Invented by Mitsunori Watanabe, Fanuc Corp

The market for Machine Learning Device and Thermal Displacement Compensation Device In recent years, the field of machine learning has witnessed significant advancements, leading to the development of various innovative devices. Two such devices that have gained considerable attention are Machine Learning Devices and Thermal Displacement Compensation Devices. These devices have revolutionized industries by enhancing efficiency, accuracy, and performance in various applications. Machine Learning Devices, also known as artificial intelligence (AI) devices, are designed to mimic human intelligence and perform tasks that typically require human intervention. These devices utilize algorithms and statistical models to analyze large datasets, learn from patterns, and make predictions or decisions without explicit programming. Machine Learning Devices have found applications in diverse fields such as healthcare, finance, manufacturing, and transportation. The market for Machine Learning Devices has experienced substantial growth in recent years and is expected to continue expanding at a rapid pace. The increasing demand for automation and the need for efficient data analysis have been the primary drivers of this growth. Machine Learning Devices offer numerous benefits, including improved accuracy, reduced human error, and enhanced productivity. These devices can process vast amounts of data in real-time, enabling businesses to make informed decisions quickly. Thermal Displacement Compensation Devices, on the other hand, are devices designed to compensate for the thermal expansion or contraction of materials. These devices are particularly crucial in industries where temperature changes can significantly impact the performance and accuracy of equipment. Thermal Displacement Compensation Devices utilize advanced sensors and algorithms to measure temperature variations and adjust the position or alignment of components accordingly. The market for Thermal Displacement Compensation Devices has witnessed steady growth due to the increasing demand for precision and reliability in various industries. These devices are widely used in sectors such as aerospace, automotive, electronics, and manufacturing. Thermal Displacement Compensation Devices help minimize the adverse effects of temperature changes on equipment, ensuring consistent performance and reducing the risk of errors or failures. The combination of Machine Learning Devices and Thermal Displacement Compensation Devices has the potential to revolutionize industries further. By integrating machine learning capabilities into thermal displacement compensation systems, businesses can achieve even higher levels of accuracy, efficiency, and reliability. These integrated devices can continuously learn from data, adapt to changing conditions, and optimize compensation strategies, resulting in improved performance and reduced downtime. The market for integrated Machine Learning Devices and Thermal Displacement Compensation Devices is expected to witness significant growth in the coming years. The increasing adoption of automation and the growing demand for precision in various industries are driving this growth. Businesses are increasingly recognizing the benefits of combining machine learning capabilities with thermal displacement compensation systems to enhance their operations and gain a competitive edge. In conclusion, the market for Machine Learning Devices and Thermal Displacement Compensation Devices is experiencing rapid growth and is poised for further expansion. These devices offer numerous benefits, including improved accuracy, efficiency, and reliability. The integration of machine learning capabilities with thermal displacement compensation systems has the potential to revolutionize industries by enabling businesses to achieve higher levels of performance and productivity. As technology continues to advance, the market for these devices is expected to witness continuous innovation and growth.

The Fanuc Corp invention works as follows

The machine learning device includes a measurement data acquisition module that acquires measured data groups; a thermal distance acquisition module that acquires thermal displacement measurements about a machine component; a storage system that stores input data using the measured group acquired by this unit; labels the input data with the thermal measured value acquired by the thermal acquisition unit; and stores input data as well as the label as teaching data.

Background for Machine Learning Device and Thermal Displacement Compensation Device

Field of Invention

The present invention is a machine-learning device and a thermal displacement compensator used in a machine tool.

Related Art

The thermal expansion of machine elements in the machine tool is one of the major factors in a machining mistake. Spindle, spindle unit and bed are all examples of components in the machine. The spindle is thermally deformed, by the generation of heat through the rotation of the motor or spindle. Other components, such as the coolant, supplied by the coolant supplier, absorb heat, and the lubrication supplied by the lubrication supplier, can also do this. In some cases, this causes a relative thermal displacement between the tooling and the workpiece.

According to a conventional technique used to address this issue, the numerical controller is compensated for the influence of thermal expansion on a spindle due to various heat sources, such as the heat source in the machine tool near the spindle and the outside air temperature. This increases machine accuracy. (See Patent Document 1, by way of example).

However Patent Document 1 only mentions the installation of multiple sensors to obtain a characteristic value of the machine tool. Patent Document 1’s technique does not guarantee highly accurate compensation.

Further, it takes time to transfer the heat detected by the temperature sensors and cause thermal expansion. It is therefore necessary to calculate time delay in order to provide highly accurate compensation. This leads to a complex compensation formula.

Also, the structure or member of the machine where numerical controllers will be installed may change.” A thermal displacement compensation formula that is optimal for a particular machine type will be changed.

Also, the use of an environment such as ambient heat or a gate can change the external heat source. It is necessary to change the location of a temperature detector, and therefore change an optimal compensation formula. Increasing the number of measuring instruments to locate the temperature sensor in the correct position will increase production costs and maintenance costs.

The present invention aims to provide both a thermal displacement compensator and a machine-learning device that can not only derive a highly accurate formula for compensation but also achieve highly accurate compensation at a low cost.

The machine learning unit comprises: a measurement data acquisition unit (measured data acquisition unit 11 described later for example) which acquires the measured data group; a thermal displacement acquisition (thermal displacement acquired 12 described later for example), which acquires a thermal displacement actual measured value about the machine component; a storage (storage 13 described later for example) using the input data and the label of the thermal data group. The machine learning device includes: a measurement data acquisition (measured data unit 11 described in detail later) which acquires the data group; a thermo displacement acquisition (thermal displacement unit 12 described in detail later) which acquires the actual thermal displacement value about a machine element; a teaching unit (storage 13 described in depth later) which stores input data as well as the label and labels the input data with the label. The calculation formula unit determines the thermal distance estimation calculation formula by comparing a thermal estimated value calculated about the machine component by substituting the measurement data group for a predetermined time period in the storage data to the thermal estimation calculation formula, and the actual thermal displacement measured value about that machine component in the predetermined time period in the storage data.

(2) The machine learning device described at (1) may acquire a second group of measured data by adding or excluding measured values from the group. The second measured data group may be stored by the measured data acquisition unit as input data in the storage unit (storage 13 described later, for instance). The calculation formula-learning unit (calculation unit learning unit 14, described later, for instance) can also set a second thermal distance estimation calculation formula that is used to calculate the thermal displacement of a machine tool using the second measured data.

(3) The machine-learning device described in (2) can include a unit for determining contribution (contribution unit 15, described later, as an example), which determines the contribution of measured data to the estimation of thermal displacement in the group of measured data. The contribution determination unit can determine the contribution as a calculation target by comparing a difference between two errors. The first error is the difference between a first estimated thermal displacement value and an actual thermal displacement measured value. The first estimated thermal displacement value is calculated by using a set of first thermal displacement calculation formulas based on the group of measured data containing the measured values as a calculation target. The second error is the difference between a second estimated value and an actual thermal displacement measured value. The second estimated thermal displacement value is calculated by using the second calculation formula for thermal displacement estimation based on the group of second measured data from which the contribution calculation target was excluded.

(4) The machine-learning device described in (3) can include an optimized data selection unit (optimized data selection unit 16, for example), which selects an optimal measured data group consisting of a combination measured data pieces that belong to a current measured data group and achieves the highest degree of accuracy by using a certain number of measured pieces. The optimized data selection unit can select a first group of measured data by excluding the data that has the lowest contribution, as determined by the unit for contribution determination. The optimized data selection unit can select the (i+1),-th group of measured data by excluding the measured data with smallest contribution, as determined by contribution determination unit, from the i-th (1? i) measured group. This selection is repeated, selecting the optimized data group that contains the predetermined number measured data pieces.

(5) The thermal displacement estimation formula in the machine-learning device described in (1 to 4) may use a first order lag element within the measured data of the measured data group.

(6) The thermal displacement estimation formula in the machine-learning device described in (1 to 5) may include a time shift component in the measured data of the measured data group.

(7) The thermal displacement estimation formula can be set in the machine-learning device described in (1 to 6) based on machine-learning using a neural net.

The calculation formula-learning unit (calculation unit learning unit 14, for example, described later) can set the thermal displacement calculation formula using machine learning and L2 regularization multiple regress analysis.

(9) The calculation formula learning module (calculation formula-learning unit 14, described later) in the machine learning device described under (1) to (6) may set the thermal displacement estimate calculation formula by sparse regularization.

(10) The machine-learning device described in (9) can further include a detection unit that detects the measured data within the group of measured data that does not increase the accuracy in thermal displacement estimation. The detection unit can detect the measured data using the thermal displacement calculation formula that is set up by sparse regularization.

(11) A machine learning device (as described in (1)-10) can be integrated into a controller for machining tool (as described later in the article, for instance) (as described later in the article, an example of machinary tool is 35).

Click here to view the patent on Google Patents.