Induction coil structural unit, method for monitoring induction heating process of induction coil structural unit, self-learning system for monitoring induction heating process, and method for training self-learning system.

Abstract

This invention relates to an induction coil structure unit and a method for monitoring the induction heating process of the induction coil structure unit. The invention also relates to a self-learning system for monitoring the induction heating process and a method for training the self-learning system. To improve the automation level and operational safety of the induction coil structure unit, it is proposed to provide at least one, and in particular multiple, different characteristic quantities of the induction process, particularly coil current and / or coil voltage and / or input current and / or input voltage and / or intermediate circuit current and / or intermediate circuit voltage. These at least one or more different characteristic quantities are input into a learning system, particularly a neural network, and then the learning system uses these at least one or more different characteristic quantities to determine monitoring characteristic quantities, such as: temperature characteristic quantities introduced into the sleeve portion of the induction coil and / or sleeve characteristic quantities introduced into the sleeve portion of the induction coil and / or time characteristic quantities of the heating process introduced into the sleeve portion of the induction coil.

Description

[0001] This invention relates to an induction coil structure unit and a method for monitoring the induction heating process of the induction coil structure unit. The invention also relates to a self-learning system for monitoring the induction heating process and a method for training the self-learning system.

[0002] This type of induction coil structure unit is known from the prior art, for example from EP 1 867 211 A1.

[0003] These known induction coil structural units are used to thermally expand a tool holder by means of an alternating magnetic field generated by the induction coil and the eddy currents induced therein in the tool holder within the induction coil introduced into the induction coil structural unit, so that a tool can be inserted there in this expanded state of the tool holder, which is then firmly and symmetrically held by the tool holder after the tool holder has cooled down. This process is also known (abbreviated) as inductively clamping the tool in the tool holder and is known per se.

[0004] For a further description of the technical and operational background, please refer to the aforementioned patent application.

[0005] However, a problem with this known induction coil unit is that, in order for the induction coil unit to operate efficiently—specifically, to inductively clamp the tool in the tool holder, especially when heating the tool holder—the induction coil unit must be individually adjusted for different operating or clamping parameters, particularly the heating duration (or clamping / heating frequency, clamping / heating temperature, maximum heating time, or (clamping / heating) energy), for the tool holder just now / currently held there. This requires a high degree of manual intervention and can therefore significantly extend the cycle time for changing tools in different types of tool holders. Furthermore, manual intervention always presents a potential source of error.

[0006] On the other hand, if this adjustment of the currently held tool holder is not performed or is performed with errors, the operation of the induction coil structure unit may be inefficient because the supplied eddy current cannot be induced in the tool holder in a suitable manner. In particularly adverse cases, such as excessively long heating duration, this may even cause the tool holder to overheat and thus be damaged or destroyed.

[0007] In addition, self-learning systems are also known.

[0008] A self-learning system is a system that can learn from experience or data and automatically improve based on that, without requiring explicit programming instructions. These systems adjust their operation based on new information to optimize their performance.

[0009] A common example of a self-learning system is a machine learning algorithm (such as a neural network, decision tree, or support vector machine), which can identify patterns from large amounts of data and make predictions or decisions. Unlike traditional procedures where all rules and processes are pre-defined, self-learning systems continuously improve through training, and their predictive accuracy or efficiency increases with the amount of data they are exposed to.

[0010] For example, self-learning systems can classify images, translate text, and even make decisions in dynamic environments without continuous human intervention.

[0011] As mentioned above, a neural network is a machine learning model whose principle is based on how the human brain works, which is also known in current technology.

[0012] This type of neural network consists of interconnected "neurons" organized in multiple layers to process data and recognize patterns. It is particularly well-suited for tasks such as image recognition, speech processing, translation, and prediction.

[0013] The main components of a neural network are the layered, structured neurons described above (input layer, output layer, hidden layer). A neuron (node) is the basic processing unit of the network. Each neuron receives input data, performs calculations, and passes the results to the neurons in the next layer.

[0014] The input layer receives input data (such as images or text) and passes it to the next layer. The output layer outputs the model's final result or prediction, such as image classification. Hidden layers are located between the input and output layers and perform the main computations. Complex patterns and features are extracted from the input data in these layers.

[0015] Each connection between neurons has a weight that determines the strength of the signal transmitted from one neuron to the next. In addition, there are bias terms that help shift the output of neurons and improve the flexibility of the model.

[0016] After a neuron processes the input data, an activation function is typically applied to determine whether the neuron is activated (i.e., "fires"). Examples of activation functions include the Sigmoid function, ReLU (Rectified Linear Unit), and Tanh.

[0017] Neural networks are improved through training. During this process, a large amount of data is used to adjust the weights and biases so that the network can predict or classify as accurately as possible. This process is typically accomplished through a method called backpropagation, in which the error between the network's predictions and the actual results is passed back through the network to adjust the weights.

[0018] Neural networks, especially deeper networks (i.e., so-called deep learning networks), can perform very complex tasks with high precision; however, they also require a lot of data and considerable computing resources.

[0019] Decision trees are also known as self-learning systems. In the field of artificial intelligence, particularly machine learning, decision trees are a commonly used algorithm for classification and regression.

[0020] A decision tree is a system model used to represent and analyze decision-making processes. It represents decisions and their possible outcomes in a tree structure. The starting point is the initial node, from which different branches branch out, representing different action plans or events.

[0021] Here, decision nodes (usually represented by squares) and opportunity nodes (usually represented by circles) are distinguished. Decision nodes involve making a definite choice, while opportunity nodes may result in different events with varying probabilities. The endpoints of the tree show the possible outcomes, which are typically evaluated in terms of costs, benefits, or gains.

[0022] Decision trees are used to make complex decisions transparent by comparing various options and selecting the best course of action based on probabilities and expected outcomes.

[0023] Like other self-learning systems (such as neural networks, see above), decision trees require training. Known training methods / algorithms available for this purpose include: top-down induction (recursive partitioning), C4.5 and its successors (such as CART, ID3), Bagging, Random Forest, or Boosting (such as AdaBoost, gradient boosting).

[0024] In addition, implementation methods for designing and optimizing decision tree structures are known, such as Scikit-learn.

[0025] Scikit-learn provides a suitable implementation of the (above) CART algorithm (classification and regression tree) for both classification and regression problems.

[0026] With the help of Scikit-learn, decision trees can be easily trained simply by passing the data and target variable to the corresponding class (such as the decision tree classifier or the decision tree regressor). Here, Scikit-learn calculates metrics such as Gini impurity or information gain to find the optimal splitting method.

[0027] The purpose of this invention is to provide an induction coil structure unit, a method for monitoring the induction heating process of the induction coil structure unit, a self-learning system for monitoring the induction heating process, and a training method for such a self-learning system, which can overcome the above-mentioned disadvantages of such induction coil structure units known in the prior art, wherein the induction coil structure unit has a higher degree of automation and thus can operate with higher operational safety and shorter cycle time (reliable process).

[0028] This objective is achieved by an induction coil structure unit having the features of the respective independent claims and a method for monitoring the induction heating process of the induction coil structure unit, as well as a self-learning system for monitoring the induction heating process and a method for training such a self-learning system. Advantageous improvements of the invention are the subject of the dependent claims and the following description, and relate to an induction coil structure unit and a method for controlling the induction heating process of the induction coil structure unit, as well as a self-learning system for monitoring the induction heating process and a method for training such a self-learning system.

[0029] Terms used where necessary, such as “upper,” “lower,” “front,” “rear,” “left,” or “right,” should be understood according to their usual meaning, taking into account the accompanying drawings, unless otherwise explicitly defined. Terms such as “radial” and “axial,” if used but not explicitly defined, should be understood with reference to the central axis or axis of symmetry of the component / part described herein, taking into account the accompanying drawings.

[0030] When used, the term “substantially” (according to the Supreme Court’s interpretation) can be understood to mean “to a degree that is still substantially substantial in practice.” Due to factors such as manufacturing or assembly tolerances, there may be unintentional (i.e., without functional justification) deviations from the precision implied by the term.

[0031] An induction coil structure unit and a method for monitoring the induction heating process of the induction coil structure unit are provided in the induction coil structure unit, and the sleeve portion of the tool holder can be introduced into the induction coil.

[0032] The core of this invention lies in a self-learning system, especially a neural network or decision tree, which can provide at least one, especially multiple, different characteristic quantities of the sensing process for the self-learning system, especially current and / or voltage, such as coil current and / or coil voltage and / or input current and / or input voltage and / or intermediate loop current and / or intermediate loop voltage.

[0033] In addition, at least one additional feature quantity can be provided for the self-learning system, especially neural networks or decision trees, such as: the instantaneous change of coil current and / or the instantaneous change of intermediate loop current, and / or the instantaneous rate of change of coil current and / or the instantaneous rate of change of intermediate loop current, and / or the absolute value of coil voltage and / or the absolute value of coil current, and / or the energy of coil current flowing since the start of heating and / or the energy of intermediate loop current flowing since the start of heating, and / or the integral of instantaneous coil current (active current) flowing since the start of heating and / or the integral of intermediate loop current flowing since the start of heating, and / or the heating time since the start of heating, and / or the instantaneous inductance of the induction coil or the instantaneous inductance of the overall system consisting of the induction coil and the sleeve, and / or the instantaneous phase angle between the instantaneous coil current (active current) and the coil voltage driving the coil current.

[0034] Another such feature (provided / possibly provided to self-learning systems, especially neural networks or decision trees) could be the time required for a pre-defined change (e.g., an increase) in the coil current or intermediate loop current. For example, this could be the time required for the coil current or intermediate loop current to rise from a first (lower) limit (e.g., 0 A) to a second (higher) limit (e.g., a pre-defined maximum current). This could occur, for example, during the application of a test pulse (see below for more information on test pulses).

[0035] Then, the self-learning system uses at least one or more different characteristic quantities (and, if necessary, at least one additional characteristic quantity) to determine the monitoring characteristic quantity, such as: the temperature characteristic quantity of the sleeve portion introduced into the induction coil and / or the sleeve characteristic quantity of the sleeve portion introduced into the induction coil and / or the time characteristic quantity of the heating process of the sleeve portion introduced into the induction coil.

[0036] In this invention, the term "characteristic quantity" can refer to a (one-dimensional) value or (multi-dimensional) vector of a (process) quantity or variable, especially a variable that changes over time, such as current, voltage, temperature, inductance, energy, etc.

[0037] Using multiple hidden layers, especially more than 10 hidden layers, especially more than 15 hidden layers, especially more than 20 hidden layers, especially more than 25 hidden layers, even more than 30 hidden layers, and most importantly more than 32 hidden layers, can optimize self-learning systems or neural networks / decision trees.

[0038] Furthermore, it is advantageous to use the ReLU function and / or the Sigmoid function as activation functions to further improve self-learning systems or neural networks.

[0039] Self-learning systems or neural networks / decision trees can be trained using the aforementioned features (methods for training self-learning systems) to determine monitoring features, such as: temperature features of the sleeve portion introduced into the induction coil and / or sleeve features of the sleeve portion introduced into the induction coil and / or time features of the heating process of the sleeve portion introduced into the induction coil.

[0040] Training is typically performed using so-called training data (or in the form of so-called training data vectors) and as much of this data as possible. For example, training can be conducted using features provided to a self-learning system or a neural network / decision tree (see at least one or more of the different features mentioned above, or at least one additional feature mentioned above, such as, in particular, current and / or voltage, such as, in particular, coil current and / or coil voltage and / or input current and / or input voltage and / or intermediate loop current and / or intermediate loop voltage, and / or, in particular, the instantaneous change of coil current and / or the rate of instantaneous change of coil current and / or the absolute value of coil voltage and / or the absolute value of coil current and / or the energy of the coil current flowing since the start of heating and / or the instantaneous linear energy flowing since the start of heating). The integral of the coil current (active current) and / or the heating time since the start of heating and / or the instantaneous inductance of the induction coil or the instantaneous inductance of the overall system consisting of the induction coil and the sleeve section and / or the instantaneous phase angle between the instantaneous coil current (active current) and the coil voltage driving the coil current and / or the (sustained) time required for the coil current or intermediate loop current to change (e.g., rise) in a preset manner, especially including / using monitoring feature quantities provided by the self-learning system or neural network / decision tree accordingly (see, for example, the temperature feature quantity introduced into the sleeve section of the induction coil and / or the sleeve feature quantity introduced into the induction coil and / or the time feature quantity of the heating process of the sleeve section introduced into the induction coil) for training (“training data / training vector”).

[0041] These training data can be determined by a heating process performed in a defined manner, particularly by measurements taken during the heating process or by determining the quantities measured during the heating process. In particular, these data can be obtained by performing a heating process on a measuring cone, i.e., by obtaining the (measured) characteristic quantities and their corresponding (measured) monitoring characteristic quantities during or after these "defined" heating processes.

[0042] For example, these (measured) training data can be obtained at a "resolution" of 5, 10, or 20 values ​​per second during the heating process.

[0043] The corresponding monitoring features can also be obtained through user feedback and / or user interaction. In other words, user feedback and / or user interaction can be used to train a self-learning system or a neural network / decision tree.

[0044] To improve the performance of self-learning systems or neural networks, it is proposed to use the backpropagation method for training.

[0045] Furthermore, in order to improve the self-learning system or neural network, it can be retrained, especially using user feedback and / or user interaction.

[0046] In any case, incorporating multiple data sources during training is important for achieving comprehensive model building. This can include real-world measurement data, simulated scenarios, or synthetic data, enabling the system to handle rare or extreme situations.

[0047] This data can be used to enable self-learning systems or neural networks / decision trees to handle a variety of scenarios, including those that may rarely occur in practice. Such data helps improve the robustness and generalization ability of the model.

[0048] This invention is based on the understanding that complex processes are difficult, if not impossible, to describe using (precise) analytical methods; therefore, the precision of process quantities that can be determined through analysis is insufficient. Consequently, control systems built upon this understanding present operational safety risks. In other words, taking the (induction) heating of a tool holder as an example, purely "analytical" monitoring and control of the (induction) heating process of the sleeve or tool holder based on analytically determined quantities (e.g., the applicant's German patent application DE 10 2024 128 298.3) may pose safety risks.

[0049] However, self-learning systems, such as neural networks or decision trees, offer a means to more accurately map or describe complex processes (e.g., the (induction) heating process of the sleeve section in this paper) by learning the process through the self-learning system or neural network / decision tree (based on (training) data). The self-learning system or neural network / decision tree creates a mapping of the process / heating flow (heating process) to a certain extent, which can then be used to determine process parameters (such as monitoring features) for monitoring and control. Based on this, the process can be monitored, controlled, and / or regulated.

[0050] In other words and in short, the self-learning system or neural network / decision tree according to the invention generates reliable process quantities (here, monitoring feature quantities) with high accuracy, on which safe and error-free (process) monitoring and control can then be carried out (see the comparison of limits for shutdown mentioned below).

[0051] Furthermore, the self-learning system or neural network / decision tree approach according to the present invention offers the possibility of predictive applications. It can not only monitor the current process status and identify anomalies and deviations from normal conditions, but also predict future developments. This is particularly valuable for proactively "maintaining" and optimizing operating parameters. Potential faults can be detected early and (if necessary, repaired) to avoid costly downtime. This improves reliability and availability.

[0052] Combining (traditional) "analysis-based" with "neural" monitoring (and control) can make the monitoring of induction heating processes more reliable.

[0053] Furthermore, this invention contributes to increased automation. If integrated into automated induction heating of tool holders, such as in automated tool changing processes, a self-learning system or neural network can automatically acquire feature quantities during / at the process. The self-learning system or neural network / decision tree then determines the monitored feature quantities without manual user intervention. This means the process can operate with high operational safety and shorter cycle times (process reliably).

[0054] method

[0055] In a method for controlling an induction heating process of an induction coil structural unit (e.g., a clamping device) (where the sleeve portion of a tool holder is introduced into the induction coil of the induction coil structural unit), it is proposed to provide at least one, and in particular multiple, different characteristic quantities of the induction process.

[0056] To better map the heating process through a self-learning system, it is advantageous to provide at least one additional characteristic quantity, such as: the instantaneous change of coil current and / or the instantaneous change of intermediate loop current, and / or the instantaneous rate of change of coil current and / or the instantaneous rate of change of intermediate loop current, and / or the absolute value of coil voltage and / or the absolute value of coil current, and / or the energy of coil current flowing since the start of heating and / or the energy of intermediate loop current flowing since the start of heating, and / or the integral of instantaneous coil current (active current) flowing since the start of heating and / or the integral of intermediate loop current flowing since the start of heating, and / or the heating time since the start of heating, and / or the instantaneous inductance of the induction coil or the instantaneous inductance of the overall system consisting of the induction coil and the sleeve, and / or the instantaneous phase angle between the instantaneous coil current (active current) and the coil voltage driving the coil current.

[0057] Another such provided / available characteristic could be the time required for a preset change (e.g., an increase) in the coil current or intermediate loop current. For example, this could be the time required for the coil current or intermediate loop current to rise from a first (lower) limit (e.g., 0 A) to a second (higher) limit (e.g., a preset maximum current). This could occur, for example, during the application of a test pulse (see below for more information on test pulses).

[0058] In any case, such a characteristic quantity can be a quantity determined using at least one or more different characteristic quantities, for example by integration, differentiation or similar methods, such as energy as an integral of coil current (active current).

[0059] Advantageously, the features provided to the self-learning system are first passed through a digital smoothing filter (e.g., a Kalman filter) before being input into the learning system. This filter is particularly suitable for real-time data processing because it continuously updates predictions and combines them with measurements to generate more accurate estimates. This filter is well-suited for dynamic systems where states can change rapidly, such as induction processes. The filtering performance can be optimized by employing adaptive algorithms that dynamically adapt to the constantly changing conditions of the heating process.

[0060] "Providing" can also mean obtaining the at least one or more different characteristic quantities and / or the at least one additional characteristic quantity by measuring them before and / or during the heating process, especially during the process, or by providing, in particular, stored or calculated values ​​for the at least one or more different characteristic quantities and / or the at least one additional characteristic quantity.

[0061] These characteristic quantities can also be “provided” through measurement, for example, through appropriate sensors or sensor technologies (especially sensors using Internet of Things technology).

[0062] The at least one or more distinct features (or these features and the at least one additional feature) are input into the learning system.

[0063] The learning system uses the input feature quantities to determine the monitored feature quantities, such as: the temperature feature quantity of the sleeve portion introduced into the induction coil and / or the sleeve feature quantity of the sleeve portion introduced into the induction coil and / or the time feature quantity of the heating process of the sleeve portion introduced into the induction coil.

[0064] For example, a temperature characteristic could be the temperature of the sleeve portion of a tool holder; a sleeve characteristic could be the geometry of the sleeve portion or the tool holder, especially the tool holder dimensions, or the classification (or description) information of the sleeve portion or the tool holder; and a time characteristic could be the remaining heating duration.

[0065] The heating process is monitored by using monitoring characteristics (such as temperature characteristics and / or sleeve characteristics and / or time characteristics).

[0066] Furthermore, it can be proposed that when a certain feature among the monitored features and / or features input to the self-learning system reaches, exceeds, or falls below a set limit, the heating of the sleeve section should be stopped (or at least altered). This ensures the reliable operation of the heating process.

[0067] Advantageously, the heating and / or shutdown parameters are set or readjusted using the determined monitoring characteristics, and the heating process is then monitored, executed, or controlled based on these parameters. Similarly, it can be proposed that heating of the sleeve be stopped (or at least altered) when such parameters reach, exceed, or fall below settable limits.

[0068] Furthermore, in order to further improve the automation of heating, it can be proposed to determine the geometry (e.g., outer diameter or other similar parameters) of the sleeve portion into which the induction coil is introduced before starting (especially automatically), thereby "identifying" the sleeve portion into which it is introduced.

[0069] For example, this is described in German patent applications 10 2015 016 831.2 and 10 2019 112 521.9, and in European patent application 19 17 6562.7. Its contents are the subject of this application.

[0070] Once the introduced sleeve section is identified, heating parameters can be set for the introduced sleeve section or its heating process, such as the provided characteristic quantities or the characteristic quantities input into the self-learning system.

[0071] These heating parameters can be, in particular, the hoop / heating frequency, and / or the hoop / heating temperature, and / or the duration of the heating process, and / or the maximum duration and / or energy (current integral) of the heating process. This can also define limits for the heating process (especially for "neural" monitoring and / or "analytical" monitoring).

[0072] Furthermore, in another preferred improvement, it can be proposed to use multiple decision trees instead of just one.

[0073] In particular, it is advantageous to use multiple decision trees with the same structure. Each decision tree can be trained on the same output or output parameters (all outputs can then be integrated into the final output, for example, proportionally), thereby improving the reliability of the monitored data or the quality of predictions by combining all decision trees.

[0074] Structural unit

[0075] According to the invention, a self-learning system is provided in the induction coil structure unit used to perform the method.

[0076] As mentioned earlier, the self-learning system determines the monitored feature quantity by using the provided feature quantity.

[0077] At least one current and / or voltage detector may also be provided in the induction coil structure unit. This at least one current and / or voltage detector is provided in the intermediate circuit of the frequency converter for supplying electrical energy to the coil circuit and / or in the coil circuit with the induction coil.

[0078] Advantageously, the induction coil structure unit is equipped with a comparator for comparing at least one of the aforementioned characteristic quantities (e.g., particularly a characteristic quantity of the monitoring characteristic quantity and / or a characteristic quantity of the input self-learning system) with a settable limit value.

[0079] To improve operational safety, it is advantageous that the induction coil structure unit also has an automatic shut-off device, which is configured to shut off (or at least alter) the heating process carried out by the induction coil when the aforementioned settable limit is reached, lowered, or exceeded during comparison.

[0080] Furthermore, in order to further improve the induction coil structure unit technically, it can be proposed that at least two, especially parallel, self-learning systems be provided in the induction coil structure unit, which can advantageously determine different monitoring characteristic quantities, such as temperature characteristic quantity and time characteristic quantity, or sleeve characteristic quantity and time characteristic quantity, or sleeve characteristic quantity and temperature characteristic quantity.

[0081] To further improve the operational safety of the induction coil structure unit, it is advantageous for the circuit to have at least one power semiconductor structure element, particularly at least one insulated gate bipolar transistor (IGBT) and / or metal oxide semiconductor field-effect transistor (MOSFET), which have good conduction behavior, high blocking voltage and robustness, and can be operated with almost no power.

[0082] This method allows for highly automated or automatic operation of the tool holder newly introduced into the induction coil structural unit. Specifically, it enables the tool to be inductively clamped within the tool holder, and particularly heats the tool holder. This eliminates the need for manual intervention to adjust operating parameters, saving time previously required. Furthermore, the automated / automated operation maintains high standards of operational safety and tolerance, ensuring the structural unit operates according to specifications. This method also provides efficient protection against overheating of the heated / expanded tool holder within the induction coil structural unit. Power output can be optimized, reducing energy consumption and improving efficiency.

[0083] The above description of the advantageous embodiments of the invention includes numerous features, which are presented in the individual dependent claims as a plurality of features in a manner partially combined. However, these features can also be advantageously viewed individually and combined into other meaningful combinations.

[0084] In particular, these features can be combined independently and in any suitable combination with the method and / or the device according to the invention.

[0085] Even if certain terms are used in the specification or in the patent claims, either in the singular or in combination with quantifiers, the scope of these terms in this invention should not be limited to the singular or the corresponding quantifiers. Furthermore, the words “an” or “a” are not to be understood as quantifiers, but rather as indefinite articles.

[0086] The features, characteristics, and advantages of the present invention described above, as well as the ways and means of implementing them, will become clearer and more readily understood in conjunction with the following detailed description of the embodiments of the present invention in conjunction with the accompanying drawings / illustrations (the same components / parts and functions have the same reference numerals in the drawings / illustrations).

[0087] These embodiments are used to illustrate the invention, and the invention is not limited to the combinations of features given therein, nor is it limited in terms of functional features. Furthermore, each feature suitable for this invention can be clearly viewed separately, removed from the embodiments, introduced into other embodiments to supplement other embodiments, and combined with any of the claims.

[0088] In the attached diagram:

[0089] Figure 1 The induction coil structure unit according to the embodiment is shown in the middle longitudinal section;

[0090] Figure 2A circuit diagram is shown for feeding an induction coil, which can be used according to... Figure 1 The induction coil structure unit;

[0091] Figure 3 Showing the monitoring data Figure 1 The neural network of the heating process in the induction coil structure unit;

[0092] Figure 4 Shown in accordance with Figure 1 In the induction coil structure unit, the coil current curves of different tool holders under test pulses; and

[0093] Figure 5 Showing the monitoring data Figure 1 Decision tree for the heating process in the induction coil structure unit.

[0094] Automatic heating control of the hoisting process using hoisting equipment / Monitoring of the induction heating process through a self-learning system (neural network or decision tree). Figures 1 to 2 as well as Figure 4 , Figure 3 and Figure 5 )

[0095] Figure 1 The basic structure of the induction coil structure unit is shown, which should also be referred to as a hoop device (or hooper) due to the function it provides here (in the following text).

[0096] Induction coil structure unit / shrinkage device

[0097] like Figure 1 The device shown has an induction coil 1 with a separate coil 2. The tool holder 4 is pushed into the center of the induction coil so that the clamping shank H of the tool W (e.g., a milling cutter) is clamped into or disengaged from the sleeve HP.

[0098] The functional principle underlying the clamping or disengagement is described in detail in German patent application DE 199 15 412 A1. Its contents are the subject of this application.

[0099] The induction coil 1 has a first outer cover 3 on its outer periphery, which is made of a non-conductive but magnetic material.

[0100] Typically, the first outer casing 3 is formed of ferrite, metal powder, or sintered metal material, the individual particles of which are electrically insulated from each other, and these particles are generally magnetic but not electrical in general.

[0101] The first outer cover 3 is also implemented such that it is completely self-closed in the circumferential direction, that is, the circumferential surface of the induction coil 1 is completely covered, so that theoretically no "magnetic gap" is left except for insignificant local perforations, such as the individual local holes and / or smaller local holes mentioned above.

[0102] like Figure 1 Furthermore, as shown, in the clamping device, the shield formed by the magnetically conductive but non-conductive material does not end with the first outer cover 3.

[0103] Instead, at least one end of the first outer cover 3, and more preferably at both ends, are connected magnetic covers 3a, 3b made of the aforementioned material, which typically contact the first outer cover 3.

[0104] Preferably, the magnetic cover 3a is formed as an integral or preferably segmentally replaceable pole shoe on the end side of the induction coil 1 away from the tool holder 4, that is, it is formed as an annular structure with a central opening that forms a through portion 7 for the tool W to be inserted or removed.

[0105] Preferably, the magnetic cover 3b is designed as a flat annular disk on the end side of the induction coil 1 facing the tool holder 4, which ideally completely surrounds the winding of the induction coil 1 and has a central through-hole for the sleeve portion HP.

[0106] In order to further improve the shielding, and such as Figure 1 As shown, the induction coil 1 and its first outer cover 3 are surrounded by a second outer cover 9 on the outer periphery of the outer cover, in such a way that the first outer cover 3 and the second outer cover 9 touch each other, ideally touching each other on the main part or the whole part of their circumferential surfaces facing each other.

[0107] This second outer casing 9 is made of a non-magnetic but conductive material (such as aluminum).

[0108] "Conductive" here is understood not only as a material that conducts electricity only locally, so-called "at the granular level," but also as a material that allows eddy currents to form within a relevant range.

[0109] The second outer cover 9 is particularly characterized in that the second outer cover is preferably designed in such a way that, under the influence of the scattered field of the induction coil 1 passing through the second outer cover, eddy currents are generated in the second outer cover, which cause a weakening of the undesirable scattered field.

[0110] Furthermore, the second outer casing 9 is surrounded in its circumference by power semiconductor structural elements 10 (which will be described in detail below), which are arranged directly in recesses 11 on the outer periphery of the second outer casing 9 (identified only).

[0111] These power semiconductor structural elements 10 have two large main surfaces and four small side surfaces. The large main surfaces are preferably more than four times the size of each of the side surfaces.

[0112] The power semiconductor structural element 10 is arranged such that its large main surface is in thermally conductive contact with the second outer casing 9 (typically on its outer periphery), wherein the large main surface of the power semiconductor structural element 10 is bonded to the peripheral surface of the second outer casing 9 by means of a thermally conductive adhesive.

[0113] Each power semiconductor structure element in the power semiconductor structure element 10 has a different terminal for supplying voltage.

[0114] In addition, such as Figure 1 As shown, capacitors 14a and 14b are grouped around the induction coil 1 on its outer periphery.

[0115] Capacitor 14a preferably relates to a smoothing capacitor, which is a direct component of the power circuit; capacitor 14b preferably relates to an oscillating circuit capacitor, which is also a direct component of the power circuit.

[0116] To enable the capacitors 14a and 14b to be electrically connected, a plurality of circuit boards 15a and 15b are provided here, each of which surrounds the outer periphery of the induction coil 1.

[0117] Each of these circuit boards 15a, 15b is preferably configured as an annular disk. Each of these circuit boards 15a, 15b is preferably made of FR4 or a similar consumable material for the circuit board.

[0118] As in Figure 1 It can also be seen that the axis of rotational symmetry of each of the two circuit boards 15a and 15b, which are implemented here as ring circuit boards, is coaxial with the longitudinal axis L of the induction coil (and the tool holder 4 / tool ​​W).

[0119] The upper circuit board 15a of the two circuit boards carries smoothing capacitors 14a, whose terminal tabs pass through the upper circuit board 15a or are connected to the upper circuit board 15a by means of SMD technology, so that the smoothing capacitors 14a hang down from the upper circuit board 15a.

[0120] The lower circuit board 15b of the two circuit boards is constructed accordingly, with the oscillation circuit capacitor 14b rising vertically from the lower circuit board.

[0121] Intuitively, the following can be summarized: the power semiconductor 10 forms an imaginary first cylinder surrounding the induction coil 1; the capacitors 14a and 14b form an imaginary second cylinder surrounding the imaginary first cylinder; the capacitors 14a and 14b, which have low sensitivity to the scattered field, form an imaginary outer cylinder, while the power semiconductor structural element 10, which points to the mounting space with as little scattered field as possible, forms an imaginary inner cylinder.

[0122] like Figure 1 As further shown, the induction coil 1 is not "completely wound" along its entire length along its longitudinal axis L. Instead, the induction coil here consists of two windings, typically cylindrical in shape. Each of these windings forms an end side of the induction coil 1. These windings are spaced apart from each other, which is, exemplarily, at least 1.5 times larger than the length of each winding in the direction of the longitudinal axis L of the induction coil 1.

[0123] Such an induction coil 1 helps to reduce reactive power because it lacks windings in the "intermediate region" that are not necessarily needed from the perspective of heating the sleeve portion HP of the tool holder as efficiently as possible. However, if they were present, these windings tend to generate additional reactive power without making a truly significant contribution to heating.

[0124] In order to power induction coil 1 with minimal loss, a system is provided. Figure 2 The circuit is shown in detail in the image.

[0125] circuit

[0126] like Figure 2 As shown, the circuit has an oscillation circuit SKS (see...). Figure 2 ).

[0127] In the oscillating circuit SKS, the majority of the required energy oscillates periodically (at high frequency) between the induction coil 1 and the capacitor units 14a and 14b. Thus, in each cycle, or periodically, only the energy lost from the oscillating circuit due to its heat and other power losses needs to be replenished to the oscillating circuit SKS. This eliminates the previously very high losses.

[0128] Power electronic devices that feed induction coil 1 (such as...) Figure 2(As shown) It is fed on the input side by the generally available grid power NST, which is 400 V / 50 Hz in Europe (three-phase AC, 3f) (the corresponding value in other countries).

[0129] like Figure 2 As shown, the current drawn from the power grid is converted into DC current by rectifier G 21, and the DC current itself is smoothed by one or more smoothing capacitors 14a (not shown).

[0130] like Figure 2 As further shown, this DC current is used to feed the actual oscillating circuit SKS.

[0131] The power semiconductor structural element 10, the oscillation circuit capacitor 14b, and the induction coil 1 for retraction and extension constitute the backbone of the oscillation circuit SKS.

[0132] The oscillation circuit SKS is controlled or regulated by the control electronics SEK, which is fed DC current by the rectifier G.

[0133] Preferably, the power semiconductor structure element 10 is implemented by a transistor of the type "Insulated-Gate Bipolar Transistor", or simply IGBT.

[0134] The control electronics SEK switches the power semiconductor structure element 10 / IGBT at a frequency that is preset to the operating frequency set on the oscillation circuit SKS.

[0135] Importantly, the SKS oscillator circuit can never operate precisely at resonance.

[0136] This could lead to rapid damage to the power semiconductor component 10 due to voltage spikes. Instead, the control electronics SEK is designed to operate the power electronics or its oscillation circuit SKS within a preset operating range that is only close to the system's resonant or natural frequency.

[0137] Preferably, the oscillation circuit (via control device 20) is controlled or adjusted to suit a value within the range of 0.9 ≤ cos φ ≤ 0.99. A value within the range of 0.95 ≤ cos φ ≤ 0.98 is particularly advantageous. This again avoids voltage spikes and thus further promotes miniaturization.

[0138] Monitoring the clamping / heating process - overheat protection

[0139] In order to operate the clamping device with specific operational safety (as automated as possible), the clamping equipment is equipped with automatic heating control / regulation that enables automated clamping operation.

[0140] The heating control / regulation is achieved by a corresponding control or regulation device 20 in the clamping device, which also includes an overheat protection device for the tool holder 4 currently inserted for clamping (to prevent possible damage to the tool holder currently being clamped due to overheating).

[0141] For the control or regulation device 20, and also for the overheat protection device, the current and voltage present in the circuit are crucial, namely, in particular the coil current M-SpA, the coil voltage M-SpV, and the intermediate circuit current M-EA and / or the intermediate circuit voltage M-EV.

[0142] To measure the coil current (M-SpA), coil voltage (M-SpV), intermediate loop current (M-EA), and intermediate loop voltage (M-EV), such as Figure 2 As shown, the circuit is equipped with corresponding current / voltage measuring instruments M-SpA (coil current), M-SpV (coil voltage), M-EA (intermediate loop current), and M-EV (intermediate loop voltage), which are installed in the circuit in the positions shown in the figure and in the corresponding manner. Here, these measuring instruments provide the corresponding measurement values ​​at a sampling rate of 10 values ​​per second.

[0143] Here, if the quantity measured or its change in the induction coil used for clamping also depends on the temperature of the sleeve portion of the introduced tool holder, then this situation or these quantities can be advantageously used for neuron-based (see...) Figure 3 Neural Network 50 Figure 5 The decision tree (60) enables (automated) monitoring or heating control, thereby avoiding "human" errors and improving the safety of the hoop-shrinking equipment through automation.

[0144] In short, the aforementioned quantities (measurable by a measuring instrument): coil current M-SpA, coil voltage M-SpV, and intermediate loop current M-EA and / or intermediate loop voltage M-EV (and other characteristic quantities, detailed below) are input to neural network 50 (see [link to neural network 50]). Figure 3 (or decision tree 60,) Figure 5 In the first exemplary embodiment, the neural network 50 outputs the temperature HP-Temp of the sleeve portion of the tool holder as a possible output quantity.

[0145] In addition, neural network 50 (see Figure 3The remaining heating duration HP-t is also output as another possible output. The remaining heating duration HP-t output by the neural network 50 at a specific / current time point t describes how long the heating process must continue from that specific / current time point to ensure "optimal hooping". "Optimal hooping" is characterized by being non-destructive, allowing tool removal and insertion, and having the shortest possible execution time.

[0146] The feature quantities that are input into the neural network 50 as other input quantities are here called variations (Deltagrößen). These variations can be derived or calculated from the current and voltage quantities mentioned above, specifically: coil current variation Δ-M-SpA, intermediate loop current variation Δ-M-EA, energy variation Δ-Ener, and Kalman variation Δ-Kalm.

[0147] The additional characteristic quantity—coil current change Δ-M-SpA—is based on the derivative of the coil current; the additional characteristic quantity—intermediate loop current change Δ-M-EA—is based on the derivative of the intermediate loop current; the additional characteristic quantity—energy change Δ-Ener—is the energy value introduced into the tool holder or sleeve section to be clamped during the heating process (see integral of coil current); the additional characteristic quantity—Kalman change Δ-Kalm—is the output of a Kalman filter with the change of intermediate loop current M-EA as input.

[0148] Then, the induction heating controlled by the control or adjustment device 20 will continue until the first of the above output quantities reaches its corresponding shut-off index. That is, until the temperature HP-Temp of the sleeve portion of the currently introduced tool holder reaches the preset maximum temperature, or until the remaining heating duration HP-t drops to "remaining time is zero". Then, the induction coil will automatically shut off.

[0149] This "neural" control or regulation device 20 or overheat protection device can also be superimposed with an analytical control or regulation device 20 or overheat protection device. For example, other characteristic quantities can be used, such as the preset maximum heating time and / or the energy value (introduced to the tool holder or sleeve portion currently being clamped). To this end, the actual current heating duration is measured separately during the heating process, and the energy value introduced into the sleeve portion up to the current time point (the integral of the instantaneous coil current (active current) flowing since the start of heating) is determined. If the actual current heating duration reaches a limit that can be preset individually (for the tool holder currently being clamped), or the energy value reaches a limit that can be preset individually, then the heating process is terminated or the induction coil is turned off.

[0150] Other characteristic quantities mentioned in this application, such as inductance or phase angle, can also be used as other turn-off indicators accordingly.

[0151] Neural Network 50

[0152] Figure 3 The neural network 50 is shown.

[0153] The neural network 50 includes an input layer 51, 32 hidden layers 52 (in this example), and an output layer 53.

[0154] The aforementioned (feature) quantities, namely: coil current M-SpA, coil voltage M-SpV, intermediate loop current M-EA, and intermediate loop voltage M-EV, are input to the input layer 51, which has eight (input) neurons, and other feature quantities, namely: the change in coil current M-SpA, the change in intermediate loop current M-EA, the change in energy, and the Kalman change, are also input.

[0155] The neural network 50 determines the monitoring features: the temperature HP-T of the sleeve portion of the introduced tool holder and the current remaining heating duration HP-Rt. These monitoring features can be acquired at the output layer 53, which has two (output) neurons.

[0156] ReLU and Sigmoid are used as activation functions.

[0157] Training Neural Networks 50

[0158] To train the neural network 50, measurements are performed on a defined measuring cone or during the heating process of a measuring cone "hooped" by a hoop. Here, in addition to measuring current and voltage, the temperature of the sleeve portion of the measuring cone is measured by a temperature sensor. The current remaining heating duration is calculated by subtracting the heating time elapsed since the start of the current "hooping" from the maximum heating duration (of the current "hooping").

[0159] For the (entire) heating duration of the "shrinkage" process, measurements (current, voltage, temperature) and the current remaining heating duration are collected based on the sampling rate (10 values ​​per second in this case).

[0160] The resulting training data vectors are then used to train the neural network.

[0161] Alternative: In addition to the neural network 50 described herein (which provides the two output / monitoring features mentioned above, namely the temperature HP-T of the sleeve portion of the introduced tool holder and the current remaining heating duration HP-Rt), two parallel neural networks may also be used.

[0162] The two neural networks are input with the aforementioned (feature) quantities: coil current M-SpA, coil voltage M-SpV, intermediate loop current M-EA, and intermediate loop voltage M-EV, respectively. Additionally, they are input with the changes in coil current M-SpA, intermediate loop current M-EA, energy change, and Kalman change. The first neural network outputs the temperature HP-T of the sleeve portion of the tool holder, and the second neural network outputs the current remaining heating duration HP-Rt.

[0163] The two networks were trained in the same way as a single “total network” (neural network 50).

[0164] (Automatically) Identifies the tool holder / sleeve portion currently being introduced into the hoisting device or its induction coil.

[0165] In the induction coil (1, see) Figure 1 and Figure 2 The sleeve part (HP, see) Figure 1 Before the heating begins, the geometry or outer diameter of the sleeve portion to be introduced into the induction coil is determined (automatically) to "identify" the introduced sleeve portion.

[0166] This is achieved in particular by applying a test pulse with a known current magnitude, waveform, frequency, and duration to the induction coil before the actual induction heating process begins in the sleeve portion introduced into the induction coil. The time / current curve of the sleeve portion introduced into the induction coil is determined for the test pulse, and the time / current curve determined for the test pulse is used as a whole as a magnetic fingerprint of the sleeve portion introduced into the induction coil (determining its geometry).

[0167] For example, this is described in German patent applications 10 2015 016 831.2 and 10 2019 112 521.9, and European patent application 19 17 6562.7. The contents of these patent applications are the subject of this application (“incorporated”).

[0168] Figure 4 Exemplary examples show the time / current curves (curves 1, 2, and 3) of three (different) sleeve portions or tool holders (tool holders 1, 2, and 3) introduced into the induction coil under a defined test pulse. Figure 4 As can be seen, the time / current curves of the three (different) sleeve sections or tool holders introduced into the induction coil exhibit their own specific trends. Therefore, the corresponding sleeve section or tool holder can be inferred from the trend of the time / current curve.

[0169] Once the introduced sleeve section is identified, (pre-determined) heating parameters can be set for the introduced sleeve section or its heating process, especially including (corresponding) clamping / heating frequency.

[0170] Alternative: The neural network is also trained on the geometry or outer diameter of the sleeve portion introduced into the induction coil, and thus can also determine the geometry or outer diameter of the sleeve portion introduced into the induction coil.

[0171] Then, the heating process of the introduced sleeve section begins at its corresponding clamping / heating frequency, wherein, simultaneously with the induction coil being energized, measurements of coil current, coil voltage, intermediate circuit current, and intermediate circuit voltage also begin (see...). Figure 2 (Circuit in the middle).

[0172] Therefore, the aforementioned "neural network" monitoring and control achieves maximum automation or automatic clamping. During this process, no manual intervention is required to set operating parameters, thus saving the time previously necessary. Furthermore, automatic / automatic operation ensures adherence to high standards of operational safety and tolerances, guaranteeing that the structural unit operates in accordance with specifications. In addition, it effectively prevents overheating of the sleeve section.

[0173] Further applications of test pulses in heating control (characteristic quantity Δt and characteristic quantity ʃ) Figure 4 )

[0174] As mentioned above, Figure 4 Exemplary time / current curves of three (different) sleeve sections or tool holders introduced into the induction coil under a defined test pulse are shown.

[0175] like Figure 4 As shown, based on this time / current curve of a certain tool holder / sleeve section, two other characteristic quantities of the tool holder / sleeve section can be determined.

[0176] - Characteristic quantity Δt

[0177] The characteristic quantity Δt describes the point at which the coil current reaches its maximum value when a test pulse is applied (see [reference]). Figure 4 The time required for "max." (see Figure 4 , Δt1, Δt2, Δt3).

[0178] -Eigenvalues ​​ʃ

[0179] The characteristic quantity ʃ describes the situation in the case of the test pulse (see...). Figure 4 Test pulse 1), after the coil current reaches its maximum value, it drops to a predetermined point or other predetermined event (here, for the application of the next test pulse, see...). Figure 4During the test pulse 2), the area under the time / current curve (see test pulse 2) Figure 4 ,ʃ1、ʃ2、ʃ3).

[0180] Alternatively, for the characteristic quantity ʃ, the entire area under the time / current curve during the test pulse can also be used.

[0181] like Figure 4 As shown, the two characteristic quantities Δt and ʃ are unique and specific to a particular tool holder / sleeve section, and are therefore suitable for describing or defining a specific tool holder. This allows for the execution of specific control functions for the tool holder / sleeve section described by these two characteristic quantities Δt and ʃ in the induction heating process of the clamping device.

[0182] Its foundation is a database containing coil current measurements on test pulses during possible (successful) clamping processes of the clamping device in the adjustable coil position, on preset (almost) tool holders covering the entire product range.

[0183] Therefore, for each tool holder, these two characteristic quantities Δt and ʃ are determined for each possible (successful) coil clamping position. Each clamping process in these defined clamping procedures can be associated with a corresponding set of (successful) clamping / heating parameters (the “dataset” contains the tool holder, characteristic quantity Δt and characteristic quantity ʃ, coil position, clamping / heating parameters).

[0184] Then, the database can be used for specific control functions in the induction heating process of the hoop equipment.

[0185] - Check coil setup using characteristic quantities Δt and ʃ

[0186] When checking the coil setup for the clamping process of a (known) clamping tool holder, insert the tool holder into the clamping device and set the induction coil to the preset (successful) coil clamping position for that known tool holder.

[0187] Apply a test pulse, measure the coil current, and determine two characteristic quantities Δt and ʃ. Check the database to see if a dataset exists where the characteristic quantities Δt and ʃ are consistent with the (basic) characteristic quantities of the test pulse.

[0188] If it exists, the coil clamping position is set correctly; if the corresponding dataset cannot be found in the database, report that the coil clamping position is set incorrectly.

[0189] -Use the characteristic quantities Δt and ʃ to determine the hoop parameters

[0190] Here, the (unknown) tool holder is inserted into the clamping device, and the induction coil is adjusted to the possible (successful) coil clamping position (from the perspective of an experienced operator).

[0191] Apply a test pulse, measure the coil current, and determine two characteristic quantities, Δt and ʃ. Then, select the dataset from the database that best matches the characteristic quantities of the test pulse.

[0192] Once the matching dataset is identified, the clamping parameters for the (unknown) tool retainer can be extracted from it, and the tool retainer can be clamped (successfully) based on these clamping parameters.

[0193] Nevertheless, in addition to the inputs already described, the two features Δt and ʃ can also be (alone or in combination) fed into the neural network described above. The training data / dataset can be adjusted accordingly.

[0194] Monitoring the hoop / heating process - using overheat protection via decision tree 60 ( Figure 5 )

[0195] Figure 5 (Exemplary) Decision tree 60 is shown.

[0196] The decision tree 60 comprises three decision layers 61a, 61b, and 61c, and one outcome layer 62. According to its structure, the decision tree (in the outcome layer 62) provides eight outcome values ​​EGW1, EGW2, ..., EGW8, which are used to correct (63) the assumed initial / starting value AW of the temperature HP-T of the sleeve portion of the introduced tool holder to the (predicted) temperature of the temperature HP-T of the sleeve portion of the introduced tool holder.

[0197] In this example, one hundred decision trees 60 with the same structure were used, and their respective correction values ​​participated in the correction with a weight of 1% (corresponding to one hundred decision trees 60).

[0198] After correcting the initial / starting value of the temperature HP-T of the sleeve portion of the introduced tool holder by all decision trees and their correction values, the (predicted) temperature HP-T of the sleeve portion of the introduced tool holder at a specific heating time point can be obtained.

[0199] For each heating time point t, decision tree 60 receives ten (including measured) characteristic quantities KG and five additional characteristic quantities wKG (derived from them) of the induction process. For example, these (characteristic) quantities include coil current M-SpA, coil voltage M-SpV, intermediate loop current M-EA, and intermediate loop voltage M-EV, as well as additional characteristic quantities: changes in coil current M-SpA, changes in intermediate loop current M-EA, energy changes, and Kalman changes. Each decision tree 60 makes a decision for its node and provides the corresponding result values ​​EGW1, EGW2, ..., EGW8.

[0200] During the sensing process, a decision tree 60 provides feature quantities at a sampling rate of 30 ms, which then determines the temperature HP-T of the sleeve portion of the introduced tool holder for these time points in the manner described above.

[0201] If the temperature HP-T of the sleeve portion of the introduced tool holder exceeds the preset limit temperature, the induction process is terminated.

[0202] Training decision tree 60

[0203] The decision tree was trained using CART / Scikit-learn, with a pre-defined structure consisting of three decision layers and one outcome layer.

[0204] The training data is formed as a 16-row vector containing the aforementioned ten features KG and five additional features wKG, as well as correction values. This training generates multiple decision trees (each with trained node parameters (i.e., specific feature values) and their decision values) and correction values. One hundred “best” decision trees are selected from these multiple decision trees. This training also generates the aforementioned initial / starting values ​​for the temperature HP-T of the sleeve portion of the introduced tool holder.

[0205] For example, the selection criterion for the decision tree 60 could be the difference in parameters between the nodes. The more differentiated the trained decision tree 60 is, the more accurate and reliable the prediction of the temperature values ​​of various possible sleeve sections will be.

[0206] If necessary, normalized feature values ​​can be used, which can also be used as a result of the training process.

[0207] Although the details of the invention have been shown and described in detail by way of preferred embodiments, the invention is not limited to the disclosed examples, and other variations can be derived therefrom without departing from the scope of protection of the invention.

Claims

1. A method for monitoring the induction heating process of an induction coil structure unit, wherein a sleeve portion of a tool holder is inserted into the induction coil of the induction coil structure unit. Its features are, - Provide at least one, and in particular multiple, different characteristic quantities of the induction process, particularly coil current and / or coil voltage and / or input current and / or input voltage and / or intermediate loop current and / or intermediate loop voltage. - The at least one or more different feature quantities are input into a learning system, particularly a neural network, which uses the at least one or more different feature quantities to determine monitored feature quantities, such as: temperature features of the sleeve portion introduced into the induction coil and / or sleeve features of the sleeve portion introduced into the induction coil and / or time features of the heating process of the sleeve portion introduced into the induction coil, and The heating process is monitored by using the monitoring characteristics, such as the temperature characteristics and / or the sleeve characteristics and / or the time characteristics.

2. The method for monitoring an induction heating process according to at least one of the preceding claims, Its features are, - Provide and / or input at least one additional characteristic quantity to the learning system, such as: the instantaneous change of the coil current and / or the intermediate loop current, and / or the instantaneous rate of change of the coil current and / or the intermediate loop current, and / or the absolute value of the coil voltage and / or the absolute value of the coil current, and / or the energy of the coil current and / or the intermediate loop current flowing since the start of heating, and / or the integral of the instantaneous coil current (active current) and / or the intermediate loop current flowing since the start of heating, and / or the heating time since the start of heating, and / or the instantaneous inductance of the induction coil or the instantaneous inductance of the overall system consisting of the induction coil and the sleeve, and / or the instantaneous phase angle between the instantaneous coil current (active current) and the coil voltage driving the coil current, and / or the time required for the coil current or intermediate loop current to undergo a preset change, such as an increase (sustained) time, the learning system determines the monitored characteristic quantity by using the at least one additional characteristic quantity.

3. The method for monitoring an induction heating process according to at least one of the preceding claims, Its features are, The at least one or more different characteristic quantities and / or the at least one additional characteristic quantity are provided by measuring before and / or during the heating process, especially during the process, or by providing, in particular, stored or calculated values ​​for the at least one or more different characteristic quantities and / or the at least one additional characteristic quantity.

4. The method for monitoring an induction heating process according to any one of the preceding claims, Its features are, The features provided to the self-learning system are first processed by a digital smoothing filter, such as a Kalman filter or normalization, before being input into the learning system.

5. The method for monitoring an induction heating process according to at least one of the preceding claims, Its features are, When a certain feature quantity among the monitored feature quantity and / or the feature quantity input to the self-learning system reaches, exceeds, or falls below a set limit value, the heating of the sleeve portion is stopped (or at least changed).

6. The method for monitoring an induction heating process according to at least one of the preceding claims, Its features are, The temperature characteristic is the temperature of the sleeve portion of the tool holder, and / or the sleeve characteristic is the geometry of the sleeve portion or the tool holder, especially the tool holder size, or the classification information of the sleeve portion or the tool holder, and / or the time characteristic is the remaining heating duration.

7. The method for monitoring an induction heating process according to at least one of the preceding claims, Its features are, Use the identified monitoring characteristics to set or readjust heating and / or shutdown parameters.

8. An induction coil structure unit for performing the method according to at least one of the preceding method claims, the induction coil structure unit having an induction coil, wherein a sleeve portion of a tool holder can be introduced into the induction coil. Its features - A self-learning system that uses at least one or more different characteristic quantities according to claim 1 to determine the monitoring characteristic quantity, such as: the temperature characteristic quantity of the sleeve portion introduced into the induction coil and / or the sleeve characteristic quantity of the sleeve portion introduced into the induction coil and / or the time characteristic quantity of the heating process of the sleeve portion introduced into the induction coil.

9. The induction coil structure unit according to at least the preceding device claim, Its features are, - The self-learning system determines the monitored feature quantity by using at least one additional feature quantity as described in claim 2.

10. The induction coil structure unit according to at least one of the preceding device claims, Its features At least one current and / or voltage detector, the at least one current and / or voltage detector being located in the intermediate circuit of the frequency converter for supplying electrical energy to the coil circuit and / or in the coil circuit having the induction coil.

11. The induction coil structure unit according to at least one of the preceding device claims, Its features A comparator is used to compare at least one of the features, such as the monitored feature, and / or one of the features input to the self-learning system according to claim 1 and / or claim 2, with a preset limit value for this purpose.

12. The induction coil structure unit according to at least one of the preceding device claims, Its features An automatic shut-off device configured to shut off (or at least alter) the heating process via the induction coil when the limit value is reached, falls below, or exceeds in the comparison.

13. The induction coil structure unit according to at least one of the preceding device claims, Its features At least two, especially parallel, self-learning systems, which determine different monitoring characteristics, especially temperature and sleeve characteristics.

14. A self-learning system, particularly a neural network or decision tree, for monitoring induction heating processes. The self-learning system is configured for use - Use at least one or more different characteristic quantities as described in claim 1 to determine the monitoring characteristic quantity, for example: temperature characteristic quantity of the sleeve portion introduced into the induction coil and / or sleeve characteristic quantity of the sleeve portion introduced into the induction coil and / or time characteristic quantity of the heating process of the sleeve portion introduced into the induction coil.

15. A self-learning system according to at least the preceding claim, particularly a neural network or decision tree, The self-learning system is configured for use - The monitoring feature quantity is also determined using at least one additional feature quantity as described in claim 2.

16. A self-learning system, particularly a neural network, according to at least one of the preceding two claims. Its features Multiple hidden layers, especially more than 10 hidden layers, especially more than 15 hidden layers, especially more than 20 hidden layers, especially more than 25 hidden layers, even more than 30 hidden layers, and most importantly more than 32 hidden layers.

17. A self-learning system, particularly a neural network, according to at least one of the preceding three claims. Its features are, The self-learning system, especially the neural network, uses the ReLU function or the Sigmoid function as the activation function.

18. A self-learning system, particularly a decision tree, according to at least one of claims 14 to 16. Its features are, Multiple decision trees are used, especially decision trees with the same structure, and in particular, multiple neural networks are selected based on the differentiation index of the decision trees.

19. A method for training a self-learning system, particularly a neural network or decision tree, according to at least one of the preceding claims relating to the self-learning system. Its features are, - The self-learning system is trained using at least one or more different feature quantities according to claim 1 to determine the monitoring feature quantities, such as: temperature feature quantity of the sleeve portion introduced into the induction coil and / or sleeve feature quantity of the sleeve portion introduced into the induction coil and / or time feature quantity of the heating process of the sleeve portion introduced into the induction coil.

20. A method for training the self-learning system, particularly a neural network or decision tree, according to at least the preceding claim. Its features are, - Training is also performed using at least one additional feature quantity as described in claim 2.

21. A method for training the self-learning system, particularly a neural network or decision tree, according to at least one of the preceding two claims. Its features are, - Training is also performed using the monitored features, such as the temperature features and / or the sleeve features and / or the time features, wherein the monitored features used for training are determined, in particular, by employing a heating process performed in a defined manner, wherein the monitored features are measured during the heating process or determined based on quantities measured during the heating process.

22. A method for training the self-learning system, particularly a neural network or decision tree, according to at least one of the preceding three claims. Its features are, The training is performed using the backpropagation method, the CART algorithm, and / or the Scikit-learn algorithm.

23. A method for training the self-learning system, particularly a neural network or decision tree, according to at least one of the preceding four claims. Its features are, The self-learning system is retrained using user feedback and / or user interaction.

Images