CN106156910A - Quality management system and method thereof - Google Patents

Abstract

A quality management system and method thereof. The quality management method comprises the steps of obtaining quality historical data of a production batch from a database, classifying the production batch into N abnormal event sets according to the quality historical data of the production batch, corresponding each abnormal event of a plurality of abnormal events in each abnormal event set to a proper abnormal event grade according to the quality historical data of the production batch, setting a risk coefficient to be estimated corresponding to each abnormal event grade, calculating the risk coefficient to be estimated corresponding to each abnormal event grade according to the quality historical data of the production batch and the abnormal event grades, generating a linear regression equation according to the abnormal event grades and the calculated risk coefficients, and predicting the risk score corresponding to a high-reliability interval of the production batch according to the linear regression equation.

Description

Quality control system and method thereof

Technical field

The present invention relates to a kind of quality control system and method thereof, particularly relate to a kind of be applied to produce batch Quality control method.

Background technology

In recent years, due to the prosperity of industrial technology, the means that batch production manufactures are by the system of tradition manpower The machine that method of making gradually is automated is replaced.It is said that in general, produce batch (production lot) Utilize production line manufacture process can through many websites, and each website can perform correspondence step and Production routine.But, in actual production process, usually have the generation of substandard products, along with machine work The time made increases, it may happen that machine deviates the phenomenon of original start in production process.Therefore produce Finished product, its specification may be undesirable.And the increase of total cost of inventory during the generation of substandard products, can be caused, And affect the fluency of production line.Therefore, in order to keyholed back plate produces the quality characteristic and yield rate criticized, usually By producing batch sampling and the data statistics in quality and analysis can be done.

It is said that in general, many anomalous events (Issue) affecting quality can occur when batch production manufactures, And the degree that various anomalous event affects quality characteristic all differs, anomalous event such as system mistake, Shake, temperature too high, even board deadlock etc..And these anomalous events will be by corresponding system board Institute's record.But, because the difference of system board, the standard of institute's recording exceptional event is the most inconsistent, this Planting skimble-scamble data will cause analysis to be difficult to.Furthermore, owing to each system board is not yet integrated, therefore When the data of utilization production batch sampling and anomalous event calculate in the lump, the result analyzed will have inaccurate Problem.

Therefore, it is very important for developing a kind of quality control method integrated by each system board.

Summary of the invention

One embodiment of the invention describes a kind of quality control method, comprises by obtaining production batch in data base Quality history data, according to produce batch quality history data, by productions criticize classifying as N number of abnormal thing Part set, according to producing the quality history data criticized, by multiple anomalous events in each anomalous event set Each anomalous event corresponding to suitable anomalous event grade, set corresponding to each anomalous event grade Risk factor to be estimated, according to produce batch quality history data and these anomalous event grades, meter Calculate the risk factor to be estimated that each anomalous event grade is corresponding, according to these anomalous event grade and meter A little risk factors after calculation, produce linear regression equations, according to linear regression equations, it was predicted that produce Criticizing the risk score corresponding to high-reliability interval of quality characteristic, wherein N is the positive integer more than 1.

Another embodiment of the present invention describes a kind of quality control system, comprises data base and processor.Number Being to store to produce the quality history data criticized according to storehouse, processor is coupled to this data base.Wherein locate Reason device is produced, by acquirement in data base, the quality history data criticized, and processor is according to producing the quality history criticized Data, classify as N number of anomalous event set by production batch, and processor is according to producing the quality history number criticized According to, by corresponding for each anomalous event of multiple anomalous events in each anomalous event set to the most abnormal Event class, processor sets the risk factor to be estimated corresponding to each anomalous event grade, processes Device, according to producing the quality history data and these anomalous event grades criticized, calculates each anomalous event grade Corresponding risk factor to be estimated, processor is according to these wind after these anomalous event grades and calculating Danger coefficient, produces linear regression equations, and processor is according to linear regression equations, it was predicted that produce batch product The risk score corresponding to high-reliability interval of matter characteristic, wherein N is the positive integer more than 1.

Accompanying drawing explanation

Fig. 1 is the element block diagram of the quality control system of inventive embodiments.

Fig. 2 describes in Fig. 1 embodiment, and processor is integrated each platform systems and produced the schematic diagram of batch data.

Fig. 3 is in Fig. 1 embodiment, produces quality characteristic and the datum line of risk score, the upper limit and lower limit Schematic diagram.

[description of reference numerals]

100 quality control systems

10 data bases

11 processors

D produces batch data

O₁To O_NAnomalous event set

I₁₁To I_NMNAnomalous event grade

B₁₁To B_NMNRisk factor

The UB upper limit

BL datum line

LB lower limit

Q_LQuality characteristic

RV_LAnd RV_RRisk score

Detailed description of the invention

Fig. 1 is the element block diagram of the quality control system 100 of the embodiment of the present invention.As it is shown in figure 1, product Matter management system 100 comprises data base 10 and processor 11.Data base 10 is in order to store production batch Quality history data.Quality history data referred herein are criticized after many platform systems, respectively for producing The data of the various different anomalous events (Issue) that platform systems is recorded.Processor 11 is coupled to number According to storehouse 10.Here the processor 11 of indication can be the processor on personal computer, in Analysis server Processor or work table on processor etc..In the present embodiment, processor 11 will utilize number According to the quality history data of the production batch in storehouse 10, algorithm is utilized to integrate the number in all platform systems According to, and analyze its statistical property and regression curve.And processor 11 is according to the result after analyzing, will be pre- Survey the risk factor corresponding to high-reliability interval producing batch quality characteristic.And the work people on production line Member i.e. can utilize this to produce the risk factor corresponding to high-reliability interval of batch quality characteristic, selects conjunction Suitable production is criticized sampling and is tested.And how quality control system 100 will predict production batch quality characteristic The risk factor corresponding to high-reliability interval, its algorithm will be described under.

Fig. 2 describes in Fig. 1 embodiment, and processor 11 is integrated each platform systems and produced showing of batch data It is intended to.In fig. 2, produce batch data D and criticize after many platform systems for producing, each platform systems The data acquisition system of the various different anomalous events (Issue) recorded, these data are stored in data base 10. And processor 11 is by extracting production batch data D in data base 10 after, according to its platform systems, will produce Batch data D classifies as N number of anomalous event set, such as the anomalous event set O in Fig. 2₁To anomalous event Set O_N.The corresponding different platform systems of each anomalous event set.In fig. 2, anomalous event collection Close O₁It is the record data for the 1st platform systems, anomalous event set O₂It is for the 2nd platform systems Record data, anomalous event set O_NIt is the record data for n-th platform systems.In this enforcement In example, each anomalous event set has the anomalous event of identical or different quantity.Such as anomalous event Set O₁In have M₁Individual anomalous event, anomalous event set O₂In have M₂Individual anomalous event, anomalous event Set O_NIn have M_NIndividual anomalous event.It is the positive integer more than 1 used herein of N, and M₁To M_NFor Positive integer.It follows that processor 11 can by the anomalous event correspondence in each anomalous event set extremely Suitably anomalous event grade (Issue Grade).Such as by the anomalous event set O in Fig. 2₁In M₁The ascending sequence of individual anomalous event, and respectively the anomalous event after these sequences is corresponded to abnormal thing Part grade I₁₁To anomalous event grade I_1M1, and anomalous event grade I₁₁For less serious anomalous event, Anomalous event grade I_1M1For more serious anomalous event.Processor 11 can be by anomalous event set O₂In M₂The ascending sequence of individual anomalous event, and respectively the anomalous event after these sequences is corresponded to abnormal thing Part grade I₂₁To anomalous event grade I_2M2, and anomalous event grade I₂₁For less serious anomalous event, Anomalous event grade I_2M2For more serious anomalous event.Similarly, processor 11 can be by anomalous event collection Close O_NIn M_NThe ascending sequence of individual anomalous event, and respectively by corresponding for the anomalous event after these sequences For anomalous event grade I_N1To anomalous event grade I_NMN, and anomalous event grade I_N1For less serious different Ordinary affair part, anomalous event grade I_NMNFor more serious anomalous event.Therefore, according to N number of anomalous event The anomalous event grade that set is corresponding, can define an anomalous event index (Issue Code), for Under:

$\mathrm{IC}=\underset{i=1}{\overset{N}{\mathrm{\Σ}}}\underset{j=1}{\overset{M}{\mathrm{\Σ}}}{I}_{\mathrm{ij}}---\left(1\right)$

In (1) formula, as i-th anomalous event set O_iInterior jth anomalous event grade I_ijIt is 0 Or the when of 1, anomalous event index IC in (1) formula i.e. represents that the anomalous event linked is closed in production batch Number.For example, production batch has suffered from the 1st anomalous event set O₁The 2nd abnormal thing Part I₁₂, the 2nd anomalous event set O₂The 3rd anomalous event I₂₃And the 3rd anomalous event collection Close O₃The 1st anomalous event I₃₁, then the numerical value of anomalous event index IC is I₁₂+I₂₃+I₃₁。

But, observe anomalous event index IC defined in (1) formula and only can know the index of anomalous event And the number of anomalous event, and cannot to obtain the risk of anomalous event in the way of quantization tight (Index) Principal characteristic.Therefore, in order to analyze the anomalous event shock extent for production risk, processor 11 further Can be by i-th anomalous event set O_iInterior jth anomalous event grade I_ij(imply that each abnormal thing Part) set corresponding risk factor B_ij(Risk Coefficient).As in figure 2 it is shown, the 1st exception Event sets O₁Anomalous event grade I₁₁To anomalous event grade I_1M1Corresponding risk factor is B₁₁Extremely B_1M1, anomalous event grade I of the 2nd anomalous event set O2₂₁To anomalous event grade I_2M2Corresponding Risk factor is B₂₁To B_2M2, n-th anomalous event set O_NAnomalous event grade I_N1To abnormal thing Part grade I_NMNCorresponding risk factor is B_N1To B_NMN.In the present embodiment, due in same exception The severity of the anomalous event in event sets is that therefore corresponding risk factor is served as reasons to longer spread by little Big to minispread, imply that risk factor is the biggest and represent that corresponding anomalous event is the lightest to the impact degree of quality Micro-, risk factor is the least represents that corresponding anomalous event is the most serious to the impact degree of quality.Therefore at Fig. 2 In, B₁₁>B₁₂…>B_1M1, B₂₁>B₂₂…>B_2M2, B_N1>B_N2…>B_NMN.Therefore, according to N number of anomalous event The risk factor of the anomalous event grade that set is corresponding, can define a risk score index (Risk Value), under:

$\mathrm{RV}=\underset{i=1}{\overset{N}{\mathrm{\Σ}}}\underset{j=1}{\overset{M_i}{\mathrm{\Σ}}}{B}_{\mathrm{ij}}{I}_{\mathrm{ij}}---\left(2\right)$

But, seriousness and the correspondence thereof of present invention anomalous event in same anomalous event set are different The grade of ordinary affair part is not limited to the ascending sequence described in embodiment, and corresponding to anomalous event grade Risk factor be also not limited to descending arrangement.For example, in other embodiments, abnormal thing Coefficient corresponding to part grade can be to longer spread by little.And in the ensuing explanation of the present embodiment, can profit By the relational expression of a matrix, release the risk system of anomalous event grade corresponding to all anomalous event set Number, then takes back in (2) formula by the risk factor calculated so that (2) formula becomes one and has The linear regression equations of individual parameter, whereinRepresent the anomalous event sum of all considerations, and line Property regression equation be that monotonicity is incremented by (Monotonically Increasing) or monotonicity and successively decreases (Monotonically Decreasing), and processor 11 will obtain according to linear regression equations About quality characteristic baseline (Base Line), the upper limit (Upper Bound) and lower limit to risk factor (Lower Bound).Processor 11 can be according to these statistical results above-mentioned, it was predicted that produce batch quality characteristic The risk score corresponding to high-reliability interval.

Here illustrate that processor 11 is integrated each platform systems and produced batch data and prediction with an example Produce the flow process of the risk factor corresponding to high-reliability interval of batch quality characteristic.Here, in order to simplify Describing, data base 10 considers the data of 2 platform systems records, as shown in the table:

In upper table, two platform systems record the anomalous event set of 2 kinds respectively, and each is different Often event sets has 3 anomalous events, and the 1st anomalous event of first category is the thing without quality loss Part, therefore its seriousness is relatively low, and the 2nd anomalous event of first category is that quality loss is below 10% Event, its seriousness is medium, and the 3rd anomalous event of first category is that quality loss is more than 10% Event, its seriousness is the highest.1st anomalous event of second category is the event without quality loss, Therefore its seriousness is relatively low, and the 2nd anomalous event of second category is quality loss thing below 5% Part, its seriousness is medium, and the 3rd anomalous event of second category is quality loss event more than 5%, Its seriousness is the highest.Quality characteristic here can be the standard of any production quality, such as yield rate (Yield) etc..

It follows that each anomalous event in each anomalous event set can be set by processor 11 Corresponding risk factor.Therefore, the 1st anomalous event of first category is set corresponding risk factor and is B₁₁, it is B that the 2nd anomalous event of first category is set corresponding risk factor₁₂, the 3rd of first category It is B that individual anomalous event is set corresponding risk factor₁₃, it is right that the 1st anomalous event of second category is set Answering risk factor is B₂₁, it is B that the 2nd anomalous event of second category is set corresponding risk factor₂₂, It is B that 3rd anomalous event of second category is set corresponding risk factor₂₃.In order to release all abnormal things The risk factor of the anomalous event grade that part set is corresponding, first can be according to 2 platforms in data base 10 The data of system record, set up an anomalous event ranking matrix.This anomalous event ranking matrix is one Individual sparse matrix (Sparse Matrix), and each element of this anomalous event ranking matrix is 0 or 1, And the definition of anomalous event ranking matrix is that producing each product criticized is classified as first category What all anomalous events of anomalous event set and the anomalous event set of second category combined presents.Citing For, if producing the product quantity criticized in embodiment is N_LOT.Representing for convenience, embodiment will produce Product in Pi is expressed as numbering 1 to N_LOTThe product of index value, then anomalous event ranking matrix can represent Under for:

Upper table is in the anomalous event set of anomalous event set and the second category considering first category, Producing batch middle quantity is N_LOTThe combination of anomalous event kind that met with of product.For the present embodiment, Because the anomalous event of each classification has three kinds, therefore produce batch anomalous event kind suffered from i.e. There is 3 × 3=9 kind possible.Therefore, if producing batch quantity N of middle product_LOT> 9, then it represents that in production batch The product having some different index value can be subjected to identical anomalous event (with the life that general yield rate is higher Producing for criticizing, quality characteristic is higher at the two the most break-even ratios of kind).Such as in upper table, rope The product drawing (1), index (2) and index (6) is just subjected to identical anomalous event (quality characteristic is lossless Lose).Therefore, the anomalous event ranking matrix set up, the dimension of its row is N_LOT.Furthermore, with For the present embodiment, a total of 6 kinds of anomalous events are considered, the anomalous event grade hence set up Matrix, the dimension of its column be 3+3=6 (but in formula, if considering N number of anomalous event set, M is had in assuming the n-th anomalous event set_nThe anomalous event of individual quantity, then anomalous event ranking matrix its The dimension of column isWherein ∑ is for even putting in marks).And by above-mentioned anomalous event grade Matrix launches to be represented by down:

$I_M=\left[\begin{array}{cccccc}1& 0& 0& 1& 0& 0\\ 1& 0& 0& 1& 0& 0\\ 0& 1& 0& 0& 0& 1\\ 1& 0& 0& 0& 1& 0\\ 0& 0& 1& 0& 0& 1\\ 1& 0& 0& 1& 0& 0\\ 0& 0& 1& 1& 0& 0\\ .& .& .& .& .& .\\ .& .& .& .& .& .\\ .& .& .& .& .& .\\ 0& 1& 0& 0& 1& 1\end{array}\right]---\left(3\right)$

Wherein I_MRepresenting anomalous event ranking matrix, its dimension is N_LOT×6。

By anomalous event ranking matrix I_MAfter foundation completes, according to the relation in (2) formula, each risk Coefficient can be corresponding to each anomalous event.Therefore, by the risk in the anomalous event set of first category Coefficient B₁₁, risk factor B₁₂, risk factor B₁₃, and by the anomalous event set of second category Risk factor B₂₁, risk factor B₂₂, risk factor B₂₃Launch, and produce a risk factor vector. In the present embodiment, i.e. there are 6 elements in risk factor vector (but in formula, if considering N number of Anomalous event set, it is assumed that have M in the n-th anomalous event set_nThe anomalous event of individual quantity, then risk I.e. have in coefficient vectorIndividual element, wherein ∑ is for even putting in marks), this risk factor vector can It is expressed as down:

$B=\left[\begin{array}{c}{B}_{11}\\ B_{12}\\ B_{13}\\ B_{14}\\ B_{15}\\ B_{16}\end{array}\right]---\left(4\right)$

Wherein B is risk factor vector, and its dimension is 6 × 1.Therefore, when by the abnormal thing in (3) formula Part ranking matrix I_MWhen being multiplied with the risk factor vector B in (4) formula, the vectorial Y of output is production At the risk factor indicator vector corresponding to the combination of each anomalous event in Pi.But, the present embodiment is raw The product quantity produced in criticizing is N_LOT, risk factor index represented in the vectorial Y of this output is N_LOT The dimension of × 1, shown in lower:

$Y=\left[\begin{array}{c}{Y}_1\\ Y_2\\ Y_3\\ Y_4\\ Y_5\\ Y_6\\ Y_7\\ .\\ .\\ .\\ Y_{N_{\mathrm{LOT}}}\end{array}\right]=\left[\begin{array}{cccccc}1& 0& 0& 1& 0& 0\\ 1& 0& 0& 1& 0& 0\\ 0& 1& 0& 0& 0& 1\\ 1& 0& 0& 0& 1& 0\\ 0& 0& 1& 0& 0& 1\\ 1& 0& 0& 1& 0& 0\\ 0& 0& 1& 1& 0& 0\\ .& .& .& .& .& .\\ .& .& .& .& .& .\\ .& .& .& .& .& .\\ 0& 1& 0& 0& 1& 1\end{array}\right]\×\left[\begin{array}{c}{B}_{11}\\ B_{12}\\ B_{13}\\ B_{21}\\ B_{22}\\ B_{23}\end{array}\right]+\left[\begin{array}{c}{\mathrm{\ϵ}}_1\\ {\mathrm{\ϵ}}_2\\ {\mathrm{\ϵ}}_3\\ {\mathrm{\ϵ}}_4\\ {\mathrm{\ϵ}}_5\\ {\mathrm{\ϵ}}_6\\ {\mathrm{\ϵ}}_7\\ .\\ .\\ .\\ {\mathrm{\ϵ}}_{N_{\mathrm{LOT}}}\end{array}\right]---\left(5\right)$

In (5) formula, B₁₁To B₂₃The number of 2 platform systems records of data base 10 internal memory can be utilized Anomalous event ranking matrix I according to this and in (3) formula_MEstimate its numerical value.Due to anomalous event ranking matrix I_MIt is not square formation, therefore can not directly do matrix inversion (Matrix Inverse).In the present embodiment, Obtain B₁₁To B₂₃Mode may utilize anomalous event ranking matrix I_MVirtual reversal process (Pseudo Inverse) or use linear regression mode, by Least squares approximation method (Least-Squared Or minimum approach error method (Minimum-Mean-Squared Error Approach) is forced Approach) Closely obtain and (in (5) formula, now need one group of micro-margin of error ε₁ExtremelyAssist and solve).Although this enforcement Example formula solves B with the framework of (5) formula₁₁To B₂₃, but the present invention is not limited to this, in other embodiments, Can also be by anomalous event ranking matrix I_MIts dimension is expanded with a constant row vector (Column Vector) Degree, or risk factor vector B is expanded its dimension with a constant, as follows:

In (6) formula, C is a constant, corresponding I_MThe row vector of the leftmost side is also constant vector, B in this case₁₁To B₂₃Still can use Least squares approximation method (Least-Squared Approach) Or minimum approach error method (Minimum-Mean-Squared Error Approach) solves.And ask Solution in Xie Yuyou (5) formula gone out only one constant offset of gap (Constant Offset), the most not The statistical property of the post analysis that can affect.

When above-mentioned (5) formula has been set up and via (5) formula by risk factor B₁₁To B₂₃After obtaining, wind Danger coefficient B₁₁To B₂₃I.e. take back the risk score index in (2) formula, it is possible to (2) formula is become a tool Having the linear regression equations of 6 parameters, wherein the quantity of parameter i.e. represents the anomalous event of all considerations Sum is (in formula, if considering N number of anomalous event set, it is assumed that have in the n-th anomalous event set The anomalous event of Mn quantity, then (2) formula i.e. becomes one and hasThe linear regression side of individual parameter Formula).Here, linear regression equations is monotonicity is incremented by (Monotonically Increasing) Or monotonicity successively decreases (Monotonically Decreasing).For example, the linear regression side obtained Formula LY is deployable as follows:

LY=91.378+0.0 × I₁₁-3.52×I₁₂-8.61×I₁₃+0.0×I₂₁-5.07×I₂₂

-32.32×I₂₃

(7)

In the present embodiment, by (5) formula, risk factor is solved one by one, and disaggregation is combined into B₁₁=0.0, B₁₂=-3.52, B₁₃=-8.61, B₂₁=0.0, B₂₂=-5.07, B₂₃=-32.32.But, the wind of the present invention Danger coefficient disaggregation credit union is along with producing that batch quality characteristic is different and difference, and above-mentioned listed solution set is only For the expression solving set a kind of in embodiment, and it is not used to limit solution procedure and the result of the present invention.This Risk factor solution calculated by embodiment also should demonstrate,prove aforementioned condition, because the present embodiment is integrating not homology During the anomalous event set that system is recorded, processor 11 can be by the exception in each anomalous event set The ascending sequence of event, therefore corresponding its size of risk factor sequence also should meet B₁₁>B₁₂>B₁₃ And B₂₁>B₂₂>B₂₃Condition.And in (7) formula, 91.378 is a predetermined constant, the most not The statistical property of the post analysis of impact.

After the linear regression equations LY in (7) formula obtains, examined and determine by risk score model, can Produce datum line (Base Line) and to should the upper limit (Upper Bound) of datum line and lower limit (Lower Bound) result, as shown in Figure 3.Fig. 3 is that the present invention produces corresponding quality characteristic and the base of risk score The schematic diagram of directrix, the upper limit and lower limit.For example, when consider new production criticize add fashionable, corresponding new Produce batch anomalous event index (Issue Code) and anomalous event grade (Issue Grade) also can It is brought in (7) formula the output producing different linear regression equations LY.And the qualitative control of the present invention System i.e. can produce such as Fig. 3 according to this in risk score and the statistics of corresponding datum line, the upper limit and lower limit etc. Characteristic.In figure 3, X-axis is risk score, and Y-axis is quality characteristic, and upper limit UB is for having circular mark The line segment of point, datum line BL is the line segment with triangle punctuate, and lower limit LB is to have rectangle punctuate Line segment.Criticize when engineering and at least require that quality characteristic meets Q_LStandard time, the risk of corresponding upper limit UB is divided Number is RVL, and the meaning representated by this risk score RVL is and meets quality characteristic Q_LStandard time, work Journey criticizes the lower limit of selected risk score.Therefore, the reliability of sampling, processor are criticized in order to increase engineering 11 meetings calculate with the angle of datum line BL and meet quality characteristic Q_LStandard time, corresponding risk score. In that case, owing to datum line BL is corresponding to quality characteristic Q_LRisk score be RVR, Gu Chu Reason device 11 will be advised that the engineering staff on production line selects and be met the risk factor of datum line BL at least The engineering of RVR criticizes sampling.

But, the quality characteristic described in Fig. 3 and the datum line of risk score, the upper limit and lower limit, its slope It is not limiting as the algorithm of the present invention.For example, if the quality characteristic in embodiment is for considering yield rate (Yield), then the slope of datum line, the upper limit and lower limit will be same as Fig. 3, and its slope is just.Otherwise, In other embodiments, if quality characteristic for consider defect rate (Defective Rate), then datum line, on The slope of limit and lower limit will be contrary with Fig. 3, and its slope is negative.

In sum, the present invention describes a kind of being applied to and produces the quality control system criticized and method thereof, Its idea is the record data integrated and produce batch different platform system, and by the data pair of these anomalous events Should be to suitable anomalous event grade and risk factor.By setting up anomalous event ranking matrix and risk Coefficient vector, can obtain all of risk factor by the way of recurrence, and then try to achieve risk score The linear regression equations of index, the high-reliability interval institute that one-step prediction of going forward side by side produces batch quality characteristic is right The risk score answered.Therefore, the engineering staff on production line utilizes the risk score of suggestion easily, Sampling is criticized in the production picking out reliability higher.

The foregoing is only the preferred embodiments of the present invention, all equivalents done according to the claims in the present invention become Change and modify, all should belong to the covering scope of the present invention.

Claims (12)

1. a quality control method, comprises:

By the quality history data obtaining a production batch in a data base；

The quality history data criticized according to this production, criticize this production and classify as N number of anomalous event set；

The quality history data criticized according to this production, by multiple anomalous events in each anomalous event set Each anomalous event is corresponding to suitable anomalous event grade；

Set the risk factor to be estimated corresponding to this each anomalous event grade；

The quality history data criticized according to this production and these anomalous event grades, calculate this each abnormal thing This risk factor to be estimated that part grade is corresponding；

According to these risk factors after these anomalous event grades and calculating, produce an equation of linear regression Formula；And

According to this linear regression equations, it was predicted that the high-reliability interval institute that quality characteristic is criticized in this production is right The risk score answered；

Wherein N is greater than the positive integer of 1.

2. the method for claim 1, also comprises:

Each anomalous event grade in each anomalous event set is added up total to produce a quality grade With.

3. the method for claim 1, wherein according to this linear regression equations, it was predicted that this is raw Produce this risk score corresponding to high-reliability interval criticized, comprise:

According to this linear regression equations, produce a datum line and to should the upper limit of datum line and lower limit Result；And

Utilize this high-reliability interval institute that this datum line, this upper limit and this this production of lower limit prediction of result are criticized Corresponding risk score.

4. the method for claim 1, wherein by the plurality of different in this each anomalous event set Each anomalous event of ordinary affair part is corresponding to suitable anomalous event grade, for by the n-th anomalous event collection In closing, quantity it is M_nThe corresponding suitably anomalous event grade of each anomalous event in individual anomalous event, N is the positive integer between 1 and N, and M_nIt it is positive integer.

5. the method for claim 1, also comprises:

According to these anomalous event grades, produce an anomalous event ranking matrix；And

According to the risk factor that these are to be estimated, produce a risk factor vector to be estimated.

6. method as claimed in claim 5, the quality history data wherein criticized according to this production and this A little anomalous event grades, calculate this risk factor to be estimated of this each anomalous event grade, are bases Quality history data that this production is criticized and this anomalous event ranking matrix, calculate this risk factor to be estimated Each risk factor in vector.

7. method as claimed in claim 5, wherein according to these anomalous event grades, produces this different Ordinary affair part ranking matrix, according to these anomalous event grades, generation dimension is N_LOTThe abnormal thing of × S Part ranking matrix, wherein N_LOTIt is the product quantity during this production is criticized, and S is

8. method as claimed in claim 5, wherein according to these risk factors to be estimated, produces One risk factor vector to be estimated, is basisIndividual risk factor to be estimated, produces dimension ForThis risk factor to be estimated vector.

9. method as claimed in claim 5, wherein according to these anomalous event grades and after calculating These risk factors, produce this linear regression equations, after utilizing these anomalous event grades and calculating These risk factors, generation hasThe linear regression equations of individual parameter, and this linear regression Equation is that monotonicity is incremented by or monotonicity is successively decreased.

10. method as claimed in claim 5, wherein this anomalous event ranking matrix is a sparse matrix, And each element of this anomalous event ranking matrix is 0 or 1.

11. methods as claimed in claim 5, the quality history data wherein criticized according to this production and should Anomalous event ranking matrix, calculates each risk factor in this risk factor vector to be estimated, is root The quality history data criticized according to this production and this anomalous event ranking matrix, utilize Least squares approximation method meter Calculate each risk factor in this risk factor vector to be estimated.

12. 1 kinds of quality control systems, comprise:

One data base, in order to store the quality history data that a production is criticized；And

One processor, is coupled to this data base；

Wherein this processor is by obtaining the quality history data that this production is criticized in this data base, and this processor depends on The quality history data criticized according to this production, criticize this production and classify as N number of anomalous event set, this process The quality history data that device is criticized according to this production, by multiple anomalous events every in each anomalous event set The corresponding suitably anomalous event grade of one anomalous event, this processor sets corresponding to this each anomalous event One risk factor to be estimated of grade, quality history data that this processor is criticized according to this production and these Anomalous event grade, calculates this risk factor to be estimated that this each anomalous event grade is corresponding, at this Reason device, according to these risk factors after these anomalous event grades and calculating, produces an equation of linear regression Formula, this processor is according to this linear regression equations, it was predicted that a high-reliability of quality characteristic is criticized in this production Risk score corresponding to interval, wherein N is greater than the positive integer of 1.