Virtual sensor based monitoring and fault detection/classification system and method for semiconductor processing equipment

A virtual sensor based monitoring system for monitoring conditions in processing equipment, comprising:

• a plurality of equipment sensors coupled to processing equipment, each of the plurality of sensors operable to measure a process condition and provide a signal representing the measured process condition;
• a plurality of filtering process units coupled to the plurality of equipment sensors, each filtering process unit operable to receive at least one signal from the plurality of equipment sensors, and each filtering process unit operable to reduce data represented by the at least one signal and provide filtered data, wherein each filtering process unit comprises;
• a data reduction filter operable to receive the at least one signal, to process the at least one signal according to a data reduction algorithm, and to provide reduced data;
• a sensor drift correction unit operable to receive the reduced data, to correct the reduced data responsive to modeled sensor drift, and to provide corrected data;
• and
• a dimensional reduction filter operable to receive the corrected data, to reduce the corrected data according to a dimensional reduction algorithm, and to provide filtered data;
• a plurality of redundant virtual sensors coupled to the plurality of filtering process units and operable to receive the filtered data, a first portion of the plurality of redundant virtual sensors modeling states of the processing equipment and a second portion of the plurality of redundant virtual sensors modeling states of a work piece in the processing equipment, and each virtual sensor operable to provide an output signal representing an estimated value for the modeled state;
• and wherein each redundant virtual sensor is comprised of a plurality of virtual sensors each of which provide an output signal, said output signals of said each virtual sensor is compared to the output signals of the other virtual sensors associated to a particular redundant virtual sensor and the output signal of each redundant virtual sensor is the output signal of the virtual sensors which have a similar output.

A method for constructing a virtual sensor based fault detection and classification system for semiconductor processing equipment, comprising:

• measuring a plurality of process conditions for a processing equipment using equipment sensors and providing a plurality of signals representing the measured process conditions;
• filtering the plurality of signals to reduce the data represented by the signals and providing filtered data;
• processing the filtered data in a plurality of virtual sensors and using the virtual sensors to model and estimate states of the processing equipment;
• providing virtual sensor output signals representing estimated values for the modeled states, said output signals obtained by comparing output signals of a plurality of redundant virtual sensors and setting said output signal to the value of said output signals of said plurality of redundant virtual sensors which have similar values;
• processing the signals provided by the equipment sensors and the output signals provided by the virtual sensors using a rule based logic system to detect and classify faults within the processing equipment.