US20130326851A1

US20130326851A1 - Method and apparatus for component assembly using continuous selection

Info

Publication number: US20130326851A1
Application number: US13/607,636
Authority: US
Inventors: Charles B. Woodhull; Ashley N. Saulsbury; Nicholas I. Reid; Phillip W. Hum; Vashist Vasanthakumar
Original assignee: Apple Inc
Current assignee: Apple Inc
Priority date: 2012-06-12
Filing date: 2012-09-07
Publication date: 2013-12-12
Also published as: CN103488101B; US10054927B2; US9134724B2; US20150331410A1; CN103488100A; US20130331975A1; CN103488101A

Abstract

A manufacturing process for providing an assembly formed of a first piece and a second piece is described. The manufacturing process is carried out by performing at least the following operations: receiving the first piece characterized in accordance with at least a first attribute, selecting the second piece from a buffer, the selecting based in part upon a matching value of a second attribute in relation to the first attribute, replacing the selected second piece with another second piece such that the number of second pieces in the buffer remains about the same, and forming a kit of the first part and the selected second part.

Description

CROSS-REFERENCE TO RELATED APPLICATION

This application claims the benefit of U.S. Provisional Application 61/658,851 filed Jun. 6, 2012, which is incorporated herein by reference in its entirety and for all purposes.

FIELD OF THE DESCRIBED EMBODIMENTS

The described embodiments generally relate to manufacturing. In particular, assembly of manufactured parts using continuous selection is described.

DESCRIPTION OF THE RELATED ART

In manufacturing products are typically assembled from multiple parts. Those parts are often made of different materials and/or constructed using different manufacturing steps. Consequently the yield of “good” assembled products is a function of at least two factors, control of the manufacturing steps to ensure functioning and/or in-specification components, and the specific tolerances of the assembled product.
In most cases, designers strive to ensure that individual components are manufactured with tight enough tolerances such that when the parts are brought together and assembled, the final product meets its overall specifications. For example, the process to cut glass for a window is usually sufficiently controlled so that it is neither too big nor too small for a corresponding window frame. Similarly, window frames are manufactured to a certain size and tolerance to ensure that the corresponding glass will fit. Accordingly, both the glass and the frame are cut to some nominal size(s) so that when assembled the gap between then is within the required gap specification. Despite variations between pieces during manufacturing, tolerances are sufficiently controlled to ensure that the pieces fit together appropriately.
However, when design tolerances approach or exceed the ability of the manufacturing processes to build individual components, the yield of assembled pieces decreases because the probability of finding two compatible components at random decreases. Situations like this can arise for cosmetic reasons, such as minimizing the gap between two pieces, or ensuring continuity of color between two different materials. For situations where the assembled design tolerances significantly exceed the manufacturing capability of components other techniques are required in order to maintain the yield of assembled products.
Therefore, accurate and reliable techniques for selecting parts for assembly of a manufactured product is desired.

SUMMARY OF THE DESCRIBED EMBODIMENTS

This paper describes various embodiments that relate to a system, and method for continuously selecting a second part for use with a first part.
In one embodiment, a manufacturing process is described. The manufacturing process can be carried out by determining an attribute value of a first part, the attribute value used to identify a matching second part from a buffer of N second parts, dividing the buffer of N second parts into two of more groups of second parts, using the first part attribute to identify a matching second part from within one of the groups of second parts, when the second part has a second part attribute value that is within an acceptable range of second part attribute values, removing the second part from the part buffer and replacing the selected second part with another second part with attributes corresponding to the group from with the second part was selected. In one embodiment, the two or more groups can be based on predefined sub ranges of values.
In another embodiment, an inventory control method can include the steps of receiving an incoming first part, determining a first and a second value of the first part, storing second parts in a buffer including two or more groups of parts based on sub-ranges of first and second attribute values and selecting a second part from the buffer when a first and second values of functions of merit related to first and second attributes are less than a predetermined amount, removing the selected matching part from the buffer and providing a kit including the first part and the selected second part.
Other aspects and advantages of the invention will become apparent from the following detailed description taken in conjunction with the accompanying drawings which illustrate, by way of example, the principles of the described embodiments.

BRIEF DESCRIPTION OF THE DRAWINGS

The invention will be readily understood by the following detailed description in conjunction with the accompanying drawings, wherein like reference numerals designate like structural elements, and in which:

FIGS. 1A-1C graphically illustrate assembly process arranged to assemble parts into an assembly that generally relies upon a selected attribute, or attributes, of the parts.

FIGS. 2A and 2B show more detailed relationship of the parts shown in FIGS. 1A-1C.

FIG. 3 shows representative yield curve in accordance with the described embodiments.

FIGS. 4A-4C illustrate the relationship between mean offset value and the yield curve shown in FIG. 3.

FIG. 5 is a graph showing multidimensional attribute space that can be used in a merit function.

FIGS. 6A-6B show representative assembly process and apparatus, respectively, in accordance with the described embodiments.

FIG. 7 shows a flowchart detailing assembly process in accordance with the described embodiments.

FIG. 8 shows a flowchart describing process for determining mean offset values for providing an optimal range of assembly yield to compensate for manufacturing process drift.

FIGS. 9-12 show representative inventory control protocols used to optimize assembly yield in accordance with the described embodiments.

FIG. 13 is a block diagram of an electronic device suitable for use with the described embodiments.

FIGS. 14A-14C are graphs showing probability distribution of attributes of a part.

FIG. 15 shows distributions of a first and a second part together.

FIG. 16 is a graph showing the second part distribution divided into groups.

FIGS. 17A-17B is a flow chart of method steps 1700 for a continuous selection manufacturing process, in accordance with one embodiment described in the specification.

DETAILED DESCRIPTION OF SELECTED EMBODIMENTS

Reference will now be made in detail to representative embodiments illustrated in the accompanying drawings. It should be understood that the following descriptions are not intended to limit the embodiments to one preferred embodiment. To the contrary, it is intended to cover alternatives, modifications, and equivalents as can be included within the spirit and scope of the described embodiments as defined by the appended claims.
The following description relates in general to a manufacturing assembly operation and process that can embody concepts that can be employed separately or in conjunction for assembling components to tolerances beyond the ability of their manufacture. A first concept is referred to as “continuous selection” by which it is meant that the best fitting two components can be selected from a continuously updated pool of components. A second concept can be described in terms of a mechanism of inventory control that can ensure matching statistical distributions of these components to minimize yield loss and ensure that as many components can be matched together as possible. It should be noted that although the embodiments are described (for simplicity) in the context of two components to be assembled, it is clear that the processes described can be extrapolated to handle the assembly of multiple components simultaneously. Furthermore, the basic principles of operation described herein can be implemented as a fully automated, a semi-automated or an entirely manual manufacturing process.
For the manufacture of any physical component, there will generally be a statistical distribution of values for any attribute of interest. Typical attributes of interest can include physical attributes such as length, height, color, weight. Other attributes can be referred to as relational attributes between various components. Relational attributes can include relative size, relative weight, and so on. Relational attributes can be useful can various components are mated in an assembly operation. For the most part, the statistical distribution of attribute values is generally Normal (Gaussian) in nature (or a close approximation). In other words, for a group of components, as the number of components in the group increase, the tendency of a particular attribute is to symmetrically group about values that are close to the mean value of the attribute for the group. Therefore, for sake of simplicity only, the following discussion presumes that all distributions described herein are essentially Gaussian in nature and that for assembly the statistical distribution of component attributes is the same. It should be noted, however, this is not a necessary factor for the described embodiments and should not be construed as limiting in any manner.
In the described embodiment, the process of continuous selection is based upon the idea that amongst a randomly selected pool of components to be assembled there is a very high probability that there will be a set of components that fit together with the best possible match of any desired attribute(s). For example, between a bag of screws and a bag of fittings, there is a high probability that there is at least one screw that best fits a given fitting. This can also be extrapolated to the assembly of multiple components. Amongst a given pool of pieces there will be a combination that fit together best. Clearly the larger the pool(s) of components from which to select from the better the fit of the best fitting components will be. Put differently, the more components from which to select, lowers the probability that the best fitting components will still be a poor fit. In the limit, with an infinite pool of components there will always be a set of components that are a perfect match. Therefore, a critical variable in this methodology is the selection of a pool size which given the statistical distribution of components and will ensure an acceptable yield of assembled parts.
In one embodiment, a continuous selection process can be implemented in an assembly operation involving a first part paired with a matching, or best fit, second part. In one implementation, the first part is identified and a pre-selected attribute is measured. A second part in the form of a best fitting component is selected from a pool (or buffer) of candidate components. As a pair formed of the first and second (or set if more than two candidate parts) of best fitting parts are identified, the best fitting second part is removed from the pool and matched with the first part. The result of the selection is to reduce the pool size (i.e., the number of components available for selection) which in turn reduces the probability of a subsequent best fit. Consequently, as a matching part is removed from the buffer, a new part(s) is inserted into the selection pool ensuring a constant pool size as well as a predicable yield of assemblies.
For example, during an assembly of a portable media device, housing for the portable media device can be selected. An opening in the housing used to accommodate a cover glass used to protect a display assembly can be measured. In this situation, a gap value being a difference in size between the opening in the housing and the cover glass can be the attribute of interest. Therefore, in accordance with the process, for a particular housing, the size of the opening can be used to find best fitting cover glass by which it is meant that a maximum gap value is less than a pre-defined design specification value. Once the best fitting cover glass is identified and removed from the buffer of cover glass(s), another cover glass can be inserted into the pool in order to maintain the size of the pool of cover glass(s) substantially constant. In this way, a probability that a subsequent housing can be matched with a best fit cover glass can be maintained as acceptable.
Variation in manufacturing processes can result in a drift, or variation, in characteristics of both incoming components and candidate component from which is selected a best match. Therefore, in some embodiments, both the distribution and/or number of candidate components can be varied in order to compensate for the shift in manufacturing process. In this way, assembly yield can be maintained at an acceptable level regardless of any variation in the manufacturing processes used to produce either the incoming components or candidate components.
In one embodiment, an inventory control protocol can be based upon a determination of process drift associated with a change in a mean value of a distribution of an attribute of interest. In another embodiment, a distribution of candidate components can be altered by replacing a matched component with a component pre-sorted in such a way that a distribution of candidate components becomes more like that of the incoming components. In this way, the likelihood of finding a matching component for an incoming component is commensurably increased. In another embodiment, an oldest not-matched part in the buffer can be removed and replaced. The removed part can be recycled when a distribution of incoming parts indicates that the removed part is likely to be matched with an incoming part. In another embodiment, the incoming parts can be characterized and if there is no part in the current buffer population that will match the incoming part and be within an outgoing specification, then the incoming part can also be set aside if and until the distribution of candidate parts in the buffer has changed such that the removed incoming part can be matched with the finished product being deemed acceptable.
These and other embodiments are discussed below with reference to FIGS. 1-17C. However, those skilled in the art will readily appreciate that the detailed description given herein with respect to these figures is for explanatory purposes only and should not be construed as limiting.
FIGS. 1A-1C graphically illustrate assembly process 100 arranged to assemble part 102 and part 104 into finished product 106 that generally relies upon a selected attribute, or attributes, of part 102 and part 104. If, for example, part 102 takes the form of peg 102 having outer diameter W_D1and part 104 takes the form of collar 104 having inner diameter W_D2, then assembly process 100 can require that peg 102 be inserted into collar 104 to form finished product 106. In order to model assembly process 100, it can be presumed that a selected attribute of any component required in assembly process 100 exhibits a random variation that conforms to a normal, or Gaussian, probability distribution. In this way, for example, peg 102 can have a selected attribute in the form of diameter W_D1that exhibits a random variation along the lines shown in graph 108 and collar 104 can have a selected attribute in the form of inner diameter W_D2that exhibits a random variation along the lines shown in graph 110 each of which are in the form of a normal or Gaussian distribution.
As well known in the art, a normal distribution can be characterized by a mean (m) and a standard deviation (S_D) that shows how much variation or dispersion exists from the mean. In other words, a low standard deviation indicates that the data points tend to be very close to the mean, whereas high standard deviation indicates that the data points are spread out over a large range of values. Accordingly, with regards to a manufacturing process, the standard deviation of the distribution of an attribute for a part can be a measure of a capability of the process with regard to that attribute used to manufacture the part. In other words, a well controlled process can produce a number of parts having an attribute whose variation about the mean is small (i.e., a low standard deviation) whereas a process that is less well controlled will produce parts having an attribute whose variation about the mean is greater and therefore has a higher standard deviation.
For example, standard device S_D1of distribution 108 can indicate an amount of control used to manufacturing peg 102. In other words, if the manufacturing process used to form peg 102 is well controlled, then the distribution of peg widths (w_D1) about mean m₁will be small as indicated by a low value for standard device S_D1and narrower graph 108. On the other hand, if the manufacturing process used to form peg 102 is less well controlled then the observed values for peg width (w_D1) are dispersed more broadly about mean m₁and standard device S_D1will have a larger value and graph 108 will be flatter and more spread out.
In order to assure that the majority of parts have a match (i.e., a peg fits into a collar), the standard distributions are made small enough or the mean values are far enough apart that the number of pegs that are too big to fit in the holes is small. Accordingly, the relationship between the distributions of components in any assembly process can have significant impact on the overall yield of that assembly process. For example, by setting the distributions of the components such that there is substantial overlap between the distributions (i.e.; S_Dis >> than Δm), then the assembly yield can be greater than 99% since the number of parts that cannot be matched due to attribute incompatibility is very small. Although the overall yield may be high, the number of assembled parts that pass an outgoing quality check based on a particular design specification, such as a gap distance between peg 102 and hole 104, can be reduced to the point where an effective yield can not meet the requirements of the design specification and the assembled part can be rejected. Rejecting already assembled parts is problematic use of resources due to the time and cost involved in processing that is ultimately to no avail.
In the context of this discussion, assembly yield can be defined as the probability that a first component (such as collar 104) can be successfully matched with a best fitting second component (such as peg 102) based upon at least one attribute. Yield can depend upon many factors such as the number of available components as well as the distribution of those components with regards to an attribute of interest. Therefore, in order to maximize yield as well as compensate for any manufacturing based drift, it can be a significant advantage in an actual manufacturing environment to be able to control, or at least modify, both the number and distribution of candidate components.
As discussed above, one technique that can be used to monitor manufacturing drift is based upon the standard deviation of a distribution of an attribute. In other words, a manufacturing process associated with a low value for standard deviation can provide a larger number of parts that do not exhibit a large variance from the mean of the distribution. However, any changes in the manufacturing process can influence the standard deviation of the distribution of the parts produced. For example, if the manufacturing process is becoming less well controlled with regards to a particular attribute (due to a variation in environmental factors, for example) then a group of parts will show on average an attribute of interest having a greater variation with respect to the mean value of that attribute that will be evidenced by an increase in the standard deviation. Therefore, a pro-active inventory control process can be utilized to compensate for any drift in the manufacturing process that can help to maintain a process yield within an acceptable range by monitoring the nature of the distribution of candidate components (using the standard deviation of the parts produced as an indicator) and adjusting the nature of the distribution(s) accordingly.
These concepts can be demonstrated using FIGS. 2A and 2B that show peg 102 and collar 104 in more detail especially with the relationship between outer diameter W_D1and inner diameter W_D2. FIG. 2A in particular shows peg 102 biased to one side of collar 104. In this situation, assembly yield Y_acan be defined in terms of the probability of finding a best fitting part from a group of N candidate parts for an incoming part. In other words, for an incoming part in the form of collar 104, any candidate part in the form of peg 102 will positively contribute to assembly yield Y_ain accordance with Eq. (2):
W _D1 <W _D2 Eq. (2).
However, in some instances, an attribute of interest can be relational. For example, still referring to FIG. 2A, if the attribute of interest is gap value (GV) of gap 202 can be represented as the difference between inner diameter W_D2(of collar 104) and outer diameter W_D1(of peg 102) according to Eq. (3):
GV=W _D2 −W _D1 Eq. (3).
As shown in FIG. 2B, for cosmetic or other reasons, it may be desirable to constrain gap value GV to a narrow range of possible gap values. In this way, gap value GV can be evaluated for acceptability by comparing gap value GV with a range of acceptable gap values that can be substantially smaller than the potential range of best fit matches between collar 104 and peg 102. This range of acceptable values (represented by “∂”) can correspond to an outgoing specification along the lines of a design specification. In this way, only those candidate components that can satisfy the design specification in accordance with Eq. (4) can be considered acceptable and can positively contribute towards an effective assembly yield Y_eff:
GV=W _D2 −W _D1<∂ Eq. (4).
It should be noted that since assembly yield Y_ais generally less than (or at best equal to) effective yield Y_eff,any assembled part (such as finished product 106) that does not meet the design specification can be considered an outgoing reject and must either be re-worked or rejected entirely. In this way, a substantial amount of processing resources that have been allocated is essentially wasted. Therefore, maintaining assembly yield Y_aand effective yield Y_effas close as possible is an important consideration in any manufacturing process.
For reasons stated above, assembly yield Y_acan be dependent upon drift in the processes used to manufacture either peg 102 or collar 104. For example, FIG. 3 shows representative yield curve 300 associated with a specific attribute (such as gap value GV) in accordance with the described embodiments. It should be noted that the characteristics of yield curve 300 can depend upon a relationship between the distribution of both incoming parts and a group of N candidate parts. Assembly yield Ya can be a function of mean value offset Am in accordance with Eq. (5):
Y _a =Y _a(Δm) Eq. (5).
As illustrated in FIGS. 4A-4C, (presuming all distributions described are essentially normal), distributions 402 and 404 can be related to assembly yield Y_ausing a mean value “offset” that can be embodied as a difference in mean values of each distribution in accordance with Eq. (6):
Δm=m ₄₀₂ −m ₄₀₄ Eq. (6).
Referring back to FIG. 3, from the results of FIGS. 4A-4C, it is evident that there is a range of mean value offsets that can be associated with what can be considered an acceptable assembly yield Y_accept. For example, offset value Δm₁corresponding assembly yield Y₃₀₂can be less than assembly yield Y₃₀₄associated with offset value Δm₂which, in turn, is greater than assembly yield Y₃₀₆associated with offset value Δm₃. Therefore, a range of offset values (shown as Δm_ato Δm_b) can be used to modify distributions 402 and 404 in order that assembly yield Ya remains within the acceptable range of assembly yield associated with acceptable assembly yield Y_accept. It should be noted that it is preferable for most assembly operations for operating point OP₁be located on a portion 308 of yield curve 300 having a shallow slope as compared to portion 310 thereby mitigating to some extent variations in manufacturing processes. Various inventory control protocols along the lines discussed below can be used to maintain a corresponding assembly process in acceptable regime.
The following discussion provides a more detailed description of the selection process during an assembly operation in which an attribute, or attributes, of a part is determined and a search for a corresponding part that most closely matches the measured attribute(s) is performed. More generally, the search can be based upon what is referred to as a merit function M. Merit function M can be related to various parameters in accordance with Eq. (7):
M=M (ω_i , A _i) Eq. (7);

- where: ω is a weighting factor; and A_iis selection attribute.

The selection attribute can be any parameter used as a selection criterion of choosing a best matching part. For example, selection attributes can be size (diameter of peg 102 for example), color, texture, gap distance, and so on. There can be any number of selection attributes used to choose a best match for a particular process. It should be noted, however, that as the number of attributes increases, so does the number of parts required in a part buffer and also the amount of selection resources (both time and computation) required. Moreover, in some embodiments, weighting factor ω_ican be used to provide relative weights between various attributes. For example, if for a particular assembly process, an attribute associated with color is substantially more important than texture, then a weighting factor for the color attribute can be greater than that for the texture attribute. Therefore, the judicious choice of selection attributes can be an important factor in an efficient implementation of the selection and assembly process.
As described above, merit function M can be multi-dimensional having various attributes such as color, size, rotation that form components in what can be referred to as a multi-dimensional attribute space S_Agraphically illustrated in FIG. 5 as attribute space 500 showing a generalized point 502. In this implementation, point 502 can represent a point in attribute space 500 having generalized attribute coordinates {ω_i, A_i} that can be associated with merit function M{ω_i, A_i} as in Eq. (7) where i is the number of attributes. Also shown are the projections of merit function M onto corresponding attribute axes a₁, a₂, and a₃.
Therefore in an automated manufacturing process, by using merit function M to select best fitting part, component parts can be manufactured with slack manufacturing tolerances but can sustain high effective yields without the need to sort or pre-sort. Moreover, an additional degree of quality control can be achieved by determining if the incoming part will have any match and if not, it can be discarded immediately thereby preserving valuable manufacturing resources. In this way, by applying merit function M it can be assured that incoming parts can be matched with a best fit counterpart that meets all design specs even though the manufacturing specs used to create the parts are looser than the target design spec. Accordingly, the number of parts required to populate a buffer can be a function of the number of attributes as well as the ratio of the design tolerance to the manufacturing capability. For example, a tight tolerance coupled with a loose manufacturing capability will require a greater number of parts to meet a corresponding design tolerance. Also, as the number of attributes increases, the number of parts required can also increase.
FIG. 6A graphically illustrates representative manufacturing operation 600 in accordance with the principals discussed above. Operation 600 can be used to manufacture any number of products using any number of components using any number of attributes. However, for the sake of clarity, operation 600 will be described in terms of basic manufacturing operation that describes a single incoming part being matched with a single component based upon a single attribute. In this example, incoming part 602 can be evaluated by determining apparatus 604 for attribute A₁. Attribute A₁can be any suitable attribute related to a measurable property of incoming part 602. Attribute A₁can be a size of incoming part 602 (or size of a particular aspect of incoming part 602), the color of incoming part 602, and so on.
In some situations, the measured attribute can be relational in that a measured property of incoming part 602 can be used in relation (i.e., compared to) to a measured property of a candidate part. One such relational attribute can be a measure of a degree of parallelism between portions of incoming part 602 and a selected best fit part. The degree of parallelism can be used as a selection criterion for choosing the best fitting part. Referring again to peg 102 and collar 104, an attribute that can be ascribed to peg 102 is circular conformity. In other words, how close does peg 102 (or collar 104) conform to a geometric circle? By comparing radii of curvature at selected points between peg 102 and collar 104, a measure of parallelism can be defined that provides an indication of just how well peg 102 fits into collar 104. In other words, the degree of conformity can also provide an indication of the uniformity of gap 202 between peg 102 and collar 104 in finished product 106.
Referring back to FIG. 6A, incoming part 602 can be scanned by determining apparatus 604. Determining apparatus 604 can be sensitive to a chosen attribute. For example, if the chosen attribute is viewable using an image capture device, then determining apparatus 604 can take the form of a camera. On the other hand, if the chosen attribute is temperature based, then determining device 604 can take the form of a thermal sensor such as an IR sensor, thermometer, etc. For the remainder of this discussion and without loss of generality, determining apparatus 604 is presumed to take the form of a visible light camera 604. In this way, camera 604 can optically scan incoming part 602. In one embodiment, camera 604 can utilize processing resources (either on-board or external) to convert images captured by camera 604 into a set of measured values (MV) that will be referred to hereinafter as to as data cloud 606. Data cloud 606 can represent a virtual rendering of incoming part 602 as the set of measured values (MV) can be used to digitally “reconstruct” incoming part 602. For convenience as well as computational efficiency, data cloud 606 can be stored locally at camera 604 or in an external data base (not shown) for later use. In any case, data cloud 606 can be digitally processed using available computational resources. In one embodiment, processing of data cloud 606 can involve comparing data cloud 606 (or a representative sample of the measured values MV) to a reference datum (such as a CAD data) used to define the geometric structure of incoming part 602 and more particularly, that aspect related to selected attribute A1. In one embodiment, the reference datum can correspond to a mean value of a statistically significant number of incoming parts. In this way, the comparison can provide a measure of the dispersion about the mean m of incoming parts 602 and provide an indication of the nature of the distribution of incoming parts with respect to the attribute of interest.
In one embodiment, a representative sample of data cloud 606 in the form of measured values MV can be compared to a set of threshold measured values. In one embodiment, the set of threshold values can correspond to those measured values of selected attribute A1 for which a best fitting part is not available. For example, when incoming part 602 is scanned by camera 604, a representative sample of the measured values MV can be forwarded to comparator 608 that can compare the measured values received to threshold values. If comparator 608 determines that a pre-determined number of measured values are outside of an acceptable range and fall into what is shown as Region I, then incoming part 602 can be removed from assembly process 600. In this way, valuable time and computational resources will not be wasted on incoming parts that cannot be matched with a corresponding best fitting part. Moreover, this concept can be extended to using an outgoing quality specification in that those incoming parts that are determined to not be able to meet an outgoing quality check when matched with an available best fitting part, the incoming part can be removed from assembly process 600. This determination can be based upon the scanned characteristics of incoming part 602 and a data base of known characteristics of parts available for potentially matching incoming part 602. If there is no available matching part, then incoming part 602 can be set aside for a subsequent attempt at matching. However, in this situation, the incoming parts removed from assembly process 600 can be set aside if and until matching parts can be found that when used will meet the outgoing quality criteria.
In those situations where incoming part 602 has been determined to be acceptable in that there is at least one matching part in buffer 610 of N parts, then selector 612 identifies and selects part 614 having an attribute A1 that satisfies merit function M that is then matched with incoming part 602 to form assembled part 616. It should be noted that by best fit it is meant that of N parts in buffer 610, selector 612 will select part 614 having characteristics that most closely match those called out in merit function M but not necessarily a “perfect” match, therefore the notation M() to indicate that the matching characteristics are not necessarily an exact match but are nonetheless the “best” match of the N parts available in buffer 610. It should be noted that prior to inclusion in buffer 610, each of the N parts included therein were scanned and characterized and the corresponding data stored and made available to selector 612. Once part 614 has been identified, replacement part 618 replaces best fitting part 614 previously removed from buffer 610. In this way, the number of parts in buffer 610 remains substantially unchanged thereby maintaining an incoming yield at an acceptable level.
It should be noted that after a period of time that depends in part upon the number of parts in buffer 610, once a substantial number of the best fitting parts are removed from buffer 610, the remaining parts will tend to be associated with portions of the distribution having characteristics that are less likely to satisfy the merit function of an incoming part. In other words, the distribution of parts within the buffer becomes distorted with respect to the distribution of incoming parts and the probability of finding a part in the buffer that is a best fit to an incoming part becomes less. Therefore, in order to assure consistent and acceptable yields, an inventory control protocol can use buffer rebalancing to improve the number of best matching parts in buffer 610 thereby increasing the monitored yield. The monitored yield can indicate a probability that an incoming part can be matched successfully with a part in buffer 610. By successful it is meant that the attribute of the best matching second part is within an acceptable range of attribute values (akin to a design spec).
The rebalancing of buffer 610 can take many forms. For example, in one embodiment, replacement parts can be “pre-binned” prior to being placed into buffer 610. By pre-binned it is meant that the distribution of buffer 610 can be modified by selecting replacement parts (parts that replace those selected) that are selected prior to inclusion in buffer 610 to have an attribute value in accordance with the attribute value of the part selected. In this way, the mean of the distribution of the parts in buffer 610 can be made closer to that of the parts most likely to be selected. This rebalancing of buffer 610 can improve yield by preferentially pre-selecting those replacement parts more likely to be selected than would replacement parts selected in a less directed, or more random manner.
In another embodiment, the distribution of parts in buffer 610 can be selectively modified by removing one or more parts from buffer 610 that have remained in buffer 610 a period of time greater than a pre-determined amount of time. In other words, those parts in buffer 610 that have not been selected are more likely than not representative of a part having an attribute value that falls out of the range of attribute values consistent with a best matching part. In this way, by removing those “oldest” parts from buffer 610 to be replaced with parts having an attribute value closer to one that would merit a match can have the effect of increasing overall yield. In other words, removing those parts in buffer 610 that have less desirable attributes can result in an increase in yield. It should be noted, however, that the parts removed can be sequestered for a later time when, perhaps, the attribute value distribution of incoming parts has shifted in such a way that renders the sequestered parts “more desirable”.
It should be noted, however, that the “age” of a part in buffer 610 can represent but one of many attributes that can be used to mark those parts in buffer 610 that possess attribute value(s) that render them undesirable with respect to the attribute values of the incoming parts. For example, some of the attributes that can be used to identify less desirable parts in buffer 610 can include a part size, a degree of mis-match with the incoming parts (i.e., the parts in buffer 610 that are the “worst matched”), and so on.
FIG. 6B graphically illustrates a manufacturing assembly operation 650 in accordance with the described embodiments. Operation 650 can be carried out using well known assembly equipment such as pick and place machines, robotic handlers, optical sensors such as cameras, and so on. Accordingly, first part 652 can be transported by transport mechanism 654. Transport mechanism 654 can take many forms. For example, transport mechanism 654 can take the form of a conveyer belt configured to carry first part 652 from a loading area (not shown) to the operation area in which assembly operation 650 can be carried out. Continuous selection apparatus 656 can include a number of operational modules and can take many forms. For example, continuous selection apparatus 656 can be distributed in nature by which it is meant that particular operational modules can be located where most efficiently used and be in communication with each other. In other embodiments, continuous selection apparatus 656 can be essentially a single unit having multiple components included therein (as shown in FIG. 6B). In some cases, continuous selection apparatus 656 can be any appropriate combination thereof.
As shown in FIG. 6B, continuous selection apparatus 656 can include determining apparatus 658 configured to determine a particular attribute, or attributes, of first part 652. The attribute can be any tangible aspect of first part 652 such as size, weight, color, smell, etc. It should be noted that determining apparatus 658 can be sensitive to one or more attributes and as such can provide a multi-dimensional data stream to computational unit 660 configured to store and process data. In this way computational unit 660 can provide a virtual rendering of first part 652 (along the lines of data cloud 606. Computational unit 660 can be in communication with comparator 662 that can be used to determine if the attributes of first part 652 are deemed to be acceptable in that the attribute(s) is within a range of acceptable attributes.
Comparator 662 can determine (with the assistance of computational unit 660, if need be) that first part 652 is acceptable, by which it is meant that the attribute(s) of part 652 is within a range of acceptable attribute values. Once deemed acceptable, instructions can be sent to selector unit 664 configured to identify and select a best matching part 666 from buffer 668. Selector unit 664 can take many forms such as a pick and place machine. Selector unit 664 can then transport selected part 666 to transporter 654 (or equivalent) for assembly with first part 652 to form finished part 668.
FIG. 7 shows a flowchart detailing process 700 in accordance with the described embodiments. Process 700 can be carried out by receiving an incoming part at 702. At 704, an attribute, or attributes, used to select a best matching part from a buffer of N parts can be associated with the incoming part. At 706, the incoming part can be scanned to determine an attribute value, or values, of the incoming part. At 708, the attribute value, or values, can be compared to database to determine if there is any part in the buffer that can be matched with the incoming part and meet acceptable criteria. For example, if the assembled part is subject to an outgoing quality check based upon, for example, a design specification, is there any part in the buffer that can be matched with the incoming part that will meet the design specification. If, at 708, it is determined that there is no part in the buffer that can be matched, then at 710, a determination is made if the incoming part is to be sent for subsequent processing. By subsequent processing, it is meant that although that there is no matching part currently in the buffer that will present an acceptable match for the incoming part, then the part can be held until such time that re-balancing the buffer will provide a matching part, then the incoming part is sent to recycle at 712, otherwise, the incoming part is rejected at 714.
Returning 708, if the incoming part is determined to be acceptable, then at 716, a determination is made if there is a best fitting part in the buffer. If there is a best fitting part in the buffer, then the best fitting part is selected at 718 and a replacement part is added to the buffer at 720. Returning back to 716, if it is determined that there is not currently a best fitting part in the buffer, then the incoming part is recycled at 712.
FIG. 8 shows a flowchart describing process 800 for determining mean offset values for providing an optimal range of assembly yield to compensate for manufacturing process drift. Process 800 can begin at 802 by selecting an attribute of interest. At 804, a desired value of the selected attribute is chosen At 806, mean offset values corresponding to an acceptable range of assembly yields is determined and at 808, distributions of incoming component and components in a buffer for storing candidate matching components are modified in accordance with the mean value offsets.
FIGS. 9-11 show representative inventory control protocols used to optimize assembly yield in accordance with the described embodiments. FIG. 9 shows a flowchart detailing process 900 for monitoring assembly yield. Process 900 begins at 902 by monitoring yield. In the described embodiment, monitored yield can be defined as a probability of an incoming first part having an associated best matching second part in a buffer having attributes within an acceptable range of attribute values (akin to a design spec). In this way, even if the first part has a best match in the buffer, the combination would not meet outgoing requirements, therefore, the first part is deemed to not have an effective best match. When at 904 it is determined that the yield is not acceptable, then the buffer of parts used to provide a best fit for an incoming part is re-balanced in accordance with any processes described by flowcharts shown in FIGS. 10-11.
In one embodiment, buffer rebalancing can include identifying the parts in the buffer that have remained unmatched for the longest period of time and removing those so identified. Accordingly, FIG. 10 shows a flowchart detailing process 1000 for determining an oldest non-matching component in the buffer begins at 1002 by evaluating an amount of time that each part in the buffer has remained unmatched. At 1004, the oldest non-matching part in the buffer is identified and removed. The removed oldest part is then replaced at 1006. In one embodiment, the removed part is subsequently recycled whereas in another embodiment, the removed part is sequestered until and if parts are received having attributes that render the sequestered part more “desirable”.
FIG. 11 shows a flowchart detailing process 1100 that modifies the distribution of parts in the buffer in an attempt to more closely align the distribution of parts in the buffer with the incoming parts. At 1102, sub-distributions (i.e., binning) of components in the buffer that more closely matches the distribution of incoming parts are identified. At 1104, the buffer is preferentially populated with components corresponding to the identified sub-distributions.
In one embodiment, buffer rebalancing can include identifying the parts in the buffer that are the “worst” matching in that those part having attribute values that differ the most from the attribute values of the incoming parts are removed from the buffer. Accordingly, FIG. 12 shows a flowchart detailing process 1200 for determining a worst matching part in the buffer that begins at 1202 determining an attribute difference value between the attribute value of the incoming part and all (or at least most) of the parts in the buffer. At 1204, the part associated with the largest difference value is identified. At 1206, the identified part is removed from the buffer and at 1208, and replaced with another part. In one embodiment, the replacement part can be pre-binned so as to modify a mean value of the distribution of parts in the buffer to more closely align with the mean value of the incoming parts.
FIG. 13 is a block diagram of a computer device 1350 suitable for use with the described embodiments. The computer device 1350 illustrates circuitry of a representative computing device. The electronic device 1350 includes a processor 1352 that pertains to a microprocessor or controller for controlling the overall operation of the electronic device 1350. The electronic device 1350 stores media data pertaining to media items in a file system 1354 and a cache 1356. The file system 1354 is, typically, a storage disk or a plurality of disks. The file system 1354 typically provides high capacity storage capability for the electronic device 1350. However, since the access time to the file system 1354 is relatively slow, the electronic device 1350 can also include a cache 1356. The cache 1356 is, for example, Random-Access Memory (RAM) provided by semiconductor memory. The relative access time to the cache 1356 is substantially shorter than for the file system 1354. However, the cache 1356 does not have the large storage capacity of the file system 1354. Further, the file system 1354, when active, consumes more power than does the cache 1356. The electronic device 1350 can also include a RAM 1370 and a Read-Only Memory (ROM) 1372. The ROM 1372 can store programs, utilities or processes to be executed in a non-volatile manner. The RAM 1370 provides volatile data storage, such as for the cache 1356.
The electronic device 1350 also includes a network/bus interface 1361 that couples to a data link 1362. The data link 1362 allows the electronic device 1350 to couple to a host computer or to peripheral devices such as a robot. The data link 1362 can be provided over a wired connection or a wireless connection. In the case of a wireless connection, the network/bus interface 1361 can include a wireless transceiver.
In some embodiments, the parts within the buffer can be characterized and arranged into groups. For example, groups can be formed by defining sub-ranges of acceptable values that were used above to determine part acceptance as described above.
A first part can have an attribute that, in turn, can have an acceptable range of values. This attribute can correspond to a like attribute on a second part such as the peg and hole example of FIG. 1A-1C. In practice, however, parts are rarely simple enough to have only one important attribute that may need to be within a proper range for proper fitment or operation. Often, parts can have two or more attributes that should be in acceptable ranges for a good fit or device operation. As a part has increasing complexity, the fitment of a second part into a first part can become more difficult. Sometimes, when applying continuous selection, the buffer of N parts can become skewed with parts (i.e., second parts) that may not be suitable for selection for a first part. One approach to managing the distribution of parts within the buffer divides the buffer into groups. These buffer groups can allow a more precise management of the part buffer. In one embodiment, the groups can be defined by non-overlapping sub-ranges of attribute values.
This may be best explained with an example. Consider a first part with an opening designed to receive a second part. The opening on the first part can have two attributes, such as a width and a height.
FIG. 14A-14C are graphs showing probability distributions of attributes of a part. FIG. 14A in particular shows the probability distribution of a first attribute of a first part. For example, the first attribute can be a width of an opening on the first part. The graph shows a probability curve 1402 that further shows a mean at 56.69 units. Furthermore, the tails of the probability curve 1402 fall off on each side of the mean such that the smallest probable width is 56.63 units and the largest probable width can be 56.75 units.
A second attribute of the first part can be height. FIG. 14B shows a probability distribution of a second attribute, such as a height of an opening on the first part. Curve 1404 shows a mean height of 10.66 units. The curve 1404 shows the smallest probable height to be 10.60 units and the largest height to be 10.72 units.
The possible values that the first and the second attributes can attain can be combined into a single graph. FIG. 14C plots possible first attribute values on the X-axis and possible second attribute values on the Y-axis. The curve 1406 represents a probable distribution of first and second attribute values for the first part. Returning to our example, the curve 1406 can represent possible values for the height and the width of an opening on a first part.
In many cases, there can be a second part that can be configured to interact with the second part. For example, the second part can be configured to fit into the opening of the first part. Naturally, the second part can have similar and related first and second attributes. Returning to our example, the second part can also have a height and a width, so that the second part can fit into the corresponding opening of the first part. Often, a designer can design the second part to be smaller, and therefore have a clearance gap, than the opening in the first part so that the second part can more easily fit into the first part. As was developed for the first part in FIGS. 14A and 14B, a curve can be derived to plot possible attribute values for the second part.
FIG. 15 shows distributions 1500 of first and second parts together. Curve 1406 which is a probable distribution of first and second part attributes for the first part. FIG. 15 can also include curve 1502 showing a probable distribution of first and second attributes of a second part. Those skilled in the art will recognize that, although illustrated in terms of width and height, the attributes and corresponding attribute values can be any possible attribute of any part, such as diameter, surface roughness, color match, planarity and so forth. FIG. 15 also shows that the second part (in this example) can be manufactured with less variance than the first part as expressed by a smaller diameter circle of curve 1502 compared to curve 1406. In other embodiments, the first part can be manufactured with less variance that the second part. The relationship of curve 1502 to curve 1406 (curve 1502 is lower than curve 1406, and positioned to the left) further shows that the second part (again, in this example) can be made smaller than first part. In some embodiments, when the second part is designed to fit within an opening of the first part, a smaller sized second part with respect to the first part attributes can be desirable to provide a clearance gap to ease assembly.
FIG. 16 is a graph 1600 showing the second part distribution divided into groups. The second part distribution can be divided into two or more groups. In one embodiment, the groups can be separated in accordance with non-overlapping sub-ranges of first and second attribute values of the second part. Although the second part is divided into groups in this example, in other embodiments the first part may be divided into groups. In some embodiments, a part having the least variance (with respect to other parts) may be selected to be divided into groups.
In one embodiment, groups may be determined by dividing attribute values into substantially even ranges. As shown in FIG. 16, the first attribute value can range from 56.60 units to 56.72 units. This range can be divided into two such that two relatively uniform sized sub-ranges of the first attribute can be formed of 0.06 units each. Similarly for the second attribute value, the range from 10.57 to 10.69 units can be divided into two relatively uniformly sized sub-ranges of 0.06 units each. Thus, in this example, a first group 1602 can have a first attribute range from 56.60 to 56.66 and a second attribute range from 10.63 to 10.69. Second group 1604, third group 1606 and fourth group 1608 can be similarly formed as illustrated in FIG. 16.
In other embodiments, the resulting groups need not be similarly sized. For example, group sub-ranges can be based upon two or more manufacturing means (averages) of a first attribute value. In another embodiment, the sub-ranges can be determined in accordance with single mean values associated with manufacturing yields. In yet other embodiments, the sub-ranges of adjacent groups may overlap. In one embodiment, as the buffer of N parts (second parts) is initially filled, each second part can be examined and assigned to be a member of one of the defined groups. Continuing the example began in FIG. 14A-14C, as second parts are included in the buffer of N parts, each second part can be assigned to one of the four groups 1602-1608. This further characterization of second parts can allow a further buffer management for a continuous selection method. This buffer management can be illustrated in FIGS. 17A-17B below.
FIGS. 17A-17B is a flow chart of method steps 1700 for a continuous selection manufacturing process, in accordance with one embodiment described in the specification. Those skilled in the art will recognize that any system configured to perform the method steps in any order is within the scope of this description. The method can begin in step 1702 when first and second attribute acceptable values are determined For example, a first attribute can be height and a second attribute can be width. In determining acceptable attribute values, a first part can have a first range of acceptable attribute values, while a second part can have a second a different range of acceptable attribute values. Referring back to FIG. 15, an acceptable range of first attribute values can be from 56.63 units to 56.75 for the first part, while an acceptable range of first attribute values for the second part can be from 56.60 units to 56.72.
In step 1704, the second parts can be divided into groups. In one embodiment, the groups can be determined by non-overlapping sub-ranges of acceptable values of a part attribute. For example, the second part can be divided into four groups as shown in FIG. 16. In step 1706, a buffer of N second parts can be formed by noting the group of the second part as the part is placed into the buffer.
In step 1708, the first part can be received. In step 1710, the first and second attribute values for the first part can be determined In step 1712, the first and second attribute values of the first part are examined to determine if the part is acceptable. For example, the first part can be scanned (step 1710) and scanned first and second attribute values compared against acceptable attribute values (step 1712) to determine if the part is acceptable. If the part is not acceptable, then in step 1714, the part is examined to determine if the first part is a recycle candidate. A recycle candidate part can be a part that can possibly have acceptable part attribute values, but the buffer may not currently have a second part that may be acceptable. If the first part is not a recycle candidate, then in step 1716, the part is rejected and the method ends. In one embodiment, the rejected parts determined in step 1716 can be considered not matched parts. These can be first parts that can be manufactured out of specification and therefore may not have any acceptable second parts.
If, on the other hand in step 1714, the part is a recycle candidate, then in step 1720 the first part is recycled and the method ends. In one embodiment, the first part can be placed back into an inventory of other first parts and the continuous selection process can be applied later by reselecting this part from inventor and applying the described process again, when the make-up of the buffer of N second parts has changed.
Returning to step 1712, if the first part is acceptable, then in step 1718, the buffer of N second parts is searched for an acceptable second part. In one embodiment, an acceptable part can be one in which an acceptable predetermined difference exists between the respective attribute values on the first and second parts. In another embodiment, an acceptable part can have attribute values that can provide an acceptable function of merit for first and second attribute values. For example, a function of merit can be a difference between like attribute values on the first and the second part. In another embodiment, an acceptable value for a function of merit can be a function of merit that is less than a predefined value. For example, in one embodiment, an acceptable function of merit can be a gap value between the first and the second part that is less than a predetermined amount.
If there is no acceptable part in the part second part buffer, then the method can proceed to step 1720. On the other hand, if there is an acceptable part in the buffer, then in step 1722, the second part can be selected from the part buffer. In one embodiment, the group associated with the selected part can be noted so that in step 1724, the part buffer can be replenished with another second part, selected in accordance with the noted group. By replenishing the buffer with second parts having the same group as the selected part, the make-up of the parts buffer can be controlled. In one embodiment, replacing selected parts is similar selected parts can prevent the parts buffer from becoming skewed with, for example, too many similar second parts. In step 1726, a kit can be formed including the first part and selected second part and the method ends.
In one embodiment, the sub-ranges that are used to define groups as described in FIG. 16 can be changed to help manage the buffer of second parts. For example, one or more sub-ranges can be narrowed or widened to decrease or increase respectively the frequency that a part from a particular group is selected. In this manner, the make-up of the buffer can be rebalanced in concert with incoming first parts having particular attributes.
In other embodiments, both the first and second parts populations can be divided into groups instead of only the second part as described above. Such an embodiment may increase selection complexity; however, improved buffer management may be realized, when compared to only one part population divided into groups.
The various aspects, embodiments, implementations or features of the described embodiments can be used separately or in any combination. Various aspects of the described embodiments can be implemented by software, hardware or a combination of hardware and software. The described embodiments can also be embodied as computer readable code on a non-transitory computer readable medium. The computer readable medium is defined as any data storage device that can store data which can thereafter be read by a computer system. Examples of the computer readable medium include read-only memory, random-access memory, CD-ROMs, DVDs, magnetic tape, and optical data storage devices. The computer readable medium can also be distributed over network-coupled computer systems so that the computer readable code is stored and executed in a distributed fashion.
The foregoing description, for purposes of explanation, used specific nomenclature to provide a thorough understanding of the described embodiments. However, it will be apparent to one skilled in the art that the specific details are not required in order to practice the described embodiments. Thus, the foregoing descriptions of the specific embodiments described herein are presented for purposes of illustration and description. They are not target to be exhaustive or to limit the embodiments to the precise forms disclosed. It will be apparent to one of ordinary skill in the art that many modifications and variations are possible in view of the above teachings.
The advantages of the embodiments described are numerous. Different aspects, embodiments or implementations can yield one or more of the following advantages. Many features and advantages of the present embodiments are apparent from the written description and, thus, it is intended by the appended claims to cover all such features and advantages of the invention. Further, since numerous modifications and changes will readily occur to those skilled in the art, the embodiments should not be limited to the exact construction and operation as illustrated and described. Hence, all suitable modifications and equivalents can be resorted to as falling within the scope of the invention.

Claims

What is claimed is:

1. A manufacturing process, comprising:

determining a first part attribute value, the attribute value used to identify a matching second part from a buffer of N second parts;

dividing the buffer of N second parts into two or more groups of second parts in accordance with predefined sub-ranges of second part attribute values;

using the first part attribute value to identify the matching second part from within one of the two or more groups of second parts when the matching second part includes a second part attribute value is within an acceptable range of second part attribute values;

removing the identified matching second part from the buffer; and

maintaining the buffer of N second parts by replacing the matching second part with a replacement second part having attribute values in accordance with sub-range corresponding to the group from which the selected second part was selected therefrom.

2. The manufacturing process of claim 1, wherein the acceptable range of second part attribute values is based on a function of merit, wherein the function of merit is a difference between measurements of one at least one point on the first attribute of the first part and a corresponding point of the first attribute of the second part.

3. The manufacturing process of claim 1, further comprising:

monitoring a yield of the manufacturing process by determining a ratio of a number of first parts and a number of not matched first parts.

4. The manufacturing process of claim 3, further comprising:

rebalancing the buffer of N second parts by changing predefined sub-ranges of second part attribute values.

5. The manufacturing process of claim 1, wherein the sub-ranges are determined in accordance with mean values associated with manufacturing yields of the second part.

6. The manufacturing process of claim 1, wherein the sub-ranges are determined by dividing the range of attribute values into two or more substantially even amounts.

7. The manufacturing process of claim 1, wherein the first part is rejected when first attribute value is greater than a predetermined amount.

8. The manufacturing process of claim 1, further comprising:

determining a second attribute value of the first part, the second attribute value and the first attribute value used to identify the matching second part from a buffer of N second parts;

wherein the buffer of N second parts is further divided into two or more groups of second parts in accordance with predefined sub-ranges of first and second attribute values, and wherein the second attribute value is used with the first attribute value to identify the matching second part from within one of the two or more groups of second parts when the matching second value is within an acceptable range of first and second attribute values.

9. The manufacturing process of claim 8, further comprising:

rejecting the first part when attribute values of the first and second attributes exceeds a predetermined amount.

10. An inventory control method performed in a continuous selection manufacturing process, comprising:

receiving an incoming first part;

determining a first and a second attribute value of the first part;

storing second parts in two or more groups based on sub-ranges of first and second attribute values, wherein the two or more groups comprise a buffer of N second parts;

selecting a matching second part from the buffer of N second parts when a value of a function of merit associated with the first attribute value and a value of a function of merit associated with the second attribute value are within an acceptable amounts;

removing the selected matching second part from the buffer; and

providing a kit including the first part and the selected second part.

11. The inventory control method as recited in claim 10, further comprising:

rejecting the first part when no matching second part is selected.

12. The inventory control method as recited in claim 11, further comprising:

monitoring a yield of the continuous selection manufacturing process by determining a ratio of a number of incoming first parts and a number of not matched first parts.

13. The inventory control method as recited in claim 12, further comprising:

modifying the sub-ranges of the groups when the yield is less than a predetermined amount.

14. The inventory control method as recited in claim 12, further comprising:

modifying the sub-ranges of the groups in accordance with changing nominal first and second attributes of the first part.

15. The inventory control method as recited in claim 10, further comprising:

recycling the first part when no matching second part is selected.

16. The inventory control method as recited in claim 12, further comprising :

selecting a replacement second part in accordance with the sub-ranges corresponding to the group from which the second part was selected therefrom.

17. The manufacturing process as recited in claim 16, rebalancing the distribution of the N second parts in the buffer when the monitored yield is less than a pre-determined value.

18. Non-transitory computer readable medium executable by a processor in a computer assisted manufacturing process, comprising:

computer code for providing an incoming first part;

computer code for determining a first part attribute value, the attribute value used to identify a matching second part from a buffer of N second parts;

computer code for using the first part attribute value to identify the matching second part from within one of the two or more groups of second parts when the matching second part attribute value is within an acceptable range of second part attribute values;

computer code for removing the identified matching second part from the buffer; and,

computer code for maintaining the buffer of N second parts by replacing the removed matching second part with a replacement second part having attribute values in accordance with sub-range corresponding to the group from which the part was selected therefrom.

19. The computer readable medium as recited in claim 18, further comprising:

computer code for monitoring a yield of the manufacturing process by determining a ratio of a number of first parts and a number of not matched first parts.

20. The computer readable medium as recited in claim 19, further comprising:

computer code for rebalancing the buffer of N second parts by changing predefined sub-ranges of second part attribute values.

21. The computer readable medium as recited in claim 18, further comprising:

computer code for rejecting the first part when attribute values of the first and attribute exceeds a predetermined amount.