RU2779855C2 - Method for optimization of performance and/or operational characteristics of fan or fan system - Google Patents

Abstract

FIELD: ventilation.

SUBSTANCE: method for optimization of performance and/or operational characteristics of a fan is proposed, while, based on part models characteristic for structural elements or functions, when taking as a base at least one algorithm, the model and data are reduced to obtain behavior models characteristic for structural elements or functions, while reduced data of behavior models in system modeling are bound or, respectively, combined to obtain a system behavior model with input and output values, while input values and corresponding output values of the fan from the system behavior model are provided to an optimizer for selection to achieve optimized system control depending on frame conditions.

EFFECT: obtaining a method for optimization of performance or operational characteristics of a fan or a fan system.

11 cl, 13 dwg

Description

The invention relates to a method for optimizing the efficiency and/or performance of a fan or fan system. In the end, we are talking about all optimizable parameters of the fan, for example also about the service life. Interactions with other ventilators, environmental conditions and/or people who are in or around the ventilator, respectively, also play a role. This also includes the so-called multi-objective optimization.

At this point, it should be noted that the more fans and/or external devices involved in the optimization, the greater the potential for efficiency gains. An example of this is load balancing in the case of a multi-fan system.

The invention is based on the fundamental idea, to provide at each operating point of the fan the best possible efficiency and the best possible performance. Due to conflicting operating parameters, this is difficult.

It is known from fan practice that the ball bearing and ball bearing lubrication are critical parameters for the life of the fan. The service life of a ball bearing and ball bearing lubrication is largely dependent on the operating temperature in or on the motor and the mechanical forces acting on the ball bearing. Since neither temperature sensors nor force sensors can be placed in the immediate vicinity of the bearing, neither the bearing temperature nor the bearing forces acting on the bearing can be measured. Therefore, it is required either to indirectly measure these parameters, or to determine them by calculation.

From DE 10 2010 002 294 A1 a system or method is known for determining the condition of a bearing in an electrical machine. Real sensor blocks determine the measured value, which is transferred to the simulation block. The simulation block determines the resulting value, which is either a bearing current value or a bearing current dependent value. The resulting value is passed to the next block for further calculation. The known system/method, due to the required sensor technology, is costly and, in the absence of sufficient structural space, can only be applied with difficulty to fans.

The invention is based on the creation of a digital display of a real fan, namely, by displaying its properties by means of mathematical calculation models and, if necessary, using known data, if necessary, real measurement data. The actual measurement data may be current measurement data from the current operating mode of each individual engine (and if necessary, its history). Also, taking into account the known relationships, characteristics, etc., at least one algorithm characteristic of the operational parameter is created and involved in further calculation.

According to the digital display by means of virtual sensors, the states of the structural elements of the fan are found or respectively calculated. These states of the structural elements are entered into an algorithm that takes into account either the specific operation or the operating parameters or the specific product, which determines and calculates the specific operating parameters of the fan from the operating states and, if necessary, gives forecasts related to the operation of the fan, for example, forecasts for the service life of the fan, derived from them. Essential is the possibility of combined use of certain states of structural elements and real measurement data.

This uses two different software components, namely a first software component relating to the digital twin and a second software component relating to a performance-specific algorithm, which can be called an "intelligent" algorithm.

The digital twin is a digital representation of a real, individual object, in the case of the fan or fan system proposed by the invention. The digital twin displays the properties of the ventilator through a calculation model and, if necessary, using known ventilator data. The task of the digital twin is to calculate the states of the fan structural elements depending on each operational state using virtual sensors. The states of the components determined from this calculation are fed into a performance-specific algorithm that determines/calculates from the performance parameters of the digital twin the performance parameters or operating states of the fan, such as bearing life and/or bearing grease life. On the basis of the result, a control adaptation appropriate to the situation is possible. Therefore, operating parameters and operating states are just as relevant when they are design values.

The above-described combination of a digital twin and a performance-specific algorithm with respect to a digital twin algorithm can be implemented on a microprocessor to be dedicated to a fan motor and thus be dedicated to that fan as a permanent component.

The digital twin algorithm is a combination of a digital twin describing a fan with a kind of intelligent algorithm, which is calculated taking into account the specific operating parameters.

With a properly designed fan, predictive maintenance can be performed to avoid fan failure due to, for example, a damaged bearing or bearing grease. We strive to adapt the system parameters to the appropriate situation, so that almost the maximum possible service life of the fan can be realized.

With the use of a digital fan display and a performance-specific algorithm, predictive maintenance aims to exhaust the life of the fan components as much as possible and at the same time avoid any fan failure. The service life of the fan is calculated by taking as a basis the design conditions of the structural elements and the resulting operating parameters.

The digital twin uses physical and/or mathematical and/or statistical and/or empirical and/or combined models to calculate the thermal and mechanical states of structural elements. In any case, they include both mathematical and physical and non-physical models. The performance-specific algorithm (intelligent algorithm) requires the states of the structural elements determined by the digital twin to determine any performance parameters, for example also to predict the failure of a fan. Since the service life of a fan is primarily dependent on ball bearings and ball bearing lubrication, a focus on ball bearing lubrication and ball bearing lubrication of the performance calculation plays a very important role.

It is known from practice that the service life of a bearing grease is largely dependent on the operating temperature. The higher the operating temperature during the entire service life, the faster the bearing lubricant is consumed. Therefore, the temperature of the bearing should be determined in order to determine the service life of the bearing grease.

To determine the bearing temperature, a temperature sensor would have to be placed in the immediate vicinity of the bearing. Due to the geometrical and functional conditions of the fan/motor, this is not possible. According to the method proposed by the invention, structural element states such as bearing temperature are calculated by means of a digital twin together with an algorithm specific to the operating parameter.

The calculation is based on a mathematical model, which, in turn, is based on a reduced coupled thermomagnetic calculation model. The combination of the digital twin, together with an algorithm specific to the operating parameter, calculates heat sources, heat sinks and the thermal state of the overall system touching the fan motor. Thus, using virtual sensors of the digital twin, the bearing lubrication temperature can be determined depending on the operating state of the fan/motor and included as an operating state in the algorithm characteristic of the operating parameter.

Both the digital twin, including its virtual sensors, and the performance-specific algorithm can be implemented in machine code (C-code) on an existing microprocessor, which gives the fan a certain machine intelligence.

The previous embodiments describe a method for determining fan operating states using a digital display (digital twin) of the fan using at least one operating parameter specific algorithm. This is the basis for the invention described below, which is based on the task of calculating operating states, which are determined by virtual sensors, based on a digital twin algorithm, and a workflow (Workflow) is defined to implement the digital twin algorithm in relation to the fan. In particular, real sensors for determining operational states should be abandoned.

In accordance with the invention, the above problem is solved using a method with the features of claim 1 of the claims, namely, a method for optimizing the efficiency and / or performance of the fan, while based on numerical detailed models characteristic of structural elements or functions, when taking as the basis of at least one algorithm, the reduction of the model and, at the same time, the data (data refinement) is carried out to obtain behavior patterns characteristic of structural elements or functions, while the reduced data of the behavior patterns in system modeling are associated or, respectively, combined with obtaining a behavior model a system having inputs and outputs, whereby the inputs and corresponding outputs of the fan from the system behavior model are provided to the optimizer for selection in order, depending on the framework conditions, to achieve an optimized system control.

In accordance with the invention, it is a digital twin algorithm in the context of an "intelligent fan" in relation to the improvement of the already considered twin algorithm.

The further development of the digital twin algorithm should be understood as an independent, situation-specific adaptation of the system parameters of the fan or, respectively, the fan system, in order to ensure the best possible efficiency and the best possible performance at each operating point.

According to the invention, numerical detailed models are first generated, for example regarding a thermal model, a magnetic circuit model or a model regarding the position of the blades and the flow or, respectively, the flow conditions. The detailed model can also be a digital twin according to an introductory explanation of the fan environment, such as a data center in the overall system context. Also, the detailed model may refer to the digital twin of the fan system. Other detailed models are possible.

In this way, detailed models can be generated, also concerning the condenser fan system. The fans are individually controlled by an optimizer who has access to a system behavior model to realize the best possible efficiency and homogeneous pressure/flow through the complementary condenser. Other detailed models are possible.

At the next stage, the reduction of detailed models occurs within the framework of the reduction of the model, namely, with the receipt of the so-called behavior models. This is accompanied by a significant reduction in the incoming data.

After that, in system modeling, there is a connection of behavior models with reduced data, the result of which is a behavior study containing a combined behavior model.

The modeling of the general system is carried out in the system space, which has a uniformly distributed combination of input values. The result is a table containing the input combinations and their corresponding output values of the system. This table reflects the behavior of the system, namely having the input values and the corresponding output values of the fan. Based on these values, optimization can take place.

In the current operating mode, the optimizer, depending on the environmental conditions, selects the best possible output values of the system, such as, for example, the efficiency of the system, in the table of the behavior model, preferably in real time. Once the best possible output value of the system has been found, the corresponding input values can be read from the table. With these input values, the system is controlled in the best possible way, preferably in real time.

In the light of the previous considerations, it is essential that the optimizer selects the optimal system efficiency from the system behavior table and supplies the necessary input quantities to the regulating system. In this way, continuous optimization is possible.

There are various possibilities to implement and improve the idea of the present invention in an advantageous manner. For this, on the one hand, reference can be made to the claims dependent on claim 1 and, on the other hand, to the following explanation of one of the preferred embodiments of the invention with the aid of a drawing. In connection with the explanation of the preferred embodiment of the invention with the aid of the drawing, general preferred embodiments and improved embodiments of the invention are also explained. The drawing shows:

1-13: steps of the method for implementing the idea of the invention, with particular differences, while the idea of the invention as an example is explained using figures 4 and 5.

1-3 serve to understand the idea of the invention and refer to the digital twin and the digital twin algorithm as the basis for a smart fan.

In a particular case, FIG. 1 shows the combination of a digital twin with at least one performance-specific algorithm, referred to below as a digital twin algorithm. This can be seen with the example of bearing and/or bearing lubrication life.

As already mentioned above, the service life of the bearing and bearing grease depends on the operating temperature and engine speed. Since a temperature sensor cannot be placed in the immediate environment of the bearing, the temperature of the bearing must be calculated by means of a model according to the invention with a digital twin algorithm, which is obtained from a combination of the digital twin and a performance-specific algorithm (intelligent algorithm).

The digital twin is nothing more than a mathematical model based on a reduced coupled thermomagnetic and mechanical calculation model. The digital twin calculates the thermal and mechanical state of the overall system in contact with the engine. With virtual sensors belonging to the digital twin, the digital twin can determine the temperature of the bearing lubricant depending on the operating state of the engine.

For further data processing, an intelligent algorithm needs the states of structural elements in order, for example, to predict the failure of a fan. From the failure characteristics, engine failure can be calculated, or at least estimated. All software related to the digital twin algorithm is implemented in machine code (C-code) on the engine's microprocessor, so no other electronics are required.

Figure 2 shows the sequence of calculating the service life of the bearing grease in the fan motor bearing. As part of creating a digital representation of a real fan, numerical detailed models are required, in a specific case, thermal models, magnetic circuit models, etc. In addition, algorithms are compiled to calculate the life of the lubricant.

After that, the detailed models are reduced to obtain behavior models so that the amount of data can be manipulated.

After that, the behavior models and the algorithm that calculates the life of the bearing grease in the system simulation are connected, namely, within the framework of the combination of a digital twin with a characteristic algorithm for the performance parameter, which in the present case calculates the life of the bearing grease. From this system simulation, a C-code is generated and the C-code is directly implemented on the motor microprocessor.

As mentioned above, the reduction of a detailed model to a behavior model is required in order to reduce the computation time. Thanks to this measure, the digital twin algorithm can be implemented on the engine microprocessor. To reduce the thermal model, different methods can be used, for example, the Krylov method. In this case, the data of the detailed model are reduced by lowering the order of the model.

The magnetic detailed model can be reduced using an algorithm or using a table. The table defines results that are pre-calculated for certain situations, so that complex calculations can be replaced by a quick search for values. Using suitably reduced models, the bearing lubrication temperature and the bearing temperature can be calculated. The calculated values use a performance-specific algorithm, here a bearing grease life calculation algorithm, to calculate the life of the bearing grease on the one hand and the bearing life on the other hand.

In addition, the used bearing life/bearing lubricant can be preferably exponentially weighted as a function of the operating temperature.

Figure 3 shows how such a weighting factor changes over the course of temperature change, using exemplary parameters such as long-term operation, bearing type, viscosity, rotational speed, grease temperature and operating time as the basis for calculating the life of the bearing grease. /life time. This calculation example, with a lifetime of four minutes, gives a usable lifetime of 15 minutes.

Reduced models according to the digital twin and performance-specific algorithm regarding bearing grease life are included in the system simulation and linked to each other. System simulation can be compiled, for example, in the MATLAB program. Using the MATLAB code generator, you can translate the system simulation into C-code and implement it on the engine's microprocessor.

Figures 4 and 5 show the individual steps of the method according to the invention, whereby figure 4 refers to the position of the angle of the blades, and figure 5 to the distribution of the load of the fans of the data center (data center). The corresponding steps of the method from the creation or, respectively, the preparation of a detailed model, a reduced model, a system simulation up to a behavior model are identical in both cases. Based on the behavior model, the optimizer selects the optimal system efficiency from the system command table and passes the corresponding input values to the controller, while the regulation is carried out in real time. This routine, along with the data, is generated in C code so that the optimization can run on a conventional processor.

In accordance with the image in figure 4, the angle of the fan blades must be adjusted so that the optimum efficiency of the system is obtained depending on the operating point. From the detailed model, when taking the appropriate algorithm as the basis, a reduced model is derived, from which, within the framework of system modeling, when taking several detailed models as the basis, a study of behavior or, accordingly, the behavior model resulting from it is obtained. The optimizer selects the optimal system efficiency from the system behavior table and passes the appropriate input values, with which optimization can be achieved, to the control mechanism. There is a regulation of the overall system in real time on the microprocessor, namely, when taking as a basis a behavior model and an algorithm serving to optimize. The data and programming algorithm is in C-code.

In accordance with the image in figure 5 we are talking about the distribution of the load system of several fans, in the selected embodiment of the distribution of the load of the fans of the data center (data center). The flow rate and thus the resulting load distribution of the individual fans must be adjusted so that the optimum system efficiency is obtained depending on the temperature prevailing in the data center. Here, too, the optimizer selects the optimum system efficiency from the system behavior table and passes appropriate input values to the control mechanism so that the overall system can be controlled in real time on the microprocessor. And here the fan model data is entered into the optimization algorithm, while the program runs in C-code on conventional processors.

The method according to the invention makes it possible, based on the data reduction discussed above, to produce a compact C-code that can run on standard microprocessors. A kind of data refinement takes place on the microprocessor (Big Data → Smart Data → smart data), and this is the result of the calculation. Only compressed refined data is further processed or, for example, sent to the cloud. Of course, this greatly reduces (reduces) the volume of streaming communications to the cloud.

Also, the performance parameters determined on the basis of the digital twin and performance-specific algorithm can be used for predictive maintenance and preventive maintenance of the fan on the one hand, and to optimize the design and operation of the fan, on the other hand, while in accordance with the idea of the claimed invention, the digital twin algorithm is improved for an independent, appropriate situation, adaptation of system parameters, namely, in order to provide the best possible efficiency and the best possible performance at each operating point.

The image in Fig.6 and 7 serves to further explain the idea of the claimed invention, which has already been explained as an example with reference to Fig.4 and 5.

Essential is the use of a digital display, namely a digital fan twin. The digital twin is obtained through data processing. It is obtained, on a case-by-case basis, from a combination of known input values or sensor measurement values with calculated values and calculations/models. On the basis of the digital twin, the temperature of structural elements, currents, losses, etc. is determined in certain specified places of the fan. On the basis of the digital twin, real values are virtually determined, such as, for example, specific temperatures of structural elements, even if there is no economically/structurally feasible measurement possibility by means of sensor technology at this specific location of the fan.

Also important for the idea of the claimed invention is the algorithm characteristic of the operational parameter. On the basis of the results or data provided by the digital twin, for example the bearing temperature, indicators such as, for example, the probability of failure or the used life of the fan or fan bearing, respectively, are determined. These figures depend on the current operating parameters of the fan and their history, i.e. in which operating points and environments the fan is/was running.

FIGS. 6 and 7 illustrate with a specific example the method of the invention for determining fan operating conditions using a digital fan display based on the foregoing considerations.

In the left column of Fig.6 are the measured and calculated input values, including the units of measurement marked on the arrows. These input quantities are measured with available standard sensors or are known.

From these input values, heat sources and heat sinks are calculated. It is based on simulation-based models that take into account heat sources such as losses in copper, iron, electronics, and heat sinks such as engine cooling (cooling wheel, circulating air and ambient temperature). This results in input quantities for a reduced thermal model having virtual sensors. All this corresponds to the digital twin in relation to the thermal model.

From the reduced thermal model with virtual sensors, the temperatures of structural elements are calculated. At the same time, the thermal model displays the physics of the fan and, using virtual sensors, calculates the temperature in the bearing, in the winding, on the magnet and in various electronic components, depending on the need.

Figure 7, as a continuation of figure 6, clearly shows that the output values from the reduced thermal model, if necessary, having other parameters, are used as input values for calculating the aging process. The underlying aging models are based on historical data and can be embedded in the form of characteristics. So the aging-limited residual life can be individually calculated or adjusted on site based on the actual history of the fan and the current operating condition.

The output of these aging calculation models is an estimated lifetime in days or hours, which, taken separately, can be pure information. Then the information related to this can be involved in a further forecast, namely, in the forecast of the remaining service life of individual structural elements or the entire fan. This prediction can then be used to intelligently optimize the residual life. In order to extend the remaining service life, measures can be taken, such as speed reduction or intelligent load sharing among several fans. These measures can be associated with any control value.

Figure 8 once again shows the digital twin down to a reduced thermal model having virtual sensors, the fan being mapped to it along with the motor. As mentioned above, the thermal model displays the physics (physical characteristics) of the fan and calculates with the help of virtual sensors different temperatures, which are involved for various purposes/purposes/applications, in keywords as follows.

For control : Determining operational parameters using virtual sensors and using them for control. These can be: warning messages, status LEDs, comments in a readable error code, displays in the cloud or software application, an image on the user interface.

For Predictive Maintenance : A way to determine the aging of a fan made up of multiple individual systems such as ball bearings, windings, electronic components, magnets, and predict remaining life. Use, for example, for scheduling maintenance intervals, to achieve the longest possible service life before a maintenance interval (i.e. not too early maintenance), automatic maintenance scheduling, maintenance notification, automatic ordering of spare parts.

For optimisation : a way to determine operational states relating to product characteristics, i.e. efficiency, temperatures of structural elements, rotational speed, output power, flow, noise, vibrations, etc.

To create an intelligent fan: reaction to certain operating conditions to refine behavior or to achieve certain goals:

- changing the operating point / changing the control parameter for optimal efficiency;

- changing the operating point to achieve the longest possible service life;

- reduction of rotational speed when the probability of failure is very high;

- changing the operating point during the day-night rhythm for the quietest possible night application;

- output of a control value for additional devices or customer devices, for example, output of a temperature for use in the control of a heat pump or for additional cooling;

- purposeful bypass of the critical states of the system (for example, resonances, elevated temperatures, etc.).

For a better understanding of the ideas of the present invention are important, on the one hand, the sequence of steps of the method, and on the other hand, their content. The sequence of each of the steps of the method can be derived from the development workflow of the underlying algorithm. This algorithm is shown in Fig. 10, and the last step can be an improvement of the method.

To create a detailed model, the following are fundamentally relevant:

A model is a reflection or approximation of reality, that is, by definition, an approximation. The model is always limited to some fragment that is interesting for the desired display. In addition, the model is fundamentally incomplete, as either it is reduced for easier application with respect to its required inputs, or individual physical behaviors are not known when the model is built. Depending on the later use and goal setting, a different kind of simulation is needed, ie, for example, a different area to be considered, a different accuracy in the results required, or a speed of calculation. There are many types of models, and in the technical field, most often the model is associated with a mathematical representation, for example, with algebraic equalities or inequalities, systems of ordinary or partial differential equations, spatial representations of states, tables, graphs.

An integral part of today's product development is virtual product development using Finite Element Simulation (FE) simulation. Classically, here the physical domain (eg, strength or heat or magnetic circuit) is mapped into a very large (on the order of 100 Gigabytes) and computationally intensive model, and the results are defined in this model in millions of places (nodes). This is one of the detailed models. A rough sequence of these detailed models can be outlined as follows:

1. import of 3D geometry, for example from the CAD area,

2. assignment of boundary conditions, i.e. permanent fixings, material definitions, contact conditions (glue points, sliding joints, thermal insulation),

3. union (dividing the geometry into a million small elements that are connected to each other),

4. creating a load, that is, forces, sources / receivers of heat, magnetic fields.

5. automatic solution of the resulting differential equations for each individual element and reduction into one large result for the general model,

6. evaluation of results.

Compilation of detailed models with virtual sensors in relation to fans/general systems with fans is as follows.

Detailed models are made to be able to represent the physics of the fan and/or the overall system. So-called virtual sensors are defined calculation points in detailed models. These virtual sensors calculate the states of structural elements, such as winding temperature in a thermally detailed fan model. Detailed models are simulation models that are costly in terms of calculation time, required processing power, and required memory. Detailed models such as, for example, thermal models, magnetic circuit models, electronics models, control models, force models or vibration models are used to calculate non-linear operating states. The physical effects of the system involve interactions between domains, so the individual models must be considered as they are related to the overall system. Calculations using detailed models in a common system are impractical in terms of calculation time, since interactions cannot be evaluated in real time. Therefore, model reduction is required.

Reduced models can be created in the following way.

Model reduction describes an already existing model in general terms in order to reduce other information, for example, to optimize it for memory requirements or computational speed. Here, depending on the specific application, there are many options, such as for example:

- approximation of simple mathematical functions, such as, for example, polynomial functions, and only storing coefficients in memory;

- bookmarking tables for different input values and then either applying these discrete values or interpolating between values;

- Approximation of statistical models that give predictions from previous values;

- graphs/logic diagrams, example: if T˃200°C, then the fan is faulty.

Creating Reduced Models - Example a)

The starting point for the reduced model is the thermal FE model, which displays the temperatures at each point in the model as a function of heat input and heat removal. In the following example, the reduction of the model is simplified to only one heat input and one heat rejection, only one detectable temperature at location A, and only "high" and "low" respectively. To do this, a study of the parameters is carried out, in which this so-called "reference table" is determined:

Temperature at location A heat input=low=1 W Heat Dissipation=High=11W Heat Dissipation=Low=1W 40°C 80°C Heat Dissipation=High=5W 20°C 60°C

After that, there are several possibilities for using the results:

- direct or discrete use of the table. Example: when a temperature of 4 W for Heat Input and 1 W for Heat Rejection is predicted at location A, the value of 40°C is directly assumed;

- using a table and linear interpolation between values. Example: when a temperature of 5 W for Heat Input and 1 W for Heat Rejection is predicted at location A, a value of 60°C is assumed by linear interpolation;

- using a table to determine the temperature prediction function through regression. Examples of objective functions are polynomial functions, linear functions, exponential functions, statistical functions, differential equations, etc. Then determine the temperature using this function.

Creating reduced models - example b)

The starting point for the reduced model is the thermal FE model, which displays the temperatures at each point in the model as a function of heat input and heat removal. After that, with the help of mathematical assumptions, calculations and transformations (for example, the LSS system (linear stationary system) or the Krylov subspace method), compact spatial models of states can be approximated. These models consist of two essential differential or, respectively, integral equations and four matrices describing the entire system (for example, 200×200 matrices filled with scalar numerical values). However, they no longer represent temperatures in millions of knots, but only in a small number of selected locations. In addition, due to approximation, there is a deviation of the results depending on the size of the spatial model of the state. In principle, it is believed that the larger the model and its matrices, the smaller the deviation.

Spatial state models are standardly available as procedures, modules, or objects in many computer programs such as Matlab, or in programming languages, so such models can be calculated by simply importing matrices. The input values here are, for example, the thermal power that enters the system and the heat sinks by convection, the output values are, for example, determined component temperatures (eg three different component temperatures).

Creating reduced models - example c)

The starting point for model reduction in this example is the results of experiments. Here, as in example a), a table would be created from the measurement results and then dealt with in an equivalent way (discrete use, linear interpolation, or regression by means of mathematical functions.

Relationships between physical domains or different models can have additional value.

Classically, in virtual product development, domains are considered separately, since the overall consideration is very computationally and memory intensive and impractical. Due to the reduction of models, it becomes possible to link models of different domains. For example, the connection of a detailed model of a magnetic circuit, the computation time of which on a high-performance computing cluster ranges from several days to weeks, with a thermal model is undesirable. In many cases, it is necessary to represent the real behavior as accurately as possible. When the detailed models are reduced, the required computing power and the need for memory become lower, and thus it becomes possible to link the models taking into account the economic aspect. In many cases, it is necessary to represent the real behavior as accurately as possible.

Communication of physical domains or different models - examples:

- the resistance of the winding is approximately linearly dependent on the temperature of the copper. Depending on the resistance of the winding, the power loss in the winding varies approximately linearly. Depending on the power loss, the thermal behavior, for example the temperature of the winding and the bearing, changes strongly non-linearly, which in turn affects the resistance of the winding. Thus, here, depending on the requirement for the results of the model, a connection is needed;

- the required torque and fan speed are highly dependent on the resistance of the installation and, for example, the pressure difference and temperature of the discharged medium. Depending on the load moment, the behavior of the magnetic circuit changes, i.e. winding currents, magnetic field, speed, etc. Then, depending on this, the power consumption, losses and achievable speed also change. And here, if the customer's requirements are taken into account, it is possible to link the behavior of the fan with the installation situation, depending on the application.

Specification of the technical implementation - example a):

- compilation of a thermal FE model of a fan → a computationally intensive and memory intensive FE model with 1,000,000 additional elements for the polynomial function. Heat sources and sinks are displayed as polynomial functions depending on the input current and speed;

- compilation of a reduced thermal model by statistical methods, which displays the temperature of the electronic components depending on the input current and speed → a polynomial function that describes the temperature depending on the input current and speed = virtual temperature sensor;

- characteristic from the data sheet on the service life of an electronic component depending on its temperature → an algorithm specific to the operational parameter, which, using a virtual temperature sensor, calculates the probability of failure;

- use of preventive maintenance to control or to optimize the operating point → intelligent algorithm.

Specification of the technical embodiment - example b):

- registration of the indicator of current and frequency of rotation of the engine by means of the integrated electronics/regulation. From here, the electromagnetic operating point is derived;

- Based on this operating point, the losses of the motor and power electronics are obtained by means of look-up tables or polynomial functions;

- the thermal model processes the loss values and determines the temperatures of critical system components such as ball bearings or semiconductor components;

- at the same time, vibrations of structural elements are recorded by means of a real sensor. Local vibrations are virtually projected by behavior models onto the overall system, as a result of which, for example, the load on the bearing from vibrations is estimated;

- with the help of algorithms specific to the operational parameter, certain temperatures and vibration values are recalculated into an assessment of the service life of structural elements and the fan;

- thus, further measures such as preventive maintenance can be ensured;

- at the same time, knowing the losses, the operating point and the efficiency of the system can be optimized by adaptations carried out according to the control technology, such as, for example, by varying the preliminary control angle.

For applications in fans and/or fan systems, the previous considerations about reducing detailed models are valid, and order reduction can be carried out using the Krylov subspace method. The goal is to minimize the computation time, the computational power required and the memory requirement so that real-time computation is possible. The virtual sensors are saved and give output values.

In accordance with Fig.9 in the next step, the reduced models are associated with obtaining a system model of the fan. In a specific case, reduced models such as a thermal model, a magnetic circuit model, a software model, an electronics model, etc. are associated with obtaining a fan system model. The fan system model displays the physics of a single fan or fan group or fan system and calculates, depending on the environmental and operating conditions, the efficiency, performance and possible interactions between the individual models.

At the next stage, the system models of the fan are linked to the model of the installation, as a result of which a general system model is obtained. The overall system model consists of several fans and one unit, for example including a compressor and/or a condenser. The installation model can be implemented with the same workflow as the fan system model. The fan system models and the plant model can then be linked to produce one common system model.

The next stage involves the study of behavior, namely, the calculation of response values using combinations of input parameters.

The aim of this study is to determine the behavior of the overall system model and use this knowledge to regulate the system in real time.

In accordance with the behavior of the overall system in the design space, the influences and influences of the input values of the models on the response values of the models are transferred or displayed, respectively.

The design space is a multidimensional space that is given by the possible input quantities. The number of input quantities corresponds to the number of dimensions of the design space. With ten input values, this means ten dimensions.

The input values of the models vary within certain limits. In this case, combinations of parameters arise that uniformly cover and thus describe the multidimensional space. Model responses such as efficiency and performance are calculated from combinations of parameters. The study of behavior provides a design space that is filled with response values depending on the input values. This space represents the behavior of the overall system.

Figure 10 shows how different input values are fed into the fan system models and the cooling circuit models, resulting in a model of the overall cooling system. The corresponding output value must be understood as the result of a model of the overall cooling system. The resulting knowledge according to FIG. 11 can be translated into a behavior table. If the behavior of the overall system is known, the input quantities can be adapted to obtain the best possible response quantity.

Adaptation or selection of the response value and the corresponding combination of input values from the behavior table is implemented by the optimizer, namely, in accordance with the next step of the method shown in Fig. 10, after which the optimizer selects the best response value and at the same time selects the best possible combination of input parameters for current operational state.

Referring to FIG. 12, the optimizer selects the best response value and determines the best possible combination of inputs for the current operational state. In other words, the optimizer, depending on the environment/operational state, chooses the best possible response model. The combinations of parameters of the input values corresponding to it are adjusted. In this way, the system can be adjusted in the best possible way. The General Cooling Behavior Table can run on any processor, preferably a fan microprocessor that is available anyway. This is how regulation is possible.

Fig. 13 relates to one of the possible advanced embodiments, according to which the table of behavior of the general cooling system in accordance with the image in Fig. 13 is expanded using system modeling of the cooling loop in the cloud. The system includes fan digital twins, cooling loop digital twins, a validator and a virtual regulator/optimizer.

The digital twins of the fan and cooling circuit physically represent the system. The virtual controller has knowledge of the table of behavior of the overall cooling system in accordance with the image in Fig.13. In addition, the virtual controller can be trained by machine learning, for example, to take into account specific customer requirements. The validator improves the digital twins by comparing the nominal value and the actual value. Thus, the system is able to model specific customer patterns and refine them according to the resulting information.

Finally, it should be directly pointed out that the above-described embodiments serve only to consider the theory for which protection is sought, however, it is not limited to these embodiments.

Claims (18)

1. A method for optimizing the efficiency and/or performance and/or service life of a fan, wherein, based on numerical detailed models characteristic of structural elements or functions, taking at least one algorithm as a basis, the model is reduced and the the reduction of data with the receipt of behaviors characteristic of structural elements or functions, at the same time, the reduced data of behavior models in system modeling are associated or, respectively, combined with obtaining a model of system behavior that has input and output values, while the input values and the corresponding output values of the fan from the system behavior model are fed to the optimizer, by means of the optimizer, depending on the framework conditions, the optimal output value of the said system behavior model is selected and the corresponding output value is supplied to the controller for optimized system control. 2. The method according to claim 1, characterized in that the detailed models include a thermal model, and/or a magnetic circuit model, and/or a blade position and flow model, and/or an environment model, and/or a digital twin model of one or multiple fans, and/or environmentally friendly overall system. 3. The method according to claim 1 or 2, characterized in that the reduced model data is provided in the form of tables or data matrices. 4. Method according to any one of claims 1 to 3, characterized in that the data of the system behavior model - the input values and the corresponding output values of the fan - are provided in the form of a table or data matrix. 5. A method according to any one of claims 1 to 4, characterized in that the optimizer framework conditions are fan environment conditions, depending on which the best possible system output value, for example system efficiency, is selected in a table or matrix preferably in real time. , under the condition of which the corresponding input value is read from the corresponding table, with the help of which the regulation is then carried out. 6. The method according to any one of claims 1 to 5, characterized in that the angle of the blades is adjusted so that the optimal system efficiency is obtained depending on the operating point, while using the optimizer, the optimal system efficiency is selected from the system behavior table and submit appropriate input values for regulation. 7. Method according to any one of claims 1 to 6, characterized in that the flow rate and thus the required load distribution of the individual fans of the fan system, for example the load distribution of the fans of the data center, is adjusted so that an optimum is obtained in the data center, in depending on the temperature, the efficiency of the system, while by means of the optimizer the optimal efficiency of the system is selected from the table of system behavior and the corresponding input values are fed for regulation. 8. A method according to any one of claims 1 to 7, characterized in that the optimized regulation is translated into C-code, thereby providing real-time regulation on a standard microprocessor. 9. The method according to any one of claims 1 to 8, wherein the basis for obtaining detailed models is the determination of the operational states of the fan using a digital display - a digital twin - of the fan and at least one algorithm specific to the operational parameter, and the determination includes the following steps of the method: create a digital display of a real fan by reproducing its properties through mathematical calculation models and, if necessary, known data; create mentioned at least one characteristic of the operational parameter of the algorithm, taking into account the known ratios, characteristics; calculate the state of the structural elements of the fan on the digital display by means of virtual sensors; the states of the structural elements are transferred to the algorithm, which calculates the operating parameters of the fan from the states of the structural elements. 10. The method according to claim 9, characterized in that the basis of the digital display as a mathematical and/or physical and/or empirical and/or statistical, or as a combined model is taken thermal and/or mechanical calculation model and/ or a magnetic circuit calculation model, wherein the corresponding calculation model is converted from the corresponding detailed model to the reduced model. 11. A method according to any one of claims 1 to 10, characterized in that the data reduction is a data refinement.

Images