Pass/fail judgment device, pass/fail judgment program, pass/fail judgment method, and multivariate statistics analyzer

A quality control apparatus which takes the form of pass/fail objects as a pass/fail judgment factor and which is used to detect a defective unit in product inspection, comprising:

• a statistical computing unit configured to receive object data representing one or more characteristics of an object and compute at least one of a first probability that the object will be classified in a first category and a second probability that the object will be classified in a second category based on the received object data wherein frequency distributions of the first probability and the second probability has a shape of a normal distribution;
• an input unit configured to receive at least one of a rate of flowout in the second category, which represents a number of objects that actually are in the second category that are actually classified as being in the first category, and a rate of overcontrol in the first category, which represents a number of objects that actually are in the first category that are judged as being in the second category;
• a calculation unit configured to calculate a discriminate function to discriminate between one or more objects classified in the first category from one or more objects classified in the second category based on the at least one of the rate of flowout in the second category and the rate of overcontrol in the first category received by the input unit and based on at least one of the first and second probabilities computed by the statistical computing unit;
• a judging unit configured to determine whether one or more objects should be classified in one of the first and second categories based on the discriminate function calculated by the calculation unit;
• and
• a communication unit configured to communicate whether the object is classified in one of the first and second categories based on the determination of the judging unit;
• wherein the judging unit has a standard deviation based on a normal distribution of the objects,
• wherein the overcontrol and flowout are separated having the normal distribution;
• and
• wherein the overcontrol and flowout are judged based on the data of the normal distribution.

A computer-implemented quality control method taking the form of pass/fail objects as a pass/fail judgment factor for detecting a defective unit in product inspection, comprising:

• receiving object data representing one or more physical characteristics of an object;
• computing with a microprocessor a first probability distribution, which represents a probability that one or more objects should be classified in a first category, based on the received object data;
• computing with the microprocessor a second probability distribution, which represents a probability that one or more objects should be classified in a second category, based on the received object data;
• wherein frequency distributions of the first category and the second category has a shape of a normal distribution,
• computing a center of distribution and distribution parameters that vary in accordance with a breadth of distribution for said variables with respect to either or both of said first category or said second category;
• receiving a rate of flowout in the second category, which represents a number of objects that should be classified in the second category, but are actually judged as being classified in the first category relative to said center of distribution and said distribution parameters;
• receiving a rate of overcontrol in the first category representing a number of objects that should be classified in the first category, but are actually judged as being classified in the second category relative to said center of distribution and said distribution parameters;
• calculating a discriminate function to discriminate between one or more objects that should classified in the first category from one or more objects that should be classified in the second category based on the first and second probability distributions computed by the statistical computing unit and based on at least one of the received rate of flowout and the received rate of overcontrol;
• determining whether one or more objects should be classified in one of the first and second categories based on the discriminate function calculated by the calculation unit;
• and
• communicating the determination of whether the one or more objects are classified in one of the first and second categories to an operator;
• wherein the rate of overcontrol and flowout are separated having normal distribution;
• and
• wherein the overcontrol and flowout are judged based on the data of the normal distribution.

A multivariate statistics analyzer which is capable of communication of data with the outside through a communication interface and executes a multivariate analysis program under the control of a predetermined operating system, wherein

• said multivariate statistics analyzer comprises a hard disk drive which is capable of accumulating the multivariate analysis program and transmitting, receiving, and accumulating data,
• said multivariate analysis program comprises modules corresponding to a mode classifying portion which includes parameter value data consisting of parameter values which are correlated with at least pass/fail judgment result data of one or more pass/fail judgment objects as a pass/fail judgment factor, and which are used to detect a defective unit in product inspection when the data is externally acquired through said communication interface and stored in said hard disk drive and are actually computed with respect to each defective unit, and subdivides categories based on the accumulated data;
• a discriminant function computing portion which eliminates multicolinearity, gives variables used to separate frequency distributions of a pass category and a fail category from a plurality of pieces of information which make pass/fail judgment factors and pass/fail judgment result information, wherein each of the frequency distributions of the pass category and the fail category has a shape of a normal distribution, and further computes discriminant functions based on said parameter value data;
• a statistical parameter computing portion which computes a center of distribution parameters and that vary in accordance with a breadth of a distribution for variables with respect to either or both of a pass category and a fail category, and the mean and standard deviation in the frequency distributions of said pass category and said fail category with respect to said discriminant functions;
• and
• a threshold determining portion for determining a threshold for providing a pass/fail judgment based on the value of a variable defined by a specific distribution probability based on at least one of a rate of flowout in the fail category, which represents a number of pass/fail judgment objects are actually in the fail category, and that are judged as being passed, and a rate of overcontrol in the pass category, which represents a number of pass/fail judgment objects that are actually in the pass category, and that are actually judged as being failed, relative to said center of distribution and said distribution parameter, and
• the threshold determining portion further performs the operations of acquiring said discriminant function data, said parameter value data, and pass/fail judgment result data, generating a histogram corresponding to a pass/fail judgment result on a category-by-category basis, computing a mean and a standard deviation of each category in the generated histogram, determining the threshold of a discriminant function corresponding to a specified rate of flowout which is set for the fail category and indicates the range in which defective units are let out, based on the mean and standard deviation computed in the fail category and the rate of flowout;
• and
• a pass/fail judgment display portion configured to display a pass/fail judgment for the one or more pass/fail judgment objects based on the threshold determined by the threshold determining portion;
• wherein the overcontrol and flowout are separated having the normal distribution;
• and
• wherein the overcontrol and flowout are judged based on the data of the normal distribution.

A pass/fail judgment method taking the form of pass/fail objects as a pass/fail judgment factor for detecting a defective unit in product inspection, comprising:

• computing with a microprocessor discriminant functions which give variables which separate frequency distributions of a pass category and a fail category from a plurality of pieces of parameter information which make pass/fail judgment factors and pass/fail judgment result information thereof wherein frequency distributions of the pass category and the fail category has a shape of a normal distribution;
• computing with the microprocessor a center of distribution and distribution parameters that vary with a breadth of a distribution for said variables with respect to either or both of said pass category and said fail category;
• determining a threshold for providing a pass/fail judgment based on the value of a variable value and giving a specific distribution probability based on at least one of a rate of flowout in the fail category, which represents a number of pass/fail judgment objects that are actually in the fail category that are judged as being passed, and a rate of overcontrol in the pass category, which represents a number of pass/fail judgment objects that are actually in the pass category that are judged as being failed, relative to said center of distribution and said distribution parameters;
• acquiring a plurality of pieces of said parameter information on one or more pass/fail judgment objects according to data of a normal distribution;
• and
• comparing the value of variables obtained by substituting the parameter information into said discriminant functions with said threshold;
• and
• displaying a pass/fail judgment for the one or more pass/fail judgment objects based on the comparing step;
• wherein the rate of overcontrol and flowout are separated having the normal distribution;
• and
• wherein the pass/fail judgment are judged based on the normal distribution.

A pass/fail judgment method taking the form of pass/fail objects as a pass/fail judgment factor for detecting a defective unit in product inspection, comprising:

• a discriminant function computing step in which discriminant functions which give variables which separate frequency distributions of a pass category and a fail category are computed with a microprocessor from a plurality of pieces of parameter information which make pass/fail judgment factors and pass/fail judgment result information thereof wherein each of the frequency distributions of the pass category and the fail category is the shape of a normal distribution;
• a statistical parameter computing step in which a center of distribution and distribution parameters that vary in accordance with a breadth of a distribution for said variables are computed with the microprocessor with respect to either or both of said pass category and said fail category;
• a threshold determining step for determining a threshold for providing a pass/fail judgment based on a value of a variable representing a specific distribution probability based on at least one of a rate of flowout in the fail category, which represents a number of pass/fail judgment objects that are actually in the fail category, and that are judged as being passed, and a rate of overcontrol in the pass category, which represents a number of pass/fail judgment objects that are actually in the pass category, and that are judged as being failed, relative to said center of distribution and said distribution parameters;
• a parameter information acquiring step in which a plurality of pieces of said parameter information on one or more pass/fail judgment objects are acquired according to data of a normal distribution;
• and
• a pass/fail judging step in which the value of variables obtained by substituting the parameter information into said discriminant functions are compared with said threshold, and a pass/fail judgment for the one or more pass/fail judgment objects is displayed based on the comparison with said threshold;
• wherein the rate of overcontrol and flowout are separated having the normal distribution;
• and
• wherein the pass/fail judging step is judged based on the normal distribution.

A computer-implemented quality control apparatus used to detect a defective unit in a product inspection, comprising:

• a detector configured to detect physical characteristics of an object and generate object data representing the detected physical characteristics of the object;
• a statistical parameter computing unit configured to compute a center of distribution and distribution parameters that vary in accordance with a breadth of distribution for variables with respect to either or both of a non-defective object probability representing a probability that one or more objects should be classified in a non-defective category based on the object data generated by the detector, compute a defective object probability representing a probability that one or more objects should be classified in a defective category based on the object data generated by the detector, classify the one or more objects as being in the non-defective category based on a match between a first pattern of object data and the computed non-defective object probability, and classify the one or more objects as being in the defective category based on a match between a second pattern of object data and the computed defective object probability;
• an input unit configured to receive a rate of flowout in the defective category, which represents a number of objects that are actually in the defective category by the statistical computing unit, and that are judged as being non-defective, and a rate of overcontrol in the non-defective category, which represents a number of objects that are actually in the non-defective category by the statistical computing unit, and that are judged as being defective, the received rate of flowout and the received rate of overcontrol including a visual observation by an operator of an actual orientation of one component of the object relative to another component of the object;
• a calculation unit configured to calculate a discriminate function to discriminate between defective and non-defective objects based on the non-defective category and defective object probability distributions computed by the statistical computing unit and based on the feedback data received by the input unit, the discriminate function being different from a midpoint between a mean value of the first probability distribution and a mean value of the second probability distribution wherein frequency distributions of the non-defective category and the defective category has a shape of a normal distribution;
• a judging unit configured to determine whether the one or more objects should be classified in one of the defective or non-defective categories based on the discriminate function calculated by the calculation unit;
• and
• classifying the one or more objects as being in the defective category based on a match between a second pattern of object data and the computed defective object probability distribution;
• receiving an rate of flowout in the defective category, which represents a number of objects that are actually in the defective category, and that are judged as being non-defective, the received rate of flowout including a visual observation by an operator of an actual orientation of one component of the object relative to another component of the object;
• receiving a rate of overcontrol in the non-defective category, which represents a number of objects that are actually in the non-defective category, and that are actually in the defective category, the received rate of overcontrol including a visual observation by an operator of an actual orientation of one component of the object relative to another component of the object;
• calculating a discriminate function to discriminate between one or more objects classified in the non-defective category from one or more objects classified in the defective category based on at least one of the received rate of flowout and the received rate of overcontrol and based on the computed non-defective and defective object probability distributions, the discriminate function being different from a midpoint between a mean value of the computed non-defective object probability distribution and a mean value of the computed defective object probability distribution wherein frequency distributions of the non-defective category and the defective category has a shape of a normal distribution;
• a display unit configured to display whether the one or more objects are classified in one of the defective or non-defective categories based on the determination of the judging unit;
• wherein the rate of overcontrol and the rate of flowout are separated having the normal distribution;
• and
• wherein the overcontrol and flowout are judged based on the data of the normal distribution.

A computer-implemented quality control method taking the form of pass/fail objects as a pass/fail judgment factor, and used to detect a defective unit in product inspection, comprising:

• detecting physical characteristics of an object;
• generating object data representing the detected physical characteristics of the object;
• computing with a microprocessor a non-defective object probability distribution representing a probability that one or more objects should be classified in a non-defective category based on the generated object data;
• computing with the microprocessor a defective object probability distribution representing a probability that one or more objects should be classified in a defective category based on the generated object data;
• classifying the one or more objects as being in the non-defective category based on a match between a first pattern of object data and the computed non-defective object probability distribution;
• determining whether one or more objects should be classified in one of the defective or non-defective categories based on the calculated discriminate function;
• and
• displaying an image that illustrates whether the one or more objects are classified in one of the defective or non-defective categories based on the determining step;
• wherein the rate of overcontrol and the rate of flowout are separated having normal distribution;
• and
• wherein the overcontrol and flowout are judged based on the data of normal distribution.