination, the machining dimensional deviation is expected to occupy 59% of the tolerance, but not out of tolerance but approaching the safety boundary (> 0.6 for early warning). The mechanism breaks through the defect that the traditional method ignores the difference of the influence of parameters, for example, when the cutting speed is increased from 90m/min to 95m/min, the weight amplification effect can lead the predicted value to reflect the actual risk jump more sharply, and the predicted value is not diluted by the secondary parameters such as the feeding amount and the like.
In summary, in the whole processing method of the gear hobbing machine tool processing precision prediction data, firstly, an error proportion mechanism converts absolute dimensional errors into relative tolerance occupancy, data incompatibilities caused by tolerance zone differences are eliminated, the same-dimensional quantitative evaluation of history and current data is realized, misjudgment caused by original error dimension confusion in the traditional method is avoided, secondly, parameter serialization recombination is independently constructed for each processing parameter, ascending historical tracks are independently constructed, an implicit trend rule (such as error nonlinear transition caused by material hardness increment) is converted into a computable topological relation, the defect that a parameter gradual change process is omitted by a statistical average method is overcome, further, based on accurate positioning of the current parameter in a sequence, only two nearest historical parameter sets are selected to serve as reference points, subdivision error proportion independently influenced by single parameter is deduced through interpolation calculation, precision abrupt change characteristics (such as error synergy effect of feeding quantity near a critical value) of a microscopic level capture parameter boundary area are achieved, finally, dynamic weight aggregation distinguishes core parameters and parameters according to process knowledge to differences, and the problem that sensitivity of the nonlinear response is influenced by a secondary parameter is solved in a prediction system.
As shown in fig. 2, in one embodiment, in S1, obtaining the error ratio corresponding to each historical processing parameter set according to the historical accuracy data corresponding to each historical processing parameter set includes:
S11, acquiring the gear design size of the target gear, and acquiring the historical processing size corresponding to each historical processing parameter set according to the historical accuracy data corresponding to each historical processing parameter set;
S12, obtaining an error proportion according to the design size and the historical processing size of the gear.
In the present embodiment, it should be noted that in S11, an accurate mapping relationship between the historical processing data and the design standard is constructed, and an original calculation basis is provided for error quantization. The method comprises the steps of firstly determining theoretical design dimensions of a target gear (namely standard values of key dimensions such as tooth thickness, tooth height and the like specified by a drawing), and then extracting corresponding actual machining dimensions of each historical machining parameter set according to records of the historical machining parameter sets, wherein the data are derived from a gear finished product physical detection report (for example, actual tooth thickness values measured by a three-coordinate measuring machine) after the actual machining of the parameter sets. This step essentially establishes a physical correspondence between "parameter combination input" and "process result output" by precisely correlating abstract process parameters (e.g., feed 0.15 mm/rev) with apparent dimensional results (e.g., actual measured tooth thickness deviation +0.005mm from standard). For example, a certain batch of gearbox gears processed by adopting a specific cutting speed and a specific feeding amount is inspected to find that the actual tooth thickness is a certain value, and the absolute difference between the value and the design size is the core output of the step, so that irreplaceable basic data is provided for subsequent error standardization.
In S12, the absolute size error is converted into a risk index which can be interpreted by a process, so that unified quantification of error influence of cross-parameter combination is realized. The step calculates the absolute dimensional deviation amount (|actual value-design value|) of each of the history parameter groups using the design size and actual machining size obtained in S11, and divides it by the maximum allowable tolerance of the gear type (for example, ±0.01mm means the tolerance zone width of 0.02 mm), to generate a dimensionless error proportional value. The ratio value breaks through the limitation of the traditional method, on one hand, the ratio value uniformly converts dimension errors (such as 0.005mm and 0.008 mm) of different orders into tolerance range occupancy (such as 50% VS80%), dimensional interference is eliminated, and on the other hand, the ratio value directly reflects the distance between a processing result and a scrapped boundary (the ratio is more than or equal to 100%, namely, out-of-tolerance), so that historical data has a risk early warning function. For example, when two sets of processing records with the same material hardness parameters are processed, one set of error proportion is 0.3 (safety), and the other set is 0.9 (imminent exceeding), even if the absolute error amounts are similar (such as 0.006mmvs0.018 mm), the index can clearly represent the substantial risk difference under the current tolerance system, and visual basis is provided for subsequent parameter optimization.
In one embodiment, the ratio of the error obtained in S12 from the gear design size and the historical machining size is expressed as:
Wherein, the method comprises the steps of,
For the error ratio corresponding to the j-th set of historical processing parameters,The gear is dimensioned so that, for a gear,For the history process size corresponding to the j-th set of history process parameters,Is the maximum error amount.
In the present embodiment, the present invention is also directed to a method for manufacturing a semiconductor device,And (5) calculating absolute deviation, and eliminating the direction influence of positive and negative deviation. When the conventional method directly uses the original error, the positive deviation (oversized) and the negative deviation (undersized) cancel each other (e.g., when averaged), resulting in the true error being underestimated. If the tooth thickness is 0.008mm (error +0.008 mm) after processing of a certain group of parameters, the tooth thickness is 0.006mm (error 0.006 mm) of another group, and the original error average value is 0.001mm, which seriously covers the actual risk. And absolute value processing ensures that all offsets are accumulated positively, avoiding directional interference.
Further, the method comprises the steps of,For normalization processing, physical dimension errors are converted into dimensionless risk ratios, and data of gears with different tolerance requirements are transversely compared. To solve dimension confusion, for example, 0.005mm error caused by feed change, in25% In the mm gear and inThe gear with the mm accounts for 50%, risk levels of the gear and the gear are different, quantitative risk levels are achieved, and out-of-tolerance early warning is triggered when the proportion value is more than or equal to 100%, so that unified decision criteria of the working conditions are achieved. At the same time, denominator selection(Tolerance zone width),Essentially the limit deviation range allowed by the design. For example, tolerance marks of + -0.01 mm, the actual allowable total deviation interval is 0.01mm (i.e., from 5mm to 4.990mm or 5.010 mm). When the actual size exceedsIn the time-course of which the first and second contact surfaces,Directly determine the out of tolerance whenIn the time-course of which the first and second contact surfaces,Safety.
As shown in fig. 3, in one embodiment, in S3, obtaining the subdivision error ratio corresponding to the i-th subdivision parameter in the current processing parameter set according to the error ratios corresponding to the two comparison parameter sets of the i-th subdivision parameter includes:
S31, obtaining a difference value of error proportions corresponding to two comparison parameter groups of the ith subdivision parameter and taking the difference value as a total error variable, and obtaining an error proportion corresponding to a smaller value of the two comparison parameter groups of the ith subdivision parameter;
S32, obtaining the difference value of the ith subdivision parameter in the two comparison parameter sets and taking the difference value as a comparison parameter variable, and obtaining the difference value of the ith subdivision parameter in the current processing parameter set and the value of the ith subdivision parameter in the comparison parameter set with a smaller value and taking the difference value as a current parameter variable;
S33, obtaining the subdivision error proportion corresponding to the ith subdivision parameter according to the current parameter variable, the comparison parameter variable, the total error variable and the error proportion corresponding to the smaller values of the two comparison parameter sets corresponding to the ith subdivision parameter.
In the present embodiment, in S31, the precision fluctuation range of the adjacent history parameter group is quantized, and a reference scale is provided for interpolation calculation. The method comprises the steps of firstly selecting two comparison parameter groups (adjacent history groups positioned by S3) of an ith subdivision parameter, calculating absolute difference of error proportions of the two comparison parameter groups as a total error variable, wherein the absolute difference represents the maximum fluctuation amplitude of precision in a parameter interval, and simultaneously identifying larger error proportion values of the two comparison groups and reflecting the upper limit of precision risk of a parameter value area. For example, in a cutting speed sequence, if the error ratio of the control group is 0.3 to 0.6, the total error variable |0.6-0.3|=0.3 reveals that the accuracy fluctuation range due to the speed change is 30% tolerance, and a larger value of 0.6 indicates that the highest risk of the interval reaches the tolerance of 60%. The step quantifies the extremely poor characteristic of the historical trend and provides a dynamic adjustment basis for subsequent interpolation.
In S32, an interpolated coordinate frame is constructed by mathematical mapping of the parameter relative positional relationship. The step of performing difference value calculation twice comprises the steps of comparing parameter variables, namely calculating the value interval of the ith parameter of two comparison parameter sets (for example, the difference value between the cutting speed of 90m/min and the cutting speed of 100m/min is 10 m/min), taking the value interval as a reference interval length, and calculating the position offset of the current parameter value relative to any comparison set (for example, the difference value between the current speed of 95m/min and the low speed of 90m/min is 5 m/min) as an interpolation coordinate point. For example, the current feed amount is 0.18mm/rev between the history group (0.15 mm/rev and 0.20 mm/rev), the reference parameter variable is 0.05mm/rev (section length), and the current parameter variable is 0.03mm/rev (position from the lower limit group). The operation converts the abstract parameter relation into a proportional spatial scale (such as 0.03/0.05=60% position), so that nonlinear influence can be approximated through the linear relation, and the problem of parameter boundary prediction distortion is directly solved.
In S33, based on the intermediate variables of the first three steps, an estimate of the accuracy under the independent influence of the current parameters is derived. The step takes the maximum error proportion of the comparison group as a reference, and the total error variable is proportionally reduced according to the position (the offset proportion of S32) of the current parameter in the value interval, so as to finally obtain the subdivision error proportion. That is, when the current parameter approaches the control group with larger error proportion, the reduction amplitude is reduced, and the result approaches the risk upper limit, and when the current parameter is far away from the high risk control group, the reduction amplitude is increased, and the result is shifted to the low risk direction.
In one embodiment, in S3, the obtaining, according to the error ratios corresponding to the two reference parameter sets of the ith subdivision parameter, the subdivision error ratio corresponding to the ith subdivision parameter in the current processing parameter set is expressed as follows:
Wherein, the method comprises the steps of,
For the subdivision error ratio corresponding to the ith subdivision parameter,For an error ratio corresponding to a reference parameter set,For the error ratio corresponding to the other control parameter set,The larger value of the i-th subdivision parameter of the two comparison parameter sets,The smaller of the i-th subdivision parameters of the two comparison parameter sets is taken,The value of the ith subdivision parameter of the current processing parameter set is obtained.
In the present embodiment, the present invention is also directed to a method for manufacturing a semiconductor device,For parameter position normalization, the relative position of the current parameter value in the historical sequence interval is quantized, and the physical quantity (such as 90 m/min-100 m/min) is converted into a continuous proportion of 0-1. At the same time cooperate withWhen (when)When (such as cutting speed ascending sequence error ratio 0.3-0.6), the expression is expressed byGenerating positive gradients, e.g. 0.3 fold increaseStacking the increment according to the position proportionWhen (such as 0.9 to 0.5 of the material hardness ascending sequence error ratio), the expression is expressed byGenerating a negative gradient, e.g. 0.4 times smallerAnd (5) proportionally reducing the basic value. The unified processing of ascending/descending trend is realized, and the limitation of the preset trend direction is avoided.
Further, the method comprises the steps of,And realizing dynamic gradient generation, and constructing a linear response gradient by using the error proportion difference of the historical control group. In a nonlinear sensitive interval (such as 0.15-0.20 mm/rev error 0.4-0.8), the gradient value is +0.4, the increment is added by +0.24 when the position proportion is 0.6, and in a negative correlation interval (such as coolant flow increase-error decrease), the gradient is automatically negative to realize attenuation calculation. Dynamically adapts to any monotonic trend (positive/negative correlation), eliminating the need for manual judgment.
Further, the method comprises the steps of,Implementing baseline plus increment mechanism to control group in historyAs a benchmark, the gradient effect is superimposed on the displacement.
To sum up, in the existing method, the parameter sequence is ignored, the sub-division parameters which do not appear in the historical processing parameter set are regarded as discrete points, and the expression realizes forced identification, for example, 95m/min is located at the midpoint of the historical data from 90m/min to 100m/min, and the trigger error ratio is estimated to be 0.45 from the midpoint of 0.3 to 0.6. Further, outputOther parameter couplings (such as speed calculation ignoring feed fluctuation) are stripped, so that the weighted aggregation of S4 can truly reflect the weight dominance of various parameters. In summary, the expression in this embodiment is compatible with the positive/negative correlation parameters, and outputs the "decoupled" subdivision error ratio, so that the multiparameter comprehensive prediction (S4) has physical rationality.
As shown in fig. 4, in one embodiment, the method further includes:
S5, obtaining the gear design size of the target gear, and obtaining prediction accuracy data according to the prediction error proportion.
In the present embodiment, in S5, the predicted theoretical risk value is converted into a physical size instruction that can directly guide production, and the final closed loop of the prediction model is completed. This step first calls for the gear design size of the target gear (i.e., the theoretical standard value determined in S11, e.g., tooth thickness of 5.000 mm), and then, in combination with the predicted error ratio output in S4 (e.g., 0.59), the fluctuation range of the intended machining size is reversely deduced. Specifically, the prediction error proportion is multiplied by the maximum allowable tolerance, the expected absolute size deviation (0.59 multiplied by 0.02mm approximately equal to 0.012mm when the maximum allowable tolerance is 0.02 mm) under the current parameter combination is calculated, and finally the deviation (5.000 mm plus or minus 0.012mm as the tooth thickness theoretical value) is superimposed on the reference design size, so that prediction precision data (the actual measurement size is expected to fall within the interval 4.988-5.012 mm) which can be directly used for quality inspection comparison is generated.
In one embodiment, the step S4 of obtaining the prediction error ratio corresponding to the current processing parameter set according to the subdivision error ratio corresponding to each subdivision parameter in the current processing parameter set is expressed as:
, Wherein, the method comprises the steps of,
For the prediction error ratio corresponding to the current set of processing parameters,For the number of categories of subdivision parameters in the current set of processing parameters,The contribution weight for the ith subdivision parameter,And the subdivision error ratio corresponding to the ith subdivision parameter.
In the present embodiment, the present invention is also directed to a method for manufacturing a semiconductor device,The method aims to quantify the differential contribution of each processing parameter to the precision and break through the limitation of equal processing of the traditional method. A slight change in cutting speed (e.g., 90m/min to 95 m/min) may increase the error ratio by 0.2, whereas an equivalent change in feed rate is only increased by 0.05-both effects are essentially different. By contributing weights(E.g. speed weight [. Times. ]=0.6), Feed amount%=0.3)), The forcing model focuses the core parameters (speed) and suppresses the secondary parameter (feed) disturbances. The weight distribution is based on a physical mechanism (e.g., cutting thermal distortion is 2 times more sensitive to speed than feed).
Further, the method comprises the steps of,For decoupled integration of the subdivision errors, the independent parameter influence estimates of the S3 outputs are fused, i.e. the subdivision error ratiosAnd constructing global precision prediction. By weighted summation it is ensured that the speed-dominant error fluctuations are not diluted by the secondary parameters. Each of which isOther parameter interference has been stripped (done in S3) such that the weighting result truly reflects the multiparameter coupling effect.
Further, the method comprises the steps of,For weight normalization constraint, the prediction error is scaledIs strictly limited to the [0,1] interval and is consistent with the definition domain of the error proportion of S1. If not normalized, the output may exceed the upper limit of the tolerance proportion (100%), and the meaning of risk early warning is lost. After normalization, ensureAlways indicates the 'occupied proportion tolerance', and can directly trigger the grading alarm.
Still provide a gear hobbing machine tool machining precision prediction data processing system, the system includes:
The data acquisition module is used for acquiring a target gear, acquiring a plurality of historical processing parameter sets corresponding to the target gear, acquiring historical precision data after gear cutting processing is completed according to each historical processing parameter set, and acquiring error proportions corresponding to each historical processing parameter set according to the historical precision data corresponding to each historical processing parameter set, wherein each historical processing parameter set comprises a plurality of subdivision parameters;
the sequence generation module is used for sequentially arranging a plurality of historical processing parameter sets according to the sequence from small to large of the values of the subdivision parameters in the historical processing parameter sets and forming a parameter set sequence corresponding to each subdivision parameter;
The data processing module is used for acquiring a current processing parameter set which corresponds to the target gear and comprises a plurality of subdivision parameters, acquiring a position where a value of an ith subdivision parameter of the current processing parameter set in a parameter set sequence corresponding to the ith subdivision parameter is located, acquiring historical processing parameter sets corresponding to two adjacent values of the position and serving as two comparison parameter sets of the ith subdivision parameter, and acquiring subdivision error proportion corresponding to the ith subdivision parameter in the current processing parameter set according to error proportion corresponding to the two comparison parameter sets of the ith subdivision parameter;
and the data prediction module is used for obtaining the prediction error proportion corresponding to the current processing parameter set according to the subdivision error proportion corresponding to each subdivision parameter in the current processing parameter set.
In one embodiment, the data acquisition module is further configured to acquire a gear design size of the target gear, acquire historical machining sizes corresponding to the historical machining parameter sets according to the historical accuracy data corresponding to the historical machining parameter sets, and acquire an error ratio according to the gear design size and the historical machining sizes.
In one embodiment, the data processing module is further configured to obtain a difference value of error ratios corresponding to two reference parameter sets of the ith subdivision parameter and serve as a total error variable, and obtain an error ratio corresponding to a smaller value of the two reference parameter sets of the ith subdivision parameter, obtain a difference value of values of the ith subdivision parameter in the two reference parameter sets and serve as a reference parameter variable, obtain a difference value of values of the ith subdivision parameter in the current processing parameter set and a difference value of values of the ith subdivision parameter in the smaller value reference parameter set and serve as a current parameter variable, and obtain a subdivision error ratio corresponding to the ith subdivision parameter according to the current parameter variable, the reference parameter variable, the total error variable and the error ratio corresponding to the smaller value of the two reference parameter sets corresponding to the ith subdivision parameter.
In the present embodiment, it is to be noted that, regarding the above-described gear hobbing machine machining precision prediction data processing system, the specific manner in which the operation is performed has been described in detail in the embodiment regarding the gear hobbing machine machining precision prediction data processing method, and will not be explained in detail here.
The preferred embodiments of the present invention have been described in detail above with reference to the accompanying drawings, but the present invention is not limited to the specific details of the above embodiments, and various simple modifications can be made to the technical solution of the present invention within the scope of the technical concept of the present invention, and all the simple modifications belong to the protection scope of the present invention.
In addition, the specific features described in the above embodiments may be combined in any suitable manner without contradiction. The various possible combinations of the invention are not described in detail in order to avoid unnecessary repetition.
Moreover, any combination of the various embodiments of the invention can be made without departing from the spirit of the invention, which should also be considered as disclosed herein.
It should be noted that the foregoing embodiments are merely illustrative of the technical solutions of the present invention and not limiting, and although the present invention has been described in detail with reference to the foregoing embodiments, it should be understood by those skilled in the art that the technical solutions described in the foregoing embodiments may be modified or some or all of the technical features may be replaced equally, and these modifications or substitutions do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention, and all the modifications or substitutions are included in the scope of the claims and the specification of the present invention.