A washing machine
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- HISENSE(SHANDONG)REFRIGERATOR CO LTD
- Filing Date
- 2025-01-24
- Publication Date
- 2026-07-24
Smart Images
Figure CN122446475A_ABST
Abstract
Description
Technical Field
[0001] Some embodiments of this application relate to washing machine technology. In particular, they relate to a washing machine. Background Technology
[0002] With technological innovation and improved living standards, people have increasingly higher demands for washing machines in terms of intelligence, functionality, and ease of use. However, the more complex the design of a washing machine, the more numerous and varied its components, and the higher the probability of failure. The drain pump is particularly critical. Users can only perceive a malfunction when the drain pump is completely damaged, meaning the washing machine can no longer drain water. Other causes of drain pump failure remain undetected, complicating subsequent repairs. Summary of the Invention
[0003] Some embodiments of this application provide a washing machine that can detect in advance whether the drain pump is malfunctioning, thereby improving maintenance efficiency.
[0004] To achieve the above objectives, some embodiments of this application provide a washing machine, including:
[0005] The drum body includes an outer drum and an inner drum disposed inside the outer drum, wherein a washing chamber for washing clothes is formed inside the inner drum;
[0006] A drain pump, connected to the outlet of the inner cylinder, is used to drain the water from the inner cylinder;
[0007] A water level sensor is used to detect the water level in the inner cylinder and output a water level signal;
[0008] A controller, electrically connected to the water level sensor and the drainage pump, is used to control the drainage pump to discharge water from the inner cylinder;
[0009] The controller is configured as follows:
[0010] The water level sensor obtains at least two water level signals in at least two detection cycles.
[0011] The at least two water level signals are input into a neural network model, and the neural network model outputs a detection result on whether the drainage process of the drainage pump is abnormal in the next detection cycle.
[0012] If the test results indicate that the drainage pump is malfunctioning during the drainage process in the next test cycle, an alarm message will be output to prompt the user to check the drainage pump.
[0013] The neural network model is obtained by training an initial network model using a training dataset, which includes water level signals when the drainage pump malfunctions.
[0014] In the aforementioned washing machine, by inputting at least two water level signals from at least two detection cycles into a neural network model, a detection result is output to indicate whether an abnormality will occur during the drainage process of the drain pump in the next detection cycle. An alarm message is then output based on the detection result, allowing the user to detect whether the drain pump is malfunctioning in advance and improving maintenance efficiency.
[0015] In some embodiments, the neural network model includes an interconnected gated loop module and a result output module, wherein:
[0016] The gated loop module is used to receive the water level signal of the current detection cycle, obtain the feature value of the water level signal of the current detection cycle based on the water level signal of the current detection cycle and the water level signal of the previous detection cycle, and output the feature value of the water level signal of the current detection cycle.
[0017] The result output module is used to receive and obtain the detection result based on the characteristic value of the water level signal in the current detection period.
[0018] It is understandable that by setting up a neural network model structure that includes a gated loop module, the water level signals of the current detection cycle and the previous detection cycle can be combined to output and obtain accurate detection results based on the feature values of the water level signals of the current detection cycle. The overall structure is simple and can improve training efficiency.
[0019] In some embodiments, the feature values of the water level signal in the current detection cycle include a first feature value and a second feature value. The gated loop module includes an update gate, a reset gate, and a calculation unit. The update gate and the reset gate are respectively connected to the calculation unit, wherein:
[0020] The reset gate is used to control the combination ratio of the water level signal of the previous detection cycle and the water level signal of the current detection cycle, so as to obtain and output the first feature value.
[0021] The update gate is used to control the degree of influence of the water level signal of the previous detection cycle on the feature value of the water level signal of the current detection cycle, and to obtain and output the second feature value.
[0022] The calculation unit is configured to receive the first feature value and the second feature value, and obtain and output the feature value of the water level signal of the current detection period based on the first feature value, the second feature value, the water level signal of the previous detection period and the water level signal of the current detection period.
[0023] It is understandable that by resetting the gate and updating the gate simultaneously, the feature value of the water level signal in the current detection period can be obtained based on the water level signal in the current detection period and the water level signal in the previous detection period. This allows control over the degree of fusion between the water level signals in the current detection period and the previous detection period, thereby obtaining a more accurate feature value of the water level signal in the current detection period and improving training efficiency.
[0024] In some embodiments, the computing unit is configured to:
[0025] Based on the first feature value, the water level signal of the previous detection period, and the water level signal of the current detection period, the candidate feature corresponding to the current detection period is obtained;
[0026] The feature value of the water level signal in the current detection period is obtained based on the candidate feature, the second feature value, and the water level signal of the previous detection period.
[0027] It is understandable that by combining the first feature value, the second feature value, the water level signal from the previous detection period, and the water level signal from the current detection period through the computing unit, the first feature value output by the update gate, the second feature value output by the reset gate, and the input data can be combined to improve training efficiency.
[0028] In some embodiments, the reset door is configured to:
[0029] The first matrix is linearly transformed according to the preset first weight matrix to obtain the target value. The first matrix includes the feature value of the water level signal in the previous detection period and the value of the water level signal in the current detection period.
[0030] The first feature value is obtained based on the activation function and the target value.
[0031] It is understandable that by performing a linear transformation on the first matrix using the first weight matrix, the corresponding first eigenvalue is obtained. Thus, by changing the specific value of the first weight matrix, the first eigenvalue can be changed, thereby controlling the combination ratio of the water level signal from the previous detection period with the water level signal from the current detection period, and improving training efficiency.
[0032] In some embodiments, the update gate is configured as follows:
[0033] The target value is obtained by linearly transforming the second matrix according to the preset second weight matrix. The second matrix includes the feature value of the water level signal in the previous detection period and the value of the water level signal in the current detection period.
[0034] The second feature value is obtained based on the activation function and the target value.
[0035] It is understandable that by performing a linear transformation on the second matrix using the second weight matrix, the corresponding second eigenvalue is obtained. Thus, by changing the specific values of the second weight matrix, the second eigenvalue can be changed, thereby controlling the degree of influence of the water level signal from the previous detection period on the water level signal of the current detection period and improving training efficiency.
[0036] In some embodiments, the result output module includes at least one fully connected layer, which is used to receive and output the detection result based on the characteristic value of the water level signal in the current detection period.
[0037] Understandably, by using at least one fully connected layer to classify the feature values of the water level signal in the current detection period, the detection results can be obtained, thus improving the reliability of the neural network model.
[0038] In some embodiments, when the drainage pump is abnormal, there are at least two types of abnormalities. When the drainage pump is determined to be abnormal based on the detection results, the detection results include the target abnormality type of the drainage pump. The change value of the water level signal of the drainage pump is different under different abnormality types, and the target abnormality type is one of the at least two abnormality types.
[0039] Understandably, by setting up a neural network model to synchronously output the specific abnormality type of the drainage pump, the user's process of detecting the specific fault type of the drainage pump can be reduced, thereby improving maintenance efficiency.
[0040] In some embodiments, the controller is configured to:
[0041] Obtain a training dataset, which includes at least two sample water level signals corresponding to each of the at least two anomaly types;
[0042] The training dataset is input into the initial network model for training to obtain the neural network model.
[0043] Understandably, when training a neural network model, setting the training dataset to include data on the changes in water level signals of drainage pumps under different anomaly types can ensure that the neural network model can output the specific anomaly type of the drainage pump, thereby improving the operational reliability of the neural network model.
[0044] In some embodiments, the detection result includes the probability value of the drainage pump malfunctioning during the drainage process in the next detection cycle.
[0045] Understandably, by setting the detection results to include the probability value of the drainage pump malfunction, the controller can be provided with data support to determine whether the drainage pump is malfunctioning, simplifying the judgment process and improving operating efficiency. Attached Figure Description
[0046] To more clearly illustrate the implementation methods in some embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings.
[0047] Figure 1 This is a schematic diagram of the structure of a washing machine according to some embodiments of this application;
[0048] Figure 2 This is a schematic diagram of the electrical control structure of a washing machine according to some embodiments of this application;
[0049] Figure 3 This is a flowchart illustrating the control logic of the controller in some embodiments of this application;
[0050] Figure 4 This is a schematic diagram of the structure of a neural network model in some embodiments of this application;
[0051] Figure 5 This is one of the structural schematic diagrams of a gated loop module according to some embodiments of this application;
[0052] Figure 6 This is one of the schematic diagrams illustrating the implementation structure of the neural network model in some embodiments of this application;
[0053] Figure 7 This is a second schematic diagram of the structure of a gated loop module according to some embodiments of this application;
[0054] Figure 8 This is a second schematic diagram illustrating the implementation structure of a neural network model in some embodiments of this application;
[0055] Figure 9 This is a schematic diagram of the controller structure of some embodiments of this application. Detailed Implementation
[0056] To make the objectives, implementation methods and advantages of this application clearer, the exemplary implementation methods of this application will be clearly and completely described below with reference to the accompanying drawings of the exemplary embodiments of this application. Obviously, the described exemplary embodiments are only some embodiments of this application, and not all embodiments.
[0057] It should be noted that the brief descriptions of terms in this application are only for the convenience of understanding the embodiments described below, and are not intended to limit the embodiments of this application. Unless otherwise stated, these terms should be understood in their ordinary and common meaning.
[0058] Furthermore, the terms “comprising” and “having”, and any variations thereof, are intended to cover but not exclusively include, for example, a product or device that includes a series of components is not necessarily limited to those that are explicitly listed, but may include other components that are not explicitly listed or that are inherent to such product or device.
[0059] In the description of this application, it should be understood that the terms "center", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this application.
[0060] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this application, unless otherwise stated, "a plurality of" means two or more.
[0061] In the description of this application, it should be noted that, unless otherwise expressly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection between two components. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.
[0062] Figure 1 This is a schematic diagram of the structure of a washing machine provided in some embodiments of this application. The washing machine provided in this application can be a drum washing machine or a top-loading washing machine. Figure 1 The washing machine structure shown is only one embodiment and should not be construed as limiting this application.
[0063] In some embodiments, the washing machine may include a housing 110, which is typically made of steel plate, stainless steel, or plastic as the main material, and its exterior is usually coated or plated to increase corrosion resistance and aesthetics. The housing 110 is the external protective structure of the washing machine, providing support and protection for the internal components while also having an attractive appearance. It also helps reduce noise and vibration. The housing 110 protects the internal electrical components, supports other parts, and ensures the stability of the washing machine. The housing 110 is connected to other components, especially to the outer drum 120 disposed within the housing 110, via screws, clips, or other means.
[0064] In some embodiments, the washing machine may include a drum. The drum of the washing machine provided in this application may include an outer drum 120 disposed in the housing 110 and an inner drum 130 disposed in the outer drum 120. The inner drum 130 has a washing chamber for washing clothes. The inner drum 130 rotates under the drive of the motor 140 so that the clothes and detergent are washed in the inner drum 130 by friction, tumbling and other means.
[0065] In some embodiments, the outer cylinder 120 is typically made of stainless steel, plastic, or composite materials. The outer cylinder 120 surrounds and supports the inner cylinder 130, holds water and provides protection, while keeping the inner cylinder 130 isolated from the outside environment. The inner cylinder 130 is typically made of stainless steel or plastic and can be secured inside the outer cylinder 120 by a support shaft.
[0066] In some embodiments, the washing machine may include a controller ( Figure 1 (Not shown in the image), the controller controls the operating status and parameters of the motor 140 to drive the inner cylinder 130 to rotate. The motor 140 can be directly driven to connect to and drive the inner cylinder 130 to rotate, or the motor 140 can drive the inner cylinder 130 to rotate through a transmission structure such as a belt or pulley, which is not limited here.
[0067] In some embodiments, the washing machine may include a detergent dispenser 150, which can be used to hold powdered detergents, such as laundry powder. Optionally, the detergent dispenser 150 can also be used to hold liquid detergents, such as washing liquid, etc., without limitation. Generally, powdered detergents are more difficult to dissolve in washing water than liquid detergents. The washing machine provided in this application can control the motor 140 to drive the inner drum 130 to rotate at a preset speed. The water flow generated between the inner and outer drums prevents the powdered detergent from sinking to the bottom of the outer drum 120, promoting detergent dissolution and improving detergent utilization and washing effect. For liquid detergents, the washing machine provided in this application also has the ability to accelerate their uniform diffusion in washing water.
[0068] In some embodiments, the washing machine may include a water inlet pipe 170, which is connected to a water inlet valve 160 via a washing box 150 for introducing washing water and detergent into the outer drum 120.
[0069] In some embodiments, the washing machine may include a water inlet valve 160, which may be connected to an external water source, such as a tap water pipe or a water storage device, so as to introduce an appropriate amount of washing water as needed during the washing process.
[0070] In some embodiments, the washing machine may also include a main water inlet valve, one end of which is connected to an external water source and the other end of which is connected to the water inlet valve 160. Thus, when both the main water inlet valve and the water inlet valve 160 are in the open state, the washing water and detergent are guided into the outer drum 120 through the washing box 150 and the water inlet pipe 170.
[0071] The washing machine controller 210 provided in this application controls the motor to rotate the inner drum at a preset speed before the washing water and detergent are introduced into the outer drum by opening the water inlet valve. The water flow generated between the inner and outer drums prevents the detergent from sinking to the bottom of the outer drum, promotes detergent dissolution, and improves the utilization rate of detergent and washing effect.
[0072] In some embodiments, the washing machine may include a drain pump 180 disposed at the outlet of the drum to drain water from the inner drum.
[0073] In some embodiments, the washing machine may include a water level sensor 220, which is used to determine the water level based on the pressure value generated by the water level in the inner drum.
[0074] In some embodiments, the water level sensor 220 can sense the water level by air pressure. That is, an air chamber is set at the bottom of the outer drum of the washing machine. As the water level rises, the air in the air chamber is compressed, generating pressure. The air chamber is connected to the water level sensor 220 through a flexible hose. Thus, the water level sensor 220 can output a corresponding water level signal by sensing the air pressure in the air chamber.
[0075] In some embodiments, the water level sensor 220 senses the water level by means of changes in inductance. The water level sensor 220 contains an inductor coil and a movable ferrite core. When the air pressure changes, the position of the magnetic core changes, thereby changing the inductance of the inductor coil.
[0076] In some embodiments, the controller 210 can be connected to an inductor via an oscillation circuit, thereby sensing the frequency of change in the inductance of the inductor through the oscillation circuit. The oscillation circuit includes an inductor and a capacitor, so the controller 210 can output a corresponding water level signal by detecting the frequency of change in the inductance of the capacitor. This can be achieved according to the formula... Since the capacitance C is constant, the larger the inductance L of the inductor coil, the lower the frequency f of the inductance change.
[0077] In some embodiments, the washing machine may include a controller 210, such as Figure 2 As shown, the controller 210 is electrically connected to the water level sensor 220 and the drain pump 180 respectively, and is used to obtain the water level signal output by the water level sensor 220 and control the working state of the drain pump 180. The working state includes a draining state and a non-draining state. When the drain pump 180 is in the draining state, the drain pump 180 will drain the water in the inner cylinder. When the drain pump 180 is in the non-draining state, the drain pump 180 will not drain the water in the inner cylinder.
[0078] The controller 210 is the control center of the washing machine. It connects all parts of the washing machine via various interfaces and lines. By running or executing software programs and / or modules stored in the memory 240, and by calling data stored in the memory 240, it performs various functions and processes data, thereby providing overall monitoring of the washing machine. Optionally, the controller 210 may include one or more processing units; preferably, the controller 210 may integrate an application processor and a modem processor, wherein the application processor mainly handles the operating system, user interface, and applications, and the modem processor mainly handles wireless communication. It is understood that the modem processor may not be integrated into the controller 210.
[0079] The motor 140 is electrically connected to the controller 210, so the controller 210 can control the motor 140 to drive the inner drum 130 to rotate. The controller 210 can control the operating status and operating parameters of the motor 140, and promote the full dissolution of detergent and improve the utilization rate of detergent when the motor 140 drives the inner drum 130 to rotate.
[0080] The water inlet valve 160 is electrically connected to the controller 210, so that the controller 210 can control the washing water to enter the washing chamber of the inner drum 130 through the outer drum 120 by controlling the operating state of the water inlet valve 160.
[0081] The communication device 230 is electrically connected to the controller 210. The communication device 230 is a component used to communicate with external devices or servers according to various communication protocol types. For example, the communication device 230 may include at least one of the following: a wireless communication technology (WiFi) module, a Bluetooth module, a wired Ethernet module, a near-field communication (NFC) module, or other network communication protocol chips or NFC chips, as well as an infrared receiver. The communication device 230 can be used to communicate with other devices or communication networks (such as Ethernet, radio access network (RAN), wireless local area network (WLAN), etc.). The controller 210 can communicate with the user's terminal device (such as a mobile phone) through the communication device to obtain the user's adjustment commands for the washing machine, thereby realizing remote control of the washing machine.
[0082] The memory 240 is electrically connected to the controller 210. The memory 240 can be used to store software programs and modules. The controller 210 executes various functional applications and data processing of the electronic device by running the software programs and modules stored in the memory 240. The memory 240 may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function, etc.; the data storage area may store data created based on the use of the washing machine (e.g., operating parameters of the motor 140 and the opening time of the water inlet valve 160), etc. In addition, the memory 240 may include high-speed random access memory, and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other volatile solid-state storage device.
[0083] In some possible embodiments, the system architecture of the washing machine may also include a timer, which is electrically connected to the memory 240. The timer can detect the working duration of each electrical component, such as the opening duration of the water inlet valve 160, and store the acquired duration information in the memory 240 so that the controller 210 can perform water level control. The timer can also be used to realize functions such as timed opening or closing of the washing machine, which is not limited here.
[0084] Normally, the proper functioning of a washing machine's drainage process depends on the functionality of the drain pump 180. Only when the drain pump 180 is working correctly can the washing machine drain properly. If the drain pump 180 malfunctions, the drainage process will be obstructed. However, in related technologies, the user can only perceive the abnormality of the drain pump 180 when it completely fails, meaning it cannot drain at all. At this point, repairs will inevitably affect the user experience.
[0085] Based on the above problems, this application provides a washing machine that inputs at least two water level signals in at least two detection cycles into a neural network model, thereby outputting a detection result indicating whether the drain pump will experience any abnormalities during the drainage process in the next detection cycle. This allows for early detection of any abnormalities in the drain pump, improving maintenance efficiency.
[0086] In some embodiments, such as Figure 3 As shown, the controller in the washing machine can perform the following steps:
[0087] Step S101: Obtain at least two water level signals in at least two detection cycles using a water level sensor;
[0088] Step S102: Input at least two water level signals into the neural network model, and output the detection result of whether the drainage pump is abnormal during the drainage process in the next detection cycle through the neural network model;
[0089] The neural network model was obtained by training the initial network model using a training dataset, which included water level signals when the drainage pump experienced drainage anomalies.
[0090] In some embodiments, the above-mentioned at least two detection cycles can be continuous detection cycles or non-continuous detection cycles, which can be set by those skilled in the art according to the actual situation, and the embodiments of this application do not impose any restrictions.
[0091] It should be noted that regardless of whether the at least two detection cycles are consecutive or non-consecutive, the cycle information must be identified, or the data must be sorted according to the execution order of the cycles before being input into the neural network model, so that the neural network model can detect the changing trend of the change value between at least two water level signals in at least two detection cycles.
[0092] It should be understood that the above test results indicate whether the drainage pump has any abnormalities during the drainage process in the next test cycle. That is, the test results when there is an abnormality are different from the test results when there is a normality. The controller can determine whether the drainage pump has any abnormalities during the drainage process in the next test cycle based on the content of the test results.
[0093] In some embodiments, the content of the above detection results can be reflected by probability values. That is, the above detection results may include the probability value of the drainage pump becoming abnormal during the drainage process in the next detection cycle. For example, the neural network model obtains a probability value of 0.6 for the drainage pump becoming abnormal during the drainage process in the next detection cycle based on at least two water level signals. After obtaining this probability value, the controller can determine whether the drainage pump will become abnormal during the drainage process in the next detection cycle based on the judgment of the magnitude of the probability value.
[0094] In other embodiments, the detection result may include the probability value that the drainage pump is normal during the drainage process in the next detection cycle. For example, the neural network model obtains a probability value of 0.6 that the drainage pump is normal during the drainage process in the next detection cycle based on at least two water level signals. After obtaining this probability value, the controller can determine whether the drainage pump will be abnormal during the drainage process in the next detection cycle based on the magnitude of the probability value.
[0095] In other embodiments, the content of the above detection results can be reflected by "yes" or "no" information. The detection results may include text information indicating that the drainage pump will malfunction or that the drainage pump will not malfunction. For example, the neural network model can directly output a "yes" or "no" result classification on whether the drainage pump will malfunction during the drainage process in the next detection cycle based on at least two water level signals. Then, the controller determines whether the drainage pump will malfunction during the drainage process in the next detection cycle based on the "yes" or "no" result classification.
[0096] In some embodiments, the above-mentioned at least two water level signals may be input into the neural network model simultaneously, or they may be input into the neural network model sequentially according to the periodic execution order. The specific settings can be made by those skilled in the art according to the actual situation, and the embodiments of this application do not impose any restrictions.
[0097] When a drainage pump malfunctions, it will manifest in its drainage volume. Different types of malfunctions will show different trends in water level signal changes. Therefore, by using a neural network model to detect at least two water level signals within at least two detection cycles, the specific malfunction type of the drainage pump can be determined based on the trend between the at least two water level signals.
[0098] In some embodiments, when the drainage pump is abnormal, there are at least two types of abnormality. When the detection result indicates that the drainage pump is abnormal, the detection result includes the target abnormality type of the drainage pump. The change value of the water level signal of the drainage pump is different under different abnormality types, and the target abnormality type is one of the at least two abnormality types.
[0099] For example, taking at least two anomaly types, including complete failure, congestion, and partial blockage, when the drainage pump is in the complete failure anomaly type, the drainage pump cannot drain water at all, so the amplitude of the change value between at least two water level signals shows a trend of suddenly decreasing from the normal value to zero; when the drainage pump is in the congestion anomaly type, the drainage volume of the drainage pump gradually decreases over time, so the amplitude of the change value between at least two water level signals shows a trend of gradually decreasing over time; when the drainage pump is in the partial blockage anomaly type, the drainage volume of the drainage pump will first decrease for a period of time, then return to normal, then decrease for a period of time again, and then return to normal again, so the amplitude of the change value between at least two water level signals also shows a trend of decreasing for a period of time, then returning to normal, then decreasing for a period of time again, and then returning to normal again.
[0100] Understandably, by setting up a neural network model to synchronously output the specific abnormality type of the drainage pump, the user's process of detecting the specific fault type of the drainage pump can be reduced, thereby improving maintenance efficiency.
[0101] In some embodiments, when the detection result includes probability values, the neural network model can output the probability values corresponding to all anomaly types, or it can only output the probability values of all probability values that are greater than a preset threshold. The specific settings can be made by those skilled in the art according to the actual situation, and this application does not impose any restrictions.
[0102] In order for the neural network model to output the corresponding target anomaly type when outputting the detection result, those skilled in the art need to set the training dataset to include the change values of sample water level signals under multiple different anomaly types when training the initial network model, so that the neural network model can perform the above step S102 according to the feature values between at least two water level signals under each anomaly type in the training dataset.
[0103] In some embodiments, the training dataset includes at least two sample water level signals corresponding to each anomaly type, and the controller is configured to:
[0104] The training dataset is input into the initial network model for training to obtain the neural network model.
[0105] In some embodiments, the above-mentioned at least two sample water level signals may be obtained from historical water level signals collected by the washing machine, or from a server connected to the washing machine. The server may store a sample dataset, which may be preset by those skilled in the art or obtained by collecting data from all washing machines associated with the server. The specific settings are determined by those skilled in the art according to the actual situation, and this application embodiment does not impose any restrictions.
[0106] Understandably, when training a neural network model, setting the training dataset to include data on the changes in water level signals of drainage pumps under different anomaly types can ensure that the neural network model can output the specific anomaly type of the drainage pump, thereby improving the operational reliability of the neural network model.
[0107] To enable users to detect washing machine malfunctions, the controller can output an alarm message when the detection result from the neural network model indicates a problem with the drain pump. Specifically, in some embodiments, the controller is configured to:
[0108] If the test results indicate that the drainage pump is malfunctioning during the next testing cycle, an alarm message will be output to prompt the user to check the drainage pump.
[0109] In some embodiments, the alarm information may be an audio-visual signal, a communication signal sent to an associated electronic device corresponding to the washing machine, or text and image information displayed on the washing machine's screen. The specific settings can be made by those skilled in the art according to the actual situation, and this application embodiment does not impose any restrictions.
[0110] Since the neural network model outputs the predicted value of the next detection cycle based on at least two water level signals and classifies the predicted value of the next detection cycle, commonly used classifiers in this field can be applied to the application environment of this application, such as convolutional neural networks (CNN).
[0111] In some embodiments, since this application analyzes and predicts the change values between at least two water level signals, which involves the calculation of water level signals corresponding to two adjacent detection cycles, the above neural network model can adopt the structure of recurrent neural networks (RNNs) commonly used in the art, such as LSTM (Long Short-Term Memory) or gated recurrent unit (GRU).
[0112] That is, in some embodiments, such as Figure 4 As shown, the neural network model includes an interconnected gated loop module 301 and a result output module 302, wherein:
[0113] The gated loop module 301 is used to receive the water level signal of the current detection cycle, obtain the feature value of the water level signal of the current detection cycle based on the water level signal of the current detection cycle and the water level signal of the previous detection cycle, and output the feature value of the water level signal of the current detection cycle.
[0114] The result output module 302 is used to receive and obtain the detection result based on the characteristic value of the water level signal in the current detection cycle.
[0115] It should be understood that the neural network model with gated loop module 301 can better capture the dependency between at least two water level signals, and it can acquire the signal transmission process that controls the water level signal of the current detection cycle and the water level signal of the previous detection cycle through training.
[0116] The structure of the gated recurrent module 301 varies depending on the network model. For example, the LSTM model includes a memory unit 3011, an input gate 3012, a forget gate 3013, and an output gate 3014, totaling three gates. The GRU module includes an update gate 3015 and a reset gate 3016, totaling two gates. Those skilled in the art can select the specific structure of the gated recurrent module 301 according to the structural parameter requirements. This application does not impose any restrictions on this embodiment.
[0117] The structure of the gated loop module 301 for the LSTM model and the GRU module is described below from the perspective of different embodiments.
[0118] For the LSTM model, the gated recurrent module 301 of the LSTM model includes a memory unit 3011, an input gate 3012, a forget gate 3013, and an output gate 3014, as follows: Figure 5 As shown, memory unit 3011 is a data channel that runs through the entire sequence, used to store and transmit key information. Memory unit 3011 maintains the persistence of information between detection cycles of the LSTM model, and can be regarded as a long-term memory for key information. This design enables the LSTM model to effectively remember important information in long sequences, avoiding the problem of traditional RNNs forgetting early information in long sequences. The state of memory unit 3011 is gradually updated with each detection cycle of the sequence. New information can be written, and old information can be cleared through "forgetting". This process is controlled by a gating mechanism including input gate 3012, forget gate 3013, and output gate 3014, thereby ensuring that memory unit 3011 only stores key information when it is updated. This state transmission and update gives the LSTM model the ability to capture long-term dependencies.
[0119] The LSTM model mainly relies on the input gate 3012, the forget gate 3013, and the output gate 3014 to control the flow and update of key information in the memory unit 3011. Among them, the input gate 3012 is responsible for controlling the degree of influence of the current input information on the memory unit 3011. It determines which information in the current input value and the feature value (hidden state) of the previous detection cycle needs to be added to the memory unit 3011. It maps the result value to between 0 and 1 through the activation function, thereby determining the proportion of information entering the memory unit 3011.
[0120] The forget gate 3013 determines which information is removed from memory unit 3011. In each detection cycle, the LSTM model reads the current input and the hidden state from the previous cycle, outputting a value between 0 and 1 through an activation function. This value corresponds to the proportion of information forgotten; a value closer to 1 indicates that the information is more important and should be retained, while a value closer to 0 indicates that the information can be forgotten. The forget gate 3013 mechanism allows the LSTM model to flexibly and selectively discard irrelevant historical information, thereby preventing the state of memory unit 3011 from accumulating due to irrelevant information.
[0121] Output gate 3014 determines how the state of memory cell 3011 in the current detection cycle affects the hidden state of the output. That is, it controls the output of the content of memory cell 3011 in the current detection cycle. The result of output gate 3014 is passed through an activation function, which limits its value to between 0 and 1. This output value determines the content of the current hidden state.
[0122] In the process of outputting detection results through the LSTM model, such as Figure 6 As shown, firstly, based on the current input and the hidden state of the previous detection cycle, the output (f) of the forget gate 3013 is calculated. t This determines which information in the current memory unit 3011 needs to be forgotten. Then, the value of the input gate 3012 (i) is calculated. t ) and the candidate hidden state of the current detection period The system determines which of the currently input water level signals should be added to memory unit 3011; then it adds the results of forget gate 3013 and input gate 3012 to update the state of memory unit 3011 (C). t This step ensures that memory unit 3011 retains useful information while also introducing new information; finally, based on the current input water level signal and the state of memory unit 3011 (C... t The value of output gate 3014 (o) is calculated. t ), and the new hidden state (h t This hidden state (h) t It will be passed as output to the next detection cycle, and at the same time serve as the output value of the current detection cycle.
[0123] Specifically, the forgetting gate 3013 controls which information in memory unit 3011 needs to be retained or forgotten according to the following calculation formula:
[0124] f t =σ(W f [h t-1 ,x t ]+b f );
[0125] Among them, f t This is the output value of Forgotten Gate 3013, W. f This is the weight matrix of the forgetting gate 3013, h t-1 It is the hidden state from the previous detection period, x t This is the water level signal for the current detection cycle, b f It is the bias vector of the forget gate 3013, σ is the activation function, and f t Limited to between 0 and 1.
[0126] The input gate 3012 determines the extent to which the water level signal of the current detection cycle is written into the memory unit 3011, and its calculation process includes two parts:
[0127] The first part involves using an activation function to generate the write ratio:
[0128] i t =σ(Q) i [h t-1 ,x t ]+b u );
[0129] Among them, i t It is the output value of input gate 3012, W i It is the weight matrix of the input gate 3012, h t-1 It is the hidden state from the previous detection period, x t This is the water level signal for the current detection cycle, b i It is the bias vector of the input gate 3012, σ is the sigmoid activation function, and it will set i t Limited to between 0 and 1.
[0130] The second part is to generate candidate hidden states for the current detection period, which is to activate them using the activation function tanh:
[0131]
[0132] in, W is the candidate hidden state for the current detection period. i It is the weight matrix of the input gate 3012, h t-1It is the hidden state from the previous detection period, x t This is the water level signal for the current detection cycle, b c is the bias vector of the input gate 3012, and tanh is the activation function.
[0133] Next, the output value of input gate 3012 is multiplied by the candidate hidden state of the current detection period to obtain the updated input value.
[0134] After combining the output value of forget gate 3013 with the output value of input gate 3012, update the state value C of memory unit 3011. t :
[0135]
[0136] Finally, output gate 3014 determines the output information for the current detection cycle, and its calculation formula is as follows:
[0137] o t =σ(W o [h t-1 ,x t ]+b o );
[0138] Among them, o t It is the output value of output gate 3014, W o It is the weight matrix of output gate 3014, h t-1 It is the hidden state from the previous detection period, x t This is the water level signal for the current detection cycle, b o It is the bias vector of the output gate 3014, and σ is the sigmoid activation function. t Limited to between 0 and 1.
[0139] Then, the state of memory cell 3011 is processed by the activation function and multiplied by the output value of output gate 3014 to obtain the final hidden state output value h. t :
[0140] h t =o t tanh(C t );
[0141] Furthermore, the LSTM model outputs h through the result output module 302. t After processing, the test results are obtained.
[0142] However, when using the LSTM model to process at least two water level signals, if the LSTM model makes a calculation error in the first few predictions of the water level signal in the next detection period, then in order to correct this calculation error, the gradient will propagate to the earlier detection periods of the model to update the weights associated with this error. However, because gradients can vanish during the calculation of predictions in subsequent detection periods, these gradients may gradually decrease in the network, resulting in the weights of earlier detection periods being hardly updated, thus failing to effectively learn long-term dependencies.
[0143] Therefore, in order to improve the model's prediction accuracy for earlier detection cycles, a gated recurrent unit (GRU) can be used to analyze and predict at least two water level signals.
[0144] For the GRU model, the gated loop module of the GRU model includes an update gate 3015, a reset gate 3016, and a computation unit 3017, such as... Figure 7 As shown, the reset gate 3016 is used to control the combination ratio of the water level signal of the previous detection cycle and the water level signal of the current detection cycle, to obtain and output a first feature value; the update gate 3015 is used to control the influence of the water level signal of the previous detection cycle on the feature value of the water level signal of the current detection cycle, to obtain and output a second feature value; and then the calculation unit 3017 is used to receive the first feature value and the second feature value, and to obtain and output the feature value of the water level signal of the current detection cycle based on the first feature value, the second feature value, the water level signal of the previous detection cycle and the water level signal of the current detection cycle.
[0145] The function of update gate 3015 is to determine how much information from the previous time step should be retained in the hidden state of the current time step. The output value of update gate 3015 is between 0 and 1. If the output value of update gate 3015 is close to 0, it means that the feature value (hidden state) of the water level signal of the previous detection period is ignored and more dependent on the input water level signal of the current detection period. If the output value of update gate 3015 is closer to 1, it means that more of the past hidden state is retained.
[0146] Among them, the reset gate 3016 determines the extent to which the hidden state of the water level signal in the previous detection cycle is ignored. When the output of the reset gate 3016 is close to 0, the GRU model tends to "forget" the feature value of the water level signal in the previous detection cycle and only depends on the water level signal of the current detection cycle input. When the output is close to 1, the feature value of the water level signal in the previous detection cycle will be retained more.
[0147] In the process of outputting detection results through the GRU model, such as Figure 8As shown, the corresponding output values need to be calculated first through update gate 3015 and reset gate 3016 respectively. Then, the degree of dependence on the hidden state of the water level signal of the previous detection cycle is calculated synchronously through calculation unit 3017, that is, the candidate hidden state of the current detection cycle is calculated. Finally, the hidden state of the current detection cycle is calculated using the output value of update gate 3015 and the candidate hidden state.
[0148] The reset gate 3016 is configured to: perform a linear transformation on a first matrix according to a preset first weight matrix to obtain a target value; the first matrix includes the feature values of the water level signal in the previous detection period and the values of the water level signal in the current detection period; and obtain the first feature value according to the activation function and the target value, i.e., the corresponding calculation formula is:
[0149] r t =σ(W r [h t-1 ,x t ]+b r );
[0150] Where, r t This resets the output value of door 3016, W. r It is the weight matrix (first weight matrix) of the reset gate 3016, h t-1 It is the hidden state from the previous detection period, x t This is the water level signal for the current detection cycle, [h t-1 ,x t ] is the first matrix, b r It is the bias vector of the reset gate 3016, σ is the activation function, and r t The value is limited to between 0 and 1, for example, 0.6.
[0151] The update gate 3015 is configured to: perform a linear transformation on the second matrix according to a preset second weight matrix to obtain the target value; the second matrix includes the feature values of the water level signal in the previous detection period and the values of the water level signal in the current detection period; and obtain the second feature value according to the activation function and the target value, i.e., the corresponding calculation formula is:
[0152] z t =σ(W z [h t-1 ,x t ]+b z );
[0153] Among them, z t It updates the output value of gate 3015, W. z It updates the weight matrix (second weight matrix) of gate 3015, h t-1 It is the hidden state from the previous detection period, x tThis is the water level signal for the current detection cycle, [h t-1 ,x t ] is the second matrix, b z It updates the bias vector of gate 3015, where σ is the activation function, and z... t The value is limited to between 0 and 1, for example, 0.6.
[0154] The calculation unit 3017 is configured to: obtain candidate features corresponding to the current detection period based on the first feature value, the water level signal of the previous detection period and the water level signal of the current detection period; and obtain feature values of the water level signal of the current detection period based on the candidate features, the second feature value and the water level signal of the previous detection period.
[0155] The calculation unit 3017 mainly involves two parts of the calculation process:
[0156] In the first part, the calculation unit 3017 obtains the first characteristic value r output by the reset gate 3016. t Then, based on the first eigenvalue r t The candidate features corresponding to the current detection period are obtained by combining the water level signals from the previous detection period and the current detection period.
[0157] For example, the calculation unit 3017 then calculates the output value r of the reset gate 3016. t The formula used to calculate the candidate feature values (candidate hidden states) for the current detection period is as follows:
[0158]
[0159] As can be seen from the above calculation formula, r t The smaller the value, the more information needs to be forgotten in the previous detection cycle, and the more data is discarded; r t The larger r is, the more the water level signal of the current detection period is combined with the hidden state of the previous detection period; when r t When r approaches 0, it indicates that the hidden state from the previous detection period needs to be discarded entirely, retaining only the water level signal for the current detection period. Therefore, it can be used to discard historical information irrelevant to prediction; when r... t When the value is close to 1, it indicates that the hidden state from the previous detection cycle needs to be retained.
[0160] In the second part, the computation unit 3017 obtains the second eigenvalue z output by the update gate 3015. t Then, based on candidate features Second eigenvalue z t The water level signal h from the previous detection cycle t-1 Obtain the characteristic value h of the water level signal in the current detection period. t The corresponding calculation formula is:
[0161]
[0162] As can be seen from the above calculation formula, z t The closer to 1, the more historical information is remembered; the closer to 0, the more is forgotten. Among these, (1-z t )*h t-1 This represents the forgotten information portion related to the hidden state from the previous detection cycle; This represents the selective memorization of candidate hidden states for the current detection period. It is the final output of the feature value h for the current detection period. t It will forget some information about the hidden state from the previous detection cycle and add some features of the water level signal from the current detection cycle. The feature value h of the current detection cycle... t It will then be transmitted to the water level signal in the next detection cycle to support obtaining the detection results in the next detection cycle after that.
[0163] In some embodiments, the result output module 302 includes at least one fully connected layer, which is used to receive and output the detection result based on the characteristic value of the water level signal in the current detection period.
[0164] It should be understood that a fully connected layer can transform the feature values of the current detection cycle into the final detection result output through linear transformation and nonlinear activation functions. Specifically, the fully connected layer will implement different linear transformations and nonlinear activation functions for different detection results. For example, when the detection result includes probability values, the fully connected layer implements the first linear transformation and the first nonlinear activation function. When the detection result includes the classification of "yes" or "no", the fully connected layer implements the second linear transformation and the second nonlinear activation function.
[0165] The at least one fully connected layer can be connected in series or in parallel, and the specific configuration can be determined by those skilled in the art based on the actual situation. This application does not impose any restrictions on this configuration.
[0166] Understandably, by using at least one fully connected layer to classify the feature values of the water level signal in the current detection period, the detection results can be obtained, thus improving the reliability of the neural network model.
[0167] Figure 9 This is a schematic diagram of a controller provided in an embodiment of this application. Figure 9 As shown, the controller may include a processor, memory, bus, and device interface.
[0168] The processor calls the executable program code stored in the memory to execute any of the air conditioner control methods disclosed in the embodiments of this application.
[0169] The memory stores executable program code, which is executed by the processor to implement any of the air conditioner control methods disclosed in the embodiments of this application.
[0170] The bus is used to transfer program code stored in memory to the processor for execution.
[0171] The device interface connects to the bus and is used to enable the processor and memory to connect to other devices.
[0172] Optionally, the memory may include read-only memory and random access memory, and provide instructions and data to the processor. A portion of the memory may also include non-volatile random access memory. For example, the memory may also store device type information. The processor can be used to execute instructions stored in the memory, and when the processor executes the instructions, the processor can perform the various steps and / or processes corresponding to the terminal device in the above method embodiments.
[0173] Optionally, the memory can be located outside the controller and communicate with the controller.
[0174] It should be understood that, in the embodiments of this application, the processor may be a Central Processing Unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc.
[0175] It should be noted that, Figure 9 The controller shown may also include components not shown, such as a power supply, which will not be described in detail in this embodiment.
[0176] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.
[0177] For ease of explanation, the above description has been provided in conjunction with specific embodiments. However, the above exemplary discussion is not intended to be exhaustive or to limit the embodiments to the specific forms disclosed above. Various modifications and variations can be obtained based on the above teachings. The selection and description of the above embodiments are for the purpose of better explaining the principles and practical applications, thereby enabling those skilled in the art to better utilize the described embodiments and various different variations of embodiments suitable for specific use considerations.
Claims
1. A washing machine, characterized in that, The washing machine includes: The drum body includes an outer drum and an inner drum disposed inside the outer drum, wherein a washing chamber for washing clothes is formed inside the inner drum; A drainage pump, connected to the outlet of the cylinder, is used to drain the water from the inner cylinder; A water level sensor is used to detect the water level in the inner cylinder and output a water level signal; A controller, electrically connected to the water level sensor and the drainage pump, is used to control the drainage pump to discharge water from the inner cylinder; The controller is configured as follows: The water level sensor obtains at least two water level signals in at least two detection cycles. The at least two water level signals are input into a neural network model, and the neural network model outputs a detection result on whether the drainage process of the drainage pump is abnormal in the next detection cycle. If the test results indicate that the drainage pump is malfunctioning during the drainage process in the next test cycle, an alarm message will be output to prompt the user to check the drainage pump. The neural network model is obtained by training an initial network model using a training dataset, which includes water level signals when the drainage pump malfunctions.
2. The washing machine as described in claim 1, characterized in that, The test results include the probability value of the drainage pump malfunctioning during the drainage process in the next test cycle.
3. The washing machine as described in claim 1 or 2, characterized in that, The neural network model includes interconnected gated loop modules and result output modules, wherein: The gated loop module is used to receive the water level signal of the current detection cycle, obtain the feature value of the water level signal of the current detection cycle based on the water level signal of the current detection cycle and the water level signal of the previous detection cycle, and output the feature value of the water level signal of the current detection cycle. The result output module is used to receive and obtain the detection result based on the characteristic value of the water level signal in the current detection period.
4. The washing machine as described in claim 3, characterized in that, The characteristic values of the water level signal in the current detection cycle include a first characteristic value and a second characteristic value. The gated loop module includes an update gate, a reset gate, and a calculation unit. The update gate and the reset gate are respectively connected to the calculation unit, wherein: The reset gate is used to control the combination ratio of the water level signal of the previous detection cycle and the water level signal of the current detection cycle, so as to obtain and output the first feature value. The update gate is used to control the degree of influence of the water level signal of the previous detection period on the feature value of the water level signal of the current detection period, and to obtain and output the second feature value. The calculation unit is configured to receive the first feature value and the second feature value, and obtain and output the feature value of the water level signal of the current detection period based on the first feature value, the second feature value, the water level signal of the previous detection period and the water level signal of the current detection period.
5. The washing machine as described in claim 4, characterized in that, The computing unit is configured as follows: Based on the first feature value, the water level signal of the previous detection period, and the water level signal of the current detection period, the candidate feature corresponding to the current detection period is obtained; The feature value of the water level signal in the current detection period is obtained based on the candidate feature, the second feature value, and the water level signal of the previous detection period.
6. The washing machine as described in claim 4, characterized in that, The reset door is configured as follows: The first matrix is linearly transformed according to the preset first weight matrix to obtain the target value. The first matrix includes the feature value of the water level signal in the previous detection period and the value of the water level signal in the current detection period. The first feature value is obtained based on the activation function and the target value.
7. The washing machine as described in claim 4, characterized in that, The update gate is configured as follows: The target value is obtained by linearly transforming the second matrix according to the preset second weight matrix. The second matrix includes the feature value of the water level signal in the previous detection period and the value of the water level signal in the current detection period. The second feature value is obtained based on the activation function and the target value.
8. The washing machine as described in claim 3, characterized in that, The result output module includes at least one fully connected layer, which is used to receive and output the detection result based on the characteristic value of the water level signal in the current detection period.
9. The washing machine as described in claim 1 or 2, characterized in that, When the drainage pump is abnormal, there are at least two types of abnormality. When the drainage pump is determined to be abnormal based on the detection results, the detection results include the target abnormality type of the drainage pump. The change value of the water level signal of the drainage pump is different under different abnormality types, and the target abnormality type is one of the at least two abnormality types.
10. The washing machine as described in claim 9, characterized in that, The controller is configured to: Obtain a training dataset, which includes at least two sample water level signals corresponding to each of the at least two anomaly types; The training dataset is input into the initial network model for training to obtain the neural network model.