Simulation parameter correction technique

1. A computer-readable, non-transitory storage medium storing a program for causing a computer to execute a process comprising:

first calculating, for each of a plurality of manufacturing measurement condition sets stored in a data storage unit, each candidate value of each parameter among a plurality of parameters that are statistics of predetermined product characteristics, based on of relational expressions, wherein each of the relational expressions represents a relationship between a corresponding manufacturing measurement condition set and one parameter of the plurality of parameters, and storing the candidate values of the plurality of parameters, which are calculated for the corresponding manufacturing measurement condition set, into the data storage unit;

obtaining, from a variability-aware simulation, a simulation result value of a predetermined product performance with respect to the candidate values of the plurality of parameters, which are stored in the data storage unit in association with each of the manufacturing measurement condition sets, and storing the simulation result values into the data storage unit;

second calculating a likelihood by substituting the candidate values of the plurality of parameters for the respective manufacturing measurement condition sets, the simulation result values of the respective manufacturing measurement condition sets, the statistics of measurement values of the predetermined product characteristics for the respective manufacturing measurement condition sets and measurement values of the predetermined product performance for the respective manufacturing measurement condition sets into a likelihood function that is defined by unifying, for the plurality of manufacturing measurement condition sets, respective probability density functions for the respective parameters and a probability density function for the predetermined product performance and is a function to calculate a combined likelihood of the respective parameters and the predetermined product performance for the plurality of manufacturing measurement condition sets, and storing the calculated likelihood into the data storage unit; and

searching for respective relational expressions for each of the manufacturing measurement condition sets in case where the likelihood becomes maximum by carrying out the first calculating and storing, the obtaining and storing and the second calculating and storing a plurality of times, while changing the respective relational expressions for each of the manufacturing measurement condition sets.

2. The computer-readable, non-transitory storage medium as set forth in claim 1 , wherein the relational expression includes a sum of a first coefficient and products of corresponding second coefficients and manufacturing measurement condition values included in the manufacturing measurement condition set, and

the searching comprises, for each of the manufacturing measurement condition sets,

calculating, for each of the first coefficient and the second coefficients, a slope of the likelihood at a neighborhood; and

generating a new first coefficient and new second coefficients by modifying the first coefficient and the second coefficients so as to increase the likelihood according to the slope of the likelihood.

3. A method for correcting simulation parameters, the method comprising:

first calculating, by using a computer, for each of a plurality of manufacturing measurement condition sets stored in a data storage unit, each candidate value of each parameter among a plurality of parameters that are statistics of predetermined product characteristics, based on each of relational expressions, wherein each of the relational expressions represents a relationship between a corresponding manufacturing measurement condition set and one parameter of the plurality of parameters, and storing, by using the computer, the candidate values of the plurality of parameters, which are calculated for the corresponding manufacturing measurement condition set, into the data storage unit;

obtaining, by using the computer, from a variability-aware simulation, a simulation result value of a predetermined product performance with respect to the candidate values of the plurality of parameters, which are stored in the data storage unit in association with each of the manufacturing measurement condition sets, and storing, by using the computer, the simulation result values into the data storage unit;

second calculating, by using the computer, a likelihood by substituting the candidate values of the plurality of parameters for the respective manufacturing measurement condition sets, the simulation result values of the respective manufacturing measurement condition sets, the statistics of measurement values of the predetermined product characteristics for the respective manufacturing measurement condition sets and measurement values of the predetermined product performance for the respective manufacturing measurement condition sets into a likelihood function that is defined by unifying, for the plurality of manufacturing measurement condition sets, respective probability density functions for the respective parameters and a probability density function for the predetermined product performance and is a function to calculate a combined likelihood of the respective parameters and the predetermined product performance for the plurality of manufacturing measurement condition sets, and storing, by using the computer, the calculated likelihood into the data storage unit; and

searching, by using the computer, for respective relational expressions for each of the manufacturing measurement condition sets in case where the likelihood becomes maximum by carrying out the first calculating and storing, the obtaining and storing and the second calculating and storing a plurality of times, while changing the respective relational expressions for each of the manufacturing measurement condition sets.

4. A simulation parameter correction apparatus comprising:

a first calculation unit to calculate, for each of a plurality of manufacturing measurement condition sets stored in a data storage unit, each candidate value of each parameter among a plurality of parameters that are statistics of predetermined product characteristics, based on each of relational expressions, wherein each of the relational expressions represents a relationship between a corresponding manufacturing measurement condition set and one parameter of the plurality of parameters, and to store the candidate values of the plurality of parameters, which are calculated for the corresponding manufacturing measurement condition set, into the data storage unit;

an obtaining unit to obtain, from a variability-aware simulation, a simulation result value of a predetermined product performance with respect to the candidate values of the plurality of parameters, which are stored in the data storage unit in association with each of the manufacturing measurement condition sets, and to store the simulation result values into the data storage unit;

a likelihood calculation unit to calculate a likelihood by substituting the candidate values of the plurality of parameters for the respective manufacturing measurement condition sets, the simulation result values of the respective manufacturing measurement condition sets, the statistics of measurement values of the predetermined product characteristics for the respective manufacturing measurement condition sets and measurement values of the predetermined product performance for the respective manufacturing measurement condition sets into a likelihood function that is defined by unifying, for the plurality of manufacturing measurement condition sets, respective probability density functions for the respective parameters and a probability density function for the predetermined product performance and is a function to calculate a combined likelihood of the respective parameters and the predetermined product performance for the plurality of manufacturing measurement condition sets, and to store the calculated likelihood into the data storage unit; and

a controller to search for respective relational expressions for each of the manufacturing measurement condition sets in case where the likelihood becomes maximum by causing the first calculation unit, the obtaining unit and the second calculation unit to operate a plurality of times, while changing the respective relational expressions for each of the manufacturing measurement condition sets.

5. A simulation parameter correction apparatus comprising:

a processor configured to execute a process comprising:

first calculating, for each of a plurality of manufacturing measurement condition sets stored in a data storage unit, each candidate value of each parameter among a plurality of parameters that are statistics of predetermined product characteristics, based on each of relational expressions, wherein each of the relational expressions represents a relationship between a corresponding manufacturing measurement condition set and one parameter of the plurality of parameters, and storing the candidate values of the plurality of parameters, which are calculated for the corresponding manufacturing measurement condition set, into the data storage unit;

obtaining, from a variability-aware simulation, a simulation result value of a predetermined product performance with respect to the candidate values of the plurality of parameters, which are stored in the data storage unit in association with each of the manufacturing measurement condition sets, and storing the simulation result values into the data storage unit;

second calculating a likelihood by substituting the candidate values of the plurality of parameters for the respective manufacturing measurement condition sets, the simulation result values of the respective manufacturing measurement condition sets, the statistics of measurement values of the predetermined product characteristics for the respective manufacturing measurement condition sets and measurement values of the predetermined product performance for the respective manufacturing measurement condition sets into a likelihood function that is defined by unifying, for the plurality of manufacturing measurement condition sets, respective probability density functions for the respective parameters and a probability density function for the predetermined product performance and is a function to calculate a combined likelihood of the respective parameters and the predetermined product performance for the plurality of manufacturing measurement condition sets, and storing the calculated likelihood into the data storage unit; and

searching for respective relational expressions for each of the manufacturing measurement condition sets in case where the likelihood becomes maximum by carrying out the first calculating and storing, the obtaining and storing and the second calculating and storing a plurality of times, while changing the respective relational expressions for each of the manufacturing measurement condition sets.