Polishing apparatus, polishing method, and machine learning apparatus

A polishing method causing a computer to execute:

• a state acquiring step of acquiring a state variable including at least data on a state of the target of data on a state of a device making up the polishing apparatus and the data on the state of the target;
• and
• an estimating step of causing a learning section that has learned a relationship between the state variable and a change in film thickness of the target using a neural network to receive the acquired state variable to predict the change and/or to receive the acquired state variable to determine that the change is abnormal,
• wherein the data on the state of a device making up the polishing apparatus includes at least one of data on the arrangement of the device, data on an operation state of the device, and data on the state of exhaustion of the device,
• the data on a state of the target includes at least one of data detected by a film thickness detection section capable of detecting a film thickness of the target, data detected by a temperature detection section capable of detecting a temperature of the target, data detected by a pressure detection section capable of detecting a pressure applied to the target, and data on characteristics of the target, and
• the data detected by the film thickness detection section includes at least one of (i) table torque applied to a rotational table capable of rotating a polishing pad to polish the target, (ii) arm torque applied to an arm holding a top ring capable of holding the target, (iii) a current value of a first electric motor for driving to rotate the rotational table, (iv) a current value of a second electric motor for driving to rotate the top ring, (v) a current value of a third electric motor for swinging the arm.

A machine learning apparatus capable of learning a change in film thickness of a target to be polished by a polishing apparatus, the machine learning apparatus comprising:

• a state acquisition section capable of acquiring a state variable including at least data on a state of the target of data on a state of a device making up the polishing apparatus and the data on the state of the target;
• and
• a learning section that has learned a relationship between the state variable and a change in film thickness of the target using a neural network, the learning section being capable of receiving the state variable from the state acquisition section to predict the change and/or receiving the state variable from the state acquisition section to determine that the change is abnormal,
• wherein the data on the state of a device making up the polishing apparatus includes at least one of data on the arrangement of the device, data on an operation state of the device, and data on the state of exhaustion of the device,
• the data on a state of the target includes at least one of data detected by a film thickness detection section capable of detecting a film thickness of the target, data detected by a temperature detection section capable of detecting a temperature of the target, data detected by a pressure detection section capable of detecting a pressure applied to the target, and data on characteristics of the target, and
• the data detected by the film thickness detection section includes at least one of (i) table torque applied to a rotational table capable of rotating a polishing pad to polish the target, (ii) arm torque applied to an arm holding a top ring capable of holding the target, (iii) a current value of a first electric motor for driving to rotate the rotational table, (iv) a current value of a second electric motor for driving to rotate the top ring, (v) a current value of a third electric motor for swinging the arm.

A machine learning apparatus capable of learning a change in film thickness of a target to be polished by a polishing apparatus, the machine learning apparatus comprising:

• a state acquisition section capable of acquiring a state variable including at least data on a state of the target of data on a state of a device making up the polishing apparatus and the data on the state of the target;
• and
• a learning section capable of learning a relationship between the state variable and a change in film thickness of the target using a neural network, and receiving the state variable from the state acquisition section to predict the change and/or receiving the state variable from the state acquisition section to determine that the change is abnormal,
• wherein the data on the state of a device making up the polishing apparatus includes at least one of data on the arrangement of the device, data on an operation state of the device, and data on the state of exhaustion of the device,
• the data on a state of the target includes at least one of data detected by a film thickness detection section capable of detecting a film thickness of the target, data detected by a temperature detection section capable of detecting a temperature of the target, data detected by a pressure detection section capable of detecting a pressure applied to the target, and data on characteristics of the target, and
• the data detected by the film thickness detection section includes at least one of (i) table torque applied to a rotational table capable of rotating a polishing pad to polish the target, (ii) arm torque applied to an arm holding a top ring capable of holding the target, (iii) a current value of a first electric motor for driving to rotate the rotational table, (iv) a current value of a second electric motor for driving to rotate the top ring, (v) a current value of a third electric motor for swinging the arm.