A machine learning device includes: a data extraction unit that extracts a first parameter and a second parameter from each of a plurality of machining programs for numerically controlling a machine tool, the first parameter being a parameter to be adjusted, the second parameter being a parameter to be used for adjusting the parameter to be adjusted; and a machine learning unit that learns a value of the first parameter according to a data set including the first parameter and the second parameter extracted by the data extraction unit.

When cutting the flat products (3) a set of the desired shapes and dimension of the products (3) is defined. Firstly at least one surface of the material (1) is scanned; scanning sets the boundaries of the available surface of the material (1). Optical scanning can be supplied by radiological scanning, preferably by a CT scanner (8). Defects (2) are identified in the scanned image and a position is assigned to them. A weight coefficient is assigned to each element from a set of the desired shapes and dimensions of the products (3). A cutting plan (4) is created; this plan (4) defines the boundaries of individual flat products (3), whereby the places with the identified defects (2) of the material (1). Optimalization of the distribution of the desired products (3) is realized with the goal of achieving the highest sum of the number of the products (3) multiplied by the weight coefficient of a given product (3) without the need to cut all the elements from a set of the desired products (3). Subsequently a cutting machine (6) is used to cut the products (3); this machine (6) cuts the material (1) without any limitation with regard to the mutual position of the cut lines of the neighboring products (3).

A detection device, detection method, and compensation method for tool wear, applied to a machine tool including a spindle connected to a tool. A first parameter set including a first cutting depth having a zero cutting depth is set, and the machine tool performs a cutting procedure with the first parameter set to record a first loading rate of the spindle. A second parameter set including a second cutting depth having a non-zero cutting depth is set, and the machine tool performs the cutting procedure with the second parameter set to record a second loading rate of the spindle. A processing device calculates an estimated cutting force according to the loading rates and a machine performance database. A fuzzy logic unit outputs a wear level according to a tool wear database and the estimated cutting force. The machine tool adjusts a cutting locus according to the wear level.

A detection device, detection method, and compensation method for tool wear, applied to a machine tool including a spindle connected to a tool. A first parameter set including a first cutting depth having a zero cutting depth is set, and the machine tool performs a cutting procedure with the first parameter set to record a first loading rate of the spindle. A second parameter set including a second cutting depth having a non-zero cutting depth is set, and the machine tool performs the cutting procedure with the second parameter set to record a second loading rate of the spindle. A processing device calculates an estimated cutting force according to the loading rates and a machine performance database. A fuzzy logic unit outputs a wear level according to a tool wear database and the estimated cutting force. The machine tool adjusts a cutting locus according to the wear level.

A machine learning device includes a data extraction unit that extracts first and second parameters from a plurality of machining programs. The machining programs numerically control a machine tool. The first parameter is a parameter to be adjusted, and the second parameter is a parameter used to adjust the first parameter. The machine learning device also includes a machine learning unit that learns a value of the first parameter according to a data set that includes the first and second parameters.

When cutting the flat products (3) a set of the desired shapes and dimension of the products (3) is defined. Firstly at least one surface of the material (1) is scanned; scanning sets the boundaries of the available surface of the material (1). Optical scanning can be supplied by radiological scanning, preferably by a CT scanner (8). Defects (2) are identified in the scanned image and a position is assigned to them. A weight coefficient is assigned to each element from a set of the desired shapes and dimensions of the products (3). A cutting plan (4) is created; this plan (4) defines the boundaries of individual flat products (3), whereby the places with the identified defects (2) of the material (1). Optimalization of the distribution of the desired products (3) is realized with the goal of achieving the highest sum of the number of the products (3) multiplied by the weight coefficient of a given product (3) without the need to cut all the elements from a set of the desired products (3). Subsequently a cutting machine (6) is used to cut the products (3); this machine (6) cuts the material (1) without any limitation with regard to the mutual position of the cut lines of the neighboring products (3).

A method and an apparatus for smart automation of robotic surface finishing of a three-dimensional surface of a workpiece is described. A finite element analysis simulation is conducted providing data for generation of a three-dimensional path along the surface of the workpiece. The finite element can include properties of the workpiece, finishing tool, and the robot configured to maneuver the finishing tool. The surface of the workpiece is finished using one or more surface finishing tools along the three-dimensional path. The surface of the workpiece includes at least a flat region and a curved region.

A machining direction determination unit determines a machining direction, i.e., whether to orient a product in a first direction along a material axis or a second direction opposite to the first direction, on the basis of constraint condition data indicating constraint conditions of front machining (first machining) and/or back machining (second machining), and product data. A time estimate output unit calculates a first time estimate required for the front machining and a second time estimate required for the back machining in accordance with the machining direction on the basis of machining condition data indicating machining conditions (including a machining speed) corresponding to tools used respectively for the front machining and the back machining, as well as material data and the product data, and outputs an estimate of a time required to manufacture the product on the basis of the first time estimate and the second time estimate.