a storage portion that stores a learned model obtained by performing machine learning on a maintenance condition for a print head based on a first data set and a second data set,

the first data set in which first nozzle surface image information obtained by photographing a nozzle plate surface of the print head at a first timing and maintenance information representing necessity of maintenance of the print head or a recommended execution timing of the maintenance are associated with each other,

the second data set in which second nozzle surface image information obtained by photographing the nozzle plate surface of the print head at a second timing and the maintenance information are associated with each other, and

the second timing is closer to a timing at which discharge defect of the print head occurred than the first timing;

an acquisition portion that acquires current nozzle surface image information obtained by photographing the nozzle plate surface of the print head at a current timing; and

a processing portion that determines an acquisition timing of the current nozzle surface image information based on discharge defect influence information, which influences a future discharge defect, and outputs the maintenance information based on the current nozzle surface image information, the discharge defect influence information, and the learned model, before the discharge defect of the print head occurs, wherein

the acquisition timing is shorter when the discharge defect influence information satisfies a condition than when the discharge defect influence information does not satisfy the condition.