Machine learning training method for dish-washing machine draining pump control and dish-washing machine

Abstract

The invention discloses a machine learning training method for dish-washing machine draining pump control and a dish-washing machine, and relates to the technical field of application of machine learning in a control system, and the machine learning training method comprises the steps of data acquisition, data processing and model training. A temperature detection part and a current detection part are arranged in the drainage pump and are used for detecting coil temperature and coil current; controlling the drainage pump to operate at different operation currents, circularly acquiring coil currents and coil temperatures under each operation current, and correspondingly forming a current-temperature combination according to timestamps; and the multi-stage neural network performs fitting training on the combined data to obtain a model for inputting the current current and the current temperature and outputting a temperature change predicted value of a later cycle period. The dish-washing machine comprises a controller provided with the model, and the controller circularly obtains the coil current and the current temperature, obtains the temperature change prediction value and further obtains the prediction temperature. The accuracy of draining pump temperature detection is improved, the safety and reliability of dishwasher operation are improved, and the service life of the draining pump is prolonged.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of application of machine learning in control systems, in particular, to a machine learning training method for dishwasher drain pump control and a dishwasher. BACKGROUND

[0002] The implementation of motor temperature protection function, the current market mainstream scheme is still using the way of external temperature sensor, the defect of this kind of way is that temperature sensor can only monitor local surface temperature, can not reflect the real temperature rise state inside the motor. When the coil inside the motor generates high temperature due to overload, the hysteresis of external sensor will cause the delay of protection action, which can not prevent the burning of insulation layer in time.

[0003] Part of high-end products use built-in temperature sensor, but need to damage the motor structure for installation, not only increase the production cost, but also affect the sealing and reliability of the motor. In addition, there is also a method of estimating temperature based on thermal model in the prior art, but this kind of method depends on accurate motor parameters and complex calculation model, and in actual application, the estimation error is often increased due to parameter drift and environmental change.

[0004] Most products still adopt over-temperature shutdown mode for temperature protection, that is, directly stop when the temperature exceeds the preset threshold, although it can prevent overheating damage, but frequent shutdown seriously affects the running efficiency of the equipment, and cannot adapt to the temperature change demand under different working conditions.

[0005] The above information disclosed in the background technology is only used to increase the understanding of the background technology of the present application, therefore, it can include the prior art known by those skilled in the art. SUMMARY

[0006] The present application aims at the problem that the coil temperature of the drain pump of the dishwasher in the prior art cannot be accurately detected and controlled, and obtains a model for predicting the temperature of the drain pump through machine learning training, detects and controls the coil temperature of the drain pump, improves the accuracy of the temperature detection of the drain pump, improves the safety and reliability of the operation of the dishwasher, and prolongs the service life of the drain pump.

[0007] In order to achieve the above-mentioned application / design purpose, the present application adopts the following technical scheme to achieve it: A machine learning training method for dishwasher drain pump control, comprising data acquisition, data processing and model training; During the data acquisition, a temperature detection part and a current detection part are arranged in the drain pump of a plurality of dishwashers, respectively, for detecting the coil temperature and coil current of the drain pump; the drain pump is controlled to operate at different operating currents, and the coil current and coil temperature under each operating current are obtained in cycles; the cycle period is the first period; The data processing forms current-temperature combinations corresponding to the obtained coil current and coil temperature by time stamp; The model training is a multi-level neural network fitting training on the current-temperature combination data, obtaining a model inputting current and temperature, and outputting temperature change prediction value after the first period.

[0008] In some specific embodiments, the maximum current and the minimum current are obtained according to the working condition of the drainage pump; A plurality of different operating currents are obtained between the maximum current and the minimum current; A protection temperature is set; During the data acquisition, the drainage pump is controlled to operate at the operating current, and the coil current and the coil temperature are cyclically acquired at the first period; when the coil temperature is higher than the protection temperature, the operating current of the drainage pump is controlled to be reduced and continue to operate, and the coil current and the coil temperature are continuously cyclically acquired at the first period.

[0009] In some specific embodiments, a normal operating temperature is set; During the data acquisition, if the coil temperature after the operating current of the drainage pump is reduced is lower than the normal operating temperature, the operating current of the drainage pump is controlled to return to the initial value.

[0010] In some specific embodiments, the current-temperature combination includes the coil current and the coil temperature at the same time and the coil temperature after the first period; The predicted temperature after the first period is equal to the sum of the temperature change prediction value after the first period and the current temperature.

[0011] In some specific embodiments, the architecture of the neural network is a single hidden layer neuron topology; The single hidden layer neuron adopts a Sigmoid activation function to construct a nonlinear mapping space; the output layer adopts a linear activation function; and a small batch random gradient descent optimizer and an L2 regularization strategy are introduced in the training process.

[0012] In some specific embodiments, it further includes test calibration, which includes: An error threshold is set; A second period is set, which is greater than the first period; the speed of the drainage pump is controlled to switch from high to low or from low to high at the second period; Cyclically acquire the coil current and the coil temperature with the first period as a cycle; input the coil current and the coil temperature into the model to obtain the temperature change prediction value; adjust the parameters of the neural network so that the error between the temperature change prediction value at the same time and the change value of the coil temperature is within the error threshold, and obtain a calibrated model.

[0013] A dishwasher comprises a drain pump, a variable frequency module, a controller, and a current detection unit; the drain pump is connected to the variable frequency module; the variable frequency module is connected to the controller, and the controller controls the variable frequency module to increase or decrease the frequency of the drain pump; the current detection unit is connected to the drain pump and the controller respectively, and is used to detect the coil current and transmit the coil current to the controller. The controller is configured with the model obtained by the machine learning training method, and is configured to: Cyclically acquire the coil current and the current temperature, and input the coil current as the current current and the current temperature into the model to obtain the temperature change prediction value in the first period from the current time; and obtain a predicted temperature from the sum of the temperature change prediction value and the current temperature.

[0014] In some specific embodiments, the controller is further configured with a protection temperature and a normal operation temperature, and is configured to: Compare the predicted temperature with the protection temperature; if the predicted temperature exceeds the protection temperature, control the drain pump to operate at a decreased frequency; during the operation of the drain pump at a decreased frequency, compare the predicted temperature with the normal operation temperature; if the predicted temperature is lower than the normal operation temperature, control the drain pump to operate at an increased frequency.

[0015] In some specific embodiments, the controller is further configured to: Acquire the coil current multiple times in the first period, calculate the average value of each coil current, and input the average value as the current current into the model to output the temperature change prediction value after the next first period.

[0016] In some specific embodiments, the current temperature is the cumulative sum of the first coil temperature and each temperature change prediction value. Apply a direct current voltage to the drain pump before starting, and obtain the coil current; the first coil temperature is obtained according to the quotient of the direct current voltage and the coil current and the relationship between resistance and temperature.

[0017] Compared with the prior art, the advantages and positive effects of the present application are: The machine learning training method for the dishwasher drain pump control of the application trains the model by collecting the coil current and the corresponding coil temperature, analyzes the dynamic relationship between the coil current and the temperature rise and fall of the coil during the operation of the drain pump, obtains the model of inputting the current temperature and outputting the temperature change prediction value after the first period, and accurately estimates the temperature rise or fall trend and quantity of the drain pump under specific coil current and current temperature environment without traditional sensors, breaking the monitoring blind area limit and accuracy defects. At the same time, since the coil current changes adaptively to the load change, temperature warning and cooling control strategy are realized through temperature rise prediction during sudden high load, irreversible damage caused by overheating is avoided, the stability and reliability of the dishwasher drain pump operation are improved, and the service life is prolonged.

[0018] Other features and advantages of the present application will become more apparent from the following detailed description of the application when read in conjunction with the accompanying drawings. BRIEF DESCRIPTION OF DRAWINGS

[0019] In order to more clearly illustrate the technical solutions in the embodiments of the application, the drawings needed in the embodiments will be briefly introduced below. Obviously, the drawings in the following description are some embodiments of the application, and other drawings can also be obtained by those skilled in the art without creative labor on the basis of these drawings.

[0020] Figure 1 is a flowchart of the training method for the dishwasher drain pump control based on machine learning according to the embodiment; Figure 2 is a temperature change and control diagram of one kind of operating condition of the drain pump during data collection according to the embodiment; Figure 3 is a temperature change and control diagram of one kind of operating condition of the drain pump during data collection according to the embodiment; Figure 4 is a temperature change diagram of one kind of operating condition of the drain pump during data collection according to the embodiment; Figure 5 is a flowchart of the machine learning training method for the dishwasher drain pump control according to the embodiment; Figure 6 is a flowchart of model test calibration according to the embodiment; Figure 7 is a flowchart of model test calibration according to the embodiment; Figure 8 is a dishwasher control flowchart according to the embodiment; Figure 9 is a dishwasher control flowchart according to the embodiment. DETAILED DESCRIPTION

[0021] The technical solutions in the embodiments of the present application will be apparently and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work belong to the protection scope of the present application.

[0022] In the description of the present application, it should be understood that the terms "center", "upper", "lower", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer" and the like indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, and are only for the purpose of facilitating the description of the present application and simplifying the description, and do not indicate or imply that the device or element referred to must have a particular orientation, be constructed and operated in a particular orientation, therefore, it cannot be understood as a limitation to the present application.

[0023] In the description of the present application, it should be noted that unless otherwise explicitly specified and limited, the terms "mounting", "connecting", "connection" should be understood in a broad sense, for example, it can be fixed connection, or detachable connection, or integral connection. For those of ordinary skill in the art, the specific meaning of the above terms in the present application can be understood according to the specific circumstances. In the description of the embodiments, specific features, structures, materials or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0024] The terms "first", "second" are only for the purpose of description, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features limited by "first", "second" can explicitly or implicitly include one or more of the features.

[0025] In the description of the present application, unless otherwise specified, the meaning of "multiple" is two or more.

[0026] Referring to Figure 1 , Figure 5 The present application discloses a machine learning training method for dishwasher drain pump control and a dishwasher. The model trained by machine learning is solidified into the controller of the dishwasher to participate in the operation control of the dishwasher.

[0027] The training method for dishwasher drain pump control based on machine learning includes data acquisition S1, data processing S2 and model training S3.

[0028] Data collection is to collect data for training the control model. Specifically, a temperature detection part, which can be a temperature sensor, is arranged in the drain pump of each dishwasher to detect the coil temperature of the motor coil of the drain pump. During data collection, the drain pump of each dishwasher is controlled to operate at different operating currents, and the coil current and coil temperature under each operating current condition are collected in a cycle. The cycle period is the first period.

[0029] That is, the drain pump of each dishwasher is controlled to operate at different operating currents in a cycle; that is, the drain pump of each dishwasher is controlled to operate at the same operating current; the operating current includes a plurality of different values. The coil current and coil temperature are collected in a cycle under each operating current condition for each dishwasher.

[0030] Data processing is to form a current-temperature combination according to the time stamp of the collected coil current and coil temperature; of course, the abnormal values are cleaned up before forming the current-temperature combination to prevent the abnormal values from affecting the results and accuracy of the trained model.

[0031] Model training is to train the obtained current-temperature combination data through a multi-level neural network to obtain a model with current current and current temperature as input and temperature change prediction value after the current time for a first period as output.

[0032] The machine learning training method for dishwasher drain pump control of the present application inputs the coil current and corresponding coil temperature into model training to analyze the dynamic relationship between the coil current and the temperature rise and fall of the coil during the operation of the drain pump, and obtains a model with current current and current temperature as input and temperature change prediction value after the first period as output. Without traditional sensors, the temperature rise or fall trend and quantity of the drain pump under a specific coil current and current temperature environment can be accurately estimated, breaking through the limitations of monitoring blind spots and accuracy defects. At the same time, since the coil current changes adaptively to the load change, temperature warning and cooling control strategy can be implemented through temperature rise prediction during sudden high load, avoiding irreversible damage caused by overheating, improving the stability and reliability of the operation of the dishwasher drain pump, and prolonging the service life.

[0033] The specific process and principle of the machine learning training method for dishwasher drain pump control of the present application will be described in detail below through specific embodiments.

[0034] In some specific embodiments, with reference to Figure 1 、 Figure 2 、 Figure 3 、 Figure 4 、 Figure 5, the running current in the data acquisition is set by the following method: obtaining the maximum current and the minimum current according to the working conditions of the drainage pump; for example, the running current of the drainage pump under the full load and high lift working condition is the maximum current; the maximum current can be the locked-rotor current of the drainage pump; the running current of the drainage pump under the no-load and low lift working condition is the minimum current; the running current of the drainage pump in the data acquisition is selected between the maximum current and the minimum current.

[0035] The data acquisition comprises: setting the protection temperature and the normal running temperature, controlling the drainage pump to run at the running current, and cyclically acquiring the coil current and the coil temperature of the drainage pump according to the first period; cyclically comparing the coil temperature with the protection temperature, when the coil temperature exceeds the protection temperature, controlling the drainage pump to reduce the running current, and continuing to cyclically acquire the coil current and the coil temperature of the drainage pump according to the first period, and comparing the coil temperature with the normal running temperature; when the coil temperature drops below the normal running temperature, controlling the drainage pump to restore to the initial running current.

[0036] The reduction of the running current can be realized by controlling the drainage pump to reduce the frequency, or directly reducing the running current. The initial value of the running current can be restored by increasing the running current or increasing the frequency.

[0037] The training data of the machine learning training method of the drainage pump control of the embodiment is no longer only the single data of the temperature rise during running or the temperature drop when stopping, but after the coil temperature reaches the protection temperature, the frequency and speed are reduced for protection, and the coil current and the coil temperature after the frequency reduction are continuously acquired, so as to ensure that the model can more comprehensively learn the dynamic relationship between the coil temperature and the coil current, and improve the prediction accuracy.

[0038] In some specific embodiments, the value of the running current in the data acquisition can be a plurality of current values uniformly selected between the maximum current and the minimum current; when the coil temperature exceeds the protection temperature, the running current is sequentially reduced according to the arrangement order of the running currents; that is, when the drainage pump runs at one of the running currents, if the coil temperature exceeds the protection temperature, the drainage pump is controlled to run at the running current adjacent to and smaller than the running current.

[0039] After the running current is reduced, the coil current and the coil temperature are cyclically acquired, the coil temperature is compared with the protection temperature and the normal running temperature; if the protection temperature is exceeded, the running current is continuously reduced; if the normal running current is lower, the running current is restored to the initial value.

[0040] Of course, the control of the drainage pump to reduce the running current can also be performed in a manner that the proportion of the reduction of the running current is less than 1.

[0041] In some specific embodiments, with reference to Figure 2 , when the operating current of the drain pump is a large current, i.e., the drain pump operates at an operating current close to the maximum current, the coil temperature increases with the extension of the operating time; when the coil temperature rises to the protection temperature or above, the operating current of the drain pump is controlled to decrease, i.e., the drain pump is controlled to operate at a reduced frequency.

[0042] The heat dissipation efficiency of the motor coil of the drain pump is related to the temperature of the motor coil itself, i.e., the higher the temperature of the motor coil, the higher the heat dissipation efficiency; the lower the temperature of the motor coil, the lower the heat dissipation efficiency. Therefore, the drain pump will correspond to a unique temperature balance state when operating at each operating current.

[0043] When the drain pump operates at a high operating current, the balance temperature of the drain pump is high, which is greater than the protection temperature; when the obtained coil temperature reaches the protection temperature to control the drain pump to operate at a reduced frequency, the heat generation efficiency of the drain pump decreases, while the heat dissipation efficiency is temporarily unchanged, the coil temperature of the drain pump decreases; the heat dissipation efficiency decreases; with the decrease of the heat dissipation efficiency, the coil of the drain pump will appear a small amplitude of temperature rise, the coil temperature rises with the decrease of the heat dissipation efficiency and decreases with the increase of the heat dissipation efficiency, and the balance state of alternating temperature rise and temperature drop appears.

[0044] In the balance state of temperature rise and temperature drop, if the coil temperature does not completely reach the normal operating temperature or below, the frequency or operating current of the drain pump is maintained. If the coil temperature completely reaches the normal operating temperature or below, the operating frequency or operating current of the drain pump is restored to the initial value, which is the best operating frequency or best operating current of the drain pump under this working condition.

[0045] The training data of the machine learning training method of the drain pump control of the embodiment includes the dynamic coil temperature of the temperature balance state, to ensure that the model can more comprehensively learn the dynamic relationship between temperature and current, so that the drain pump in the balance state can still accurately estimate the temperature change prediction value, thereby improving the accuracy of the predicted temperature; by controlling the operation of the dishwasher through the predicted temperature, the stability of the dishwasher operation is improved, and the user experience is improved.

[0046] In some specific embodiments, with reference to Figure 3 , when the operating current of the drain pump is a medium current, i.e., the drain pump operates at an operating current located at the middle position between the maximum current and the minimum current, the coil temperature slowly increases with the extension of the operating time; when the coil temperature rises to the protection temperature or above, the operating current of the drain pump is controlled to decrease, i.e., the drain pump is controlled to operate at a reduced frequency.

[0047] The heat dissipation efficiency of the motor coil of the drain pump is related to the temperature of the motor coil, that is, the higher the temperature of the motor coil, the higher the heat dissipation efficiency thereof; the lower the temperature of the motor coil, the lower the heat dissipation efficiency thereof; after the drain pump is operated at a reduced frequency, the heat generation efficiency is reduced, the heat dissipation efficiency is temporarily unchanged, the temperature of the motor coil is reduced, and with the reduction of the temperature of the motor coil, the heat dissipation efficiency thereof is reduced, the cooling speed of the motor coil is reduced to the opposite direction of slight temperature rise, and a dynamic balance is reached; if the normal operating temperature or below is reached before the dynamic balance of the coil temperature, the drain pump is controlled to be operated at the initial value.

[0048] In some specific embodiments, with reference to Figure 4 , when the operating current of the drain pump is a small current, that is, the drain pump is operated at an operating current close to the minimum current, the coil temperature slowly rises with the extension of the operating time; when it rises to a stable position, the temperature of the motor coil is below the protection temperature, and the drain pump is controlled to be operated at the current operating current.

[0049] In some specific embodiments, with reference to Figure 1 、 Figure 5 , the current-temperature combination includes the coil current at the same time, the coil temperature, and the difference between the coil temperature of the next cycle period and the coil temperature of the previous cycle period; that is, the current-temperature combination is composed of the coil temperature of the current cycle, the coil current, and the temperature change value after the next cycle period, the prediction variable is the coil temperature and the coil current of the current cycle, the response value is the temperature change value after the next cycle period, and the current-temperature combination is a two-dimensional array according to the same feature and the same column.

[0050] Then the input of the model is the current temperature and the current current, and the output is the temperature change prediction value after the first period; then the temperature prediction value after the first period is equal to the sum of the current temperature and the temperature change prediction value.

[0051] The machine learning training method of the dishwasher drain pump control of the embodiment makes the model for predicting the temperature change value of the next cycle period according to the coil current and the current temperature have generality by setting the response value of the training data as the temperature change value, improves the accuracy of the predicted temperature, and further improves the accuracy of the control, the stability and reliability of the drain pump operation, and prolongs the service life of the drain pump.

[0052] In some specific embodiments, with reference to Figure 1 、 Figure 5 , the architecture of the neural network is a single hidden layer neuron topology; the single hidden layer neuron adopts a Sigmoid activation function to construct a nonlinear mapping space; the output layer adopts a linear activation function; and a small batch stochastic gradient descent optimizer and an L2 regularization strategy are introduced in the training process.

[0053] The machine learning training method of the dishwasher drain pump control of the embodiment drives modeling through data relationship, trains based on actual operation data of the drain pump, and effectively improves adaptability and prediction accuracy of working conditions. The architecture of the neural network adopts a single hidden layer neuron topology; for example, 10 hidden layer neurons are adopted to achieve optimal balance between model complexity and computational efficiency. The hidden layer neurons adopt a Sigmoid activation function to construct a nonlinear mapping space, and the continuous derivable characteristics significantly enhance the nonlinear fitting capability of the model to multidimensional features, and are particularly suitable for complex regression analysis of temperature rise curves. The output layer adopts a linear activation function to maintain dimensional consistency of physical quantities. The training process introduces a mini-batch size=32 random gradient descent optimizer and an L2 regularization strategy to effectively suppress overfitting. Through k-fold cross-validation (k=5) and multi-condition pressure testing, the model exhibits excellent performance in key indicators such as mean square error (MSE=0.15℃) and determination coefficient (R²=0.97), confirming the reliability of its engineering application.

[0054] In some specific embodiments, with reference to Figure 5 、 Figure 6 , the machine learning training method of the dishwasher drain pump control further comprises test calibration S4, which comprises: S41, set a second period and an error threshold; the second period is greater than the first period; S42, control the speed of the drain pump to switch from high to low or from low to high with the second period as the cycle period; S43, collect the current and temperature of the motor coil of the drain pump, i.e., the coil current and coil temperature, with the first period as the cycle period; S44, input the coil current and coil temperature into the model to obtain a temperature change prediction value for the next collection period; S45, obtain the coil temperature of the next collection period and the temperature change value compared with the coil temperature of the adjacent previous collection period, and calculate the error between the temperature change prediction value of the next cycle period and the temperature change value of the next collection period; S46, determine whether the error is within the error threshold range; if yes, execute S47; if no, execute S48; S47, execute the output calibration model; S48, adjust the parameters of the neural network for the next round of calibration.

[0055] In some specific embodiments, with reference to Figure 5 、 Figure 7 , the machine learning training method of the dishwasher drain pump control further comprises test calibration S4, which comprises: S41, set a second period and an error threshold; the second period is greater than the first period; S42, control the rotation speed of the drain pump to switch from high to low or from low to high in the second cycle as a cycle period; S43, collect the current and temperature of the motor coil of the drain pump in the first cycle as a cycle period, that is, collect the coil current and coil temperature of the drain pump; S44, input the coil current and coil temperature into the model to obtain the temperature change prediction value in the next collection cycle; S45, obtain the predicted temperature of the motor coil in the next collection cycle and the coil temperature, and compare the error between the predicted temperature in the next cycle and the coil temperature in the next collection cycle; S46, determine whether the error is within the error threshold range; if yes, execute S47; if no, execute S48; S47, execute the output calibration model; S48, adjust the parameters of the neural network for the next round of calibration.

[0056] In some specific embodiments, with reference to Figure 5 、 Figure 6 、 Figure 7 , the error within the error threshold range is a continuous setting number of tests, and the error meets the error threshold requirement.

[0057] For example, the setting number is set to 5.

[0058] With reference to Figure 8 、 Figure 9 , the application also discloses a dishwasher comprising a drain pump, a frequency conversion module, a controller, a temperature detection part, and a current detection part. The motor of the drain pump is a frequency conversion motor connected with the frequency conversion module. The frequency conversion module is connected with the controller to control the change of the operation frequency or operation current of the drain pump. The temperature detection part is arranged near the coil of the drain pump and connected with the controller to detect the coil temperature and transmit it to the controller. The current detection part is connected with the drain pump and the controller to detect the coil current and transmit it to the controller.

[0059] The controller is configured with a model obtained by the machine learning training method for dishwasher drain pump control, to predict the temperature change value of the motor coil in the next cycle, that is, to obtain the temperature change prediction value in the next cycle. The sum of the temperature change prediction value and the current temperature obtains the predicted temperature in the next cycle.

[0060] The controller is further configured to: The coil current is input as the current, and the last predicted temperature is input as the current temperature.

[0061] The dishwasher of the present application effectively solves the technical defects of poor temperature detection accuracy in traditional temperature protection by configuring a model for predicting temperature changes according to real-time coil current, improves the running safety and stability of the drain pump, reduces maintenance costs, and prolongs the service life of the drain pump.

[0062] In some specific embodiments, with reference to Figure 8 、 Figure 9 , the controller is also configured with a protection temperature and a normal running temperature S100, and is configured to: S200, compare the predicted temperature with the protection temperature; if the predicted temperature exceeds the protection temperature, execute S300; S300, control the drain pump to run at a reduced frequency; S400, compare the obtained predicted temperature with the normal running temperature when running at a reduced frequency; if yes, execute S500; S500, if the predicted temperature is lower than the normal running temperature, control the drain pump to increase the frequency to the original running speed. And return to S200, continue to predict the temperature and compare and judge with the protection temperature.

[0063] The dishwasher of the present embodiment effectively reduces the temperature of the motor coil by adopting a control strategy of reducing the speed after monitoring that the predicted temperature of the motor coil exceeds the threshold value, prevents overheating damage. At the same time, after the over-temperature frequency reduction, the temperature of the drain pump and the drainage function are still continuously monitored to ensure that the temperature returns to the normal running temperature to restore to the normal working state, realize the balance of efficient operation and safety protection of the drain pump, avoid frequent shutdown affecting efficiency and service life of the drain pump, and improve user experience.

[0064] In some specific embodiments, the amplitude of the frequency reduction is the product of the original running frequency and the proportion less than 1.

[0065] In some specific embodiments, with reference to Figure 8 、 Figure 9 , the coil current is collected multiple times in the cycle period of the first cycle, and the average value of each coil current is calculated, which is the current current of the current cycle period, improving the stability and accuracy of the coil current, and further improving the accuracy of the predicted temperature.

[0066] In some specific embodiments, with reference to Figure 8 、 Figure 9 , the current temperature value is the cumulative sum of the first coil temperature and each temperature change prediction value.

[0067] The initial coil temperature is obtained by applying a DC voltage to the drain pump before starting the drain pump and obtaining the coil current under the DC voltage. Specifically, the resistance of the coil at the ambient temperature is obtained from the quotient of the DC voltage and the coil current under the DC voltage; the ambient temperature, i.e., the initial coil temperature, is calculated from the relationship between the resistance and the temperature.

[0068] For example, the initial coil temperature is calculated by calculating the drain pump coil resistance. Before the drain pump is started, a DC voltage is applied to the coil for a period of time. The coil voltage U and coil current I are measured, and the quotient is the drain pump coil resistance R. The relationship between the resistance of pure copper and temperature can be described by the formula R = R0 (1 + αT), where R0 is the resistance at a reference temperature, α is the temperature coefficient of copper (approximately 0.00393 / °C), and T is the temperature. The current temperature, i.e., the initial coil temperature, is inferred from the coil resistance.

[0069] That is, the control process of the dishwasher is as follows: S10, determine whether the drainage pump is turned on; if so, execute S20; S20, applying a DC voltage to the drainage pump to obtain a coil current under the DC voltage; S30, calculating the coil resistance from the DC voltage and the coil current; obtaining the initial coil temperature from the coil resistance and the relationship between resistance and temperature, which is the current coil temperature; S40, drainage pump starts; S50, sampling the coil current multiple times to obtain an average value; S60, obtain the temperature change prediction value of the next acquisition period by inputting the average value and the current temperature into the model, and then obtain the predicted temperature of the next acquisition period, which is the current temperature of the next acquisition period; return to S50 and loop to execute the temperature change value prediction.

[0070] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for a person skilled in the art to modify the technical solutions described in the aforementioned embodiments, or to replace some of the technical features therein with equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions claimed to be protected by the present invention.

Claims

1. A machine learning training method for dishwasher drain pump control, characterized in that: include: Data acquisition: a temperature detection unit and a current detection unit are set in the drain pumps of multiple dishwashers to detect the coil temperature and coil current of the drain pumps respectively; the drain pumps are controlled to operate at different operating currents, and the coil current and the coil temperature under each operating current are cyclically obtained; the cycle period is the first cycle; Data processing, the obtained coil current and the obtained coil temperature are formed into a current-temperature combination according to the timestamp; Model training: a multi-level neural network is trained to fit the data of the current-temperature combination to obtain a model that inputs the current current and the current temperature and outputs a predicted value of the temperature change after the first cycle.

2. The machine learning training method according to claim 1, characterized in that Obtaining a maximum current value and a minimum current value according to the working condition of the drainage pump; obtaining a plurality of different operating currents between the maximum current value and the minimum current value; Set the protection temperature; During the data collection, the drainage pump is controlled to operate at the operating current, and the coil current and the coil temperature are collected in the first cycle; When the coil temperature is higher than the protection temperature, the operation current of the drainage pump is controlled to be reduced while the drainage pump continues to operate, and the coil current and the coil temperature are continuously collected in the first cycle.

3. The machine learning training method according to claim 2, wherein: Set the normal operating temperature; During the data collection, if the coil temperature of the drainage pump after the operating current is reduced is lower than the normal operating temperature, the operating current of the drainage pump is controlled to return to an initial value.

4. The machine learning training method according to claim 3, wherein: The current-temperature combination includes the coil current and the coil temperature at the same moment and the coil temperature at the same moment after the first cycle; The predicted temperature after the first cycle is equal to the sum of the predicted temperature change value over the first cycle at the current moment and the current temperature.

5. The machine learning training method according to claim 1, wherein: The architecture of the neural network is a single hidden layer neuron topology; The single hidden layer neurons use the Sigmoid activation function to construct a nonlinear mapping space; the output layer uses a linear activation function; and the training process introduces a small batch stochastic gradient descent optimizer and an L2 regularization strategy.

6. The machine learning training method according to any one of claims 1 to 5, characterized in that: Also included is test calibration, which includes: Set the error threshold; Setting a second cycle that is greater than the first cycle; controlling the rotation speed of the drainage pump to switch from high to low or from low to high based on the second cycle; The coil current and the coil temperature are obtained in a cyclical manner with the first cycle as a period; the coil current and the coil temperature are input into the model to obtain the temperature change prediction value; the parameters of the neural network are adjusted so that the error between the temperature change prediction value and the change value of the coil temperature at the same time is within the error threshold, thereby obtaining a calibration model.

7. A dishwasher, characterized in that: The system comprises a drainage pump, a frequency conversion module, a controller, and a current detection unit; the drainage pump is connected to the frequency conversion module; the frequency conversion module is connected to the controller, and the controller controls the frequency conversion module to increase or decrease the frequency of the drainage pump; the current detection unit is connected to the drainage pump and the controller respectively, and is used to detect the coil current and transmit it to the controller; The controller is configured with a model obtained by the machine learning training method according to any one of claims 1 to 6, and is configured to: The coil current and current temperature are acquired cyclically, and the coil current is input into the model together with the current temperature as the current current to obtain a predicted value of temperature change over the first cycle from the current moment; and the predicted temperature is obtained by the sum of the predicted value of temperature change and the current temperature.

8. The dishwasher according to claim 7, characterized in that The controller is also configured with a protection temperature and a normal operating temperature, and is configured as follows: Compare the predicted temperature with the protection temperature; if the predicted temperature exceeds the protection temperature, control the drainage pump to operate at a reduced frequency; during the drainage pump's reduced frequency operation, compare the predicted temperature with the normal operating temperature; if the predicted temperature is lower than the normal operating temperature, control the drainage pump to operate at a higher frequency.

9. The dishwasher according to claim 8, characterized in that The controller is further configured to: The coil current is obtained multiple times in the first cycle, and an average value of each coil current is calculated; and the average value is input into the model as the current current to output the temperature change prediction value after the next first cycle.

10. The dishwasher according to any one of claims 7 to 9, characterized in that The current temperature is the cumulative sum of the initial coil temperature and each of the temperature change prediction values; A DC voltage is applied to the drainage pump before startup, and the coil current is obtained; the initial coil temperature is calculated based on the quotient of the DC voltage and the coil current and the relationship between resistance and temperature.

Images