Washing machine eccentricity model training method and washing machine eccentricity detection method

Abstract

The invention provides a washing machine eccentricity model training method and a washing machine eccentricity detection method, and the method comprises the steps: presetting a plurality of time windows in a dewatering time sequence, and analyzing a part of curves, corresponding to each time window, on an eccentricity curve to obtain features corresponding to each time window; respectively training eccentric models corresponding to the time windows according to the characteristics, and substituting part of the characteristics in the previous time window together when training the next eccentric model, so as to obtain a plurality of eccentric models which have mutual relevance and are sequentially arranged along the dewatering time sequence; the method comprises the following steps of: drawing an eccentricity curve taking eccentricity and dewatering time as coordinate axes in real time in the dewatering process of the washing machine, analyzing the eccentricity curve in each time window to obtain corresponding characteristics when the dewatering time passes through each time window, and substituting the characteristics into a corresponding eccentricity model to judge whether eccentricity exists or not, the actual eccentric state of the washing machine can be obtained more accurately.

Description

Technical Field

[0001] The invention belongs to the technical field of washing equipment, and in particular relates to a washing machine eccentricity model training method and a washing machine eccentricity detection method. Background Art

[0002] With the improvement of people's living standards, the popularity of various types of intelligent washing equipment is becoming higher and higher, and they are widely praised by users because they can complete most routine cleaning operations on behalf of users.

[0003] Take the washing machine as an example. Because it can complete the washing of clothes on the users' behalf, it is widely recognized by users and has become one of the "must-have" household appliances. Usually, the washing machine also has a dehydration function, which greatly reduces the drying time of clothes. However, since most washing machines currently use centrifugal force to separate clothes from water, the washing machine will inevitably vibrate during the dehydration process, which may cause noise due to the vibration. There is also a risk of damage to the washing machine due to collision.

[0004] To this end, the Chinese invention patent with application number CN202310719576.2 discloses a method for detecting eccentricity of clothes in a drum washing machine, which obtains real-time eccentricity load value and total load value; sets balanced average load value according to the size of the drum; inputs the eccentricity load value, total load value and balanced average load value into a clothing eccentricity prediction model to calculate the eccentricity mass; wherein, the training method of the clothing eccentricity prediction model includes: changing the balanced average load mass and eccentricity mass of the drum through a simulated drum experiment to obtain corresponding multiple groups of eccentricity load values ​​and total load values; using the balanced average load mass, eccentricity load value and total load value as the input layer of the BP neural network, and the eccentricity mass as the output layer of the BP neural network, and training the input and output data; although the above scheme can predict when eccentricity has not yet occurred, the accuracy of the prediction result is not high, and cannot fully meet the needs of eccentricity detection.

[0005] In view of this, the present invention is proposed. Summary of the invention

[0006] The purpose of the present invention is to provide a washing machine eccentricity model training method to address the problems in the above-mentioned prior art. By analyzing a section of eccentricity curve in different time windows respectively and associating each time window to obtain an eccentricity model corresponding to the different time windows, the eccentricity judgment requirements in different dehydration stages can be met.

[0007] Another object of the present invention is to provide a method for detecting eccentricity of a washing machine. By adopting the eccentricity model trained by the above-mentioned scheme, different eccentricity models can be used in turn for judgment as the dehydration process proceeds, thereby achieving accurate judgment of eccentricity during the dehydration process and improving judgment accuracy.

[0008] In order to achieve the above-mentioned purpose, the first aspect of the present invention provides a method for training an eccentricity model of a washing machine, wherein several time windows are preset along the dehydration time sequence, the eccentricity curve corresponding to each time window is analyzed, and each time window is associated and trained to obtain an eccentricity model corresponding to each dehydration time length.

[0009] The present invention can judge the eccentricity of the dehydration process in sections by training several eccentricity models arranged along the dehydration time sequence. Since each eccentricity model is trained by the eccentricity curve in the corresponding time window, the judgment accuracy of each eccentricity model in the corresponding time window is higher.

[0010] Furthermore, multiple time windows arranged along the dehydration time sequence are analyzed in turn to obtain several features corresponding to each time window, and at least part of the features obtained from the analysis of the previous time window are brought into the subsequent time window as features for training the eccentricity model corresponding to the subsequent time window.

[0011] In the above scheme, multiple time windows set along the dehydration time sequence are analyzed in sequence, and then when the time window after rain is analyzed and the model is established, some features of the previous time window are brought in, thereby realizing the association between the various eccentric models, thereby further improving the judgment accuracy of each eccentric model.

[0012] Furthermore, the statistical domain, spectral domain and time domain features of the eccentricity curve in each time window are extracted, the statistical domain, spectral domain and time domain features are analyzed to obtain the eccentricity probability in the corresponding time window, and the eccentricity model is obtained by training the statistical domain, spectral domain, time domain features and eccentricity probability of the eccentricity curve in each time window;

[0013] At least the eccentricity probability in the previous time window is brought into the subsequent time window as a feature for training the eccentricity model corresponding to the subsequent time window.

[0014] Furthermore, the statistical domain features include: mean of eccentricity values, standard deviation of eccentricity values, skewness and kurtosis of eccentricity curves;

[0015] The spectrum domain features include: the spectrum distance of the eccentricity curve and the spectrum fundamental frequency of the eccentricity curve;

[0016] The time domain features include: the distance between the crest and the trough in the eccentricity curve, and the number of crests.

[0017] In the above scheme, by extracting and analyzing the morphological features of the eccentricity curve, a more accurate eccentricity model can be trained. In the actual detection process, by analyzing the eccentricity curve formed in real time during the dehydration process, the eccentricity state can be accurately known, and the eccentricity can be discovered and processed in time.

[0018] Furthermore, it is defined that the number of time windows preset along the dehydration time sequence is n, and the number of features extracted from each time window is m, where m>n.

[0019] In the above scheme, the number of features extracted in each time window is greater than the number of time windows, which ensures that there are more features as references when training the model, thereby improving the accuracy and effectiveness of the model.

[0020] Furthermore, the number of features used to train the eccentricity model in the first i-th time window is k i , k i Satisfies the following relationship:

[0021] k i =M i +c i

[0022] Among them, M i represents the number of features extracted from the i-th time window; c i Represents the number of features introduced from the features obtained in the previous time window.

[0023] Furthermore, the features obtained in each time window are sorted according to their importance, and the features with the highest importance ranking are selected from the features obtained in the previous time window. i Features are introduced into the i-th time window;

[0024] Preferably,

[0025] Where n is the number of preset time windows.

[0026] In the above scheme, each feature is sorted according to its importance, and the features ranked at the front in the previous time window are introduced into the subsequent time window as the features for training the eccentricity model in the subsequent time window, so that the eccentricity models are not simply judged on a certain section of the eccentricity curve, but are judged comprehensively based on the features of the eccentricity curve before the eccentricity curve corresponding to the current eccentricity model. The association between the eccentricity models further improves the judgment accuracy of each eccentricity model.

[0027] It should also be noted that the features introduced into the later time window may come from multiple previous time windows. Since the correlation between adjacent eccentric models arranged in order of dehydration time is stronger, the number of features introduced from several previous time windows is controlled according to the change in the distance between the time windows, so that the closer the previous time window is to the later time window, the more features are introduced.

[0028] Furthermore, a plurality of time windows preset along the dehydration time sequence partially overlap or do not overlap each other.

[0029] Furthermore, in the process of training the eccentric model, the missed judgment rate of several eccentric models set along the dehydration time sequence is adjusted to gradually decrease, and the misjudgment rate is gradually increased.

[0030] In the above scheme, by controlling the missed judgment rate and false judgment rate of each eccentricity model, the judgment of eccentricity in the early stage of dehydration is relatively loose, and as the dehydration proceeds, the judgment of eccentricity by the subsequent eccentricity model is gradually tightened, taking into full consideration that a small eccentricity in the early stage may be automatically eliminated as the water content decreases, thereby avoiding the dehydration being stopped due to a small eccentricity in the early stage of dehydration, thereby improving the user experience; as the dehydration proceeds, the water content is lower and the rotation speed is higher in the later stage of dehydration. At this time, by tightening the judgment standard for eccentricity, the occurrence of eccentricity can be more accurately reflected, and a prompt can be issued to the user when the eccentricity may affect the normal progress of dehydration, or the dehydration can be stopped directly, thereby improving the intelligence level of the washing machine.

[0031] A second aspect of the present invention further provides a method for detecting eccentricity of a washing machine, which uses an eccentricity model obtained by the eccentricity model training method described above.

[0032] Specifically, after the dehydration program starts, a dehydration curve is drawn with the eccentricity and the dehydration time as coordinate axes;

[0033] When the dehydration time passes through the time window corresponding to any eccentricity model, the corresponding eccentricity model is used to judge whether eccentricity occurs. If it is judged that eccentricity does not occur, dehydration continues.

[0034] The above scheme draws the dehydration curve in real time during the dehydration process, and uses various eccentricity models to make judgments at different stages of dehydration, thereby achieving accurate detection of dehydration eccentricity.

[0035] The beneficial effects of the present invention are:

[0036] 1. Several time windows preset along the dehydration time sequence are related to each other, and each eccentricity model is not completely independent. Since the eccentricity in the dehydration process is a gradual cumulative change process, the eccentricity model arranged later in the dehydration time sequence is related to the eccentricity model arranged earlier, which improves the judgment accuracy, avoids the independence of each eccentricity model, resulting in one-sided judgment results and inability to accurately reflect the eccentricity state, and improves the judgment accuracy of each eccentricity model.

[0037] 2. Extracting statistical domain, spectral domain and time domain features from the eccentricity curve can more fully analyze the shape of the eccentricity curve. The eccentricity model trained using the above features can more accurately judge the eccentricity.

[0038] 3. Sort the features in each time window according to their importance, and bring some of the top-ranked features in the previous time window into the training process of the subsequent eccentricity model, further improving the correlation between the eccentricity models; by controlling the missed judgment rate of several eccentricity models set along the dehydration time sequence to gradually decrease and the misjudgment rate to gradually increase, it is avoided that a small eccentricity in the early stage of dehydration will lead to the dehydration being stopped, and the influence of the change of water content on the eccentricity during the dehydration process is fully considered, thereby improving the intelligence level of eccentricity detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 The present invention is a simplified flow chart of the washing machine eccentricity model training method.

[0040] Figure 2 The present invention is a flow chart of the washing machine eccentricity model training method.

[0041] Figure 3 The present invention is a flow chart of the washing machine eccentricity detection method. DETAILED DESCRIPTION

[0042] Exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Those skilled in the art will appreciate that the following embodiments are only used to explain the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0043] In the description of the present invention, unless otherwise clearly specified and limited, the terms "setting" and "connection" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a direct connection or an indirect connection through an intermediate medium. For those skilled in the art, the specific meanings of the above terms in the present invention can be understood according to the specific circumstances.

[0044] In the description of the present invention, it should be noted that the directions or positional relationships indicated by terms such as “upper”, “lower”, “front”, “back”, “left”, “right”, “vertical”, “inside” and “outside” are based on the directions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore cannot be understood as a limitation on the present invention.

[0045] The present invention provides a washing machine eccentricity model training method, such as Figure 1 and Figure 2 As shown, several time windows are preset along the dehydration time sequence, the eccentricity curve corresponding to each time window is analyzed, and each time window is associated and trained to obtain an eccentricity model corresponding to each dehydration duration.

[0046] In the above scheme, by training several eccentricity models arranged along the dehydration time sequence, different models can be used to judge different stages during the dehydration process. Since each eccentricity model is trained by the eccentricity curve in the corresponding time window, each eccentricity model is more targeted in judging the eccentricity, and the judgment accuracy in the corresponding time window is higher, so the occurrence of eccentricity can be known more accurately and timely.

[0047] The present invention is further described in detail below with reference to specific embodiments.

[0048] Embodiment 1

[0049] As an embodiment of the present invention, this embodiment provides a method for training an eccentricity model of a washing machine. In this embodiment, three time windows are arranged in sequence along the dehydration time sequence, which are defined as a first time window, a second time window and a third time window. The eccentricity curves in the three time windows are analyzed respectively to obtain a number of features corresponding to each time window. The three time windows are correlated with each other through the features, and the features are used to train to obtain eccentricity models corresponding to the three time windows, which are defined as a first eccentricity model, a second eccentricity model and a third eccentricity model respectively along the dehydration time sequence.

[0050] The training process is as follows Figure 1 and Figure 2 As shown, specifically:

[0051] First, a dehydration curve for training the model is prepared with eccentricity and dehydration time as coordinate axes, and then a corresponding section of the dehydration curve is intercepted according to the first time window, the second time window and the third time window for analysis to obtain features corresponding to each time window. Here, the feature obtained from the eccentricity curve in the first time window is defined as the first feature, the feature obtained from the eccentricity curve in the second time window is defined as the second feature, and the feature obtained from the eccentricity curve in the third time window is defined as the third feature; specifically, the feature includes extracting the statistical domain, spectral domain and time domain features of the eccentricity curve, and obtaining the eccentricity probability corresponding to the eccentricity curve in each time window according to its features; then part of the first feature is brought into the second time window as the second feature, and part of the second feature is brought into the third time window as the third feature, and the first eccentricity model, the second eccentricity model and the third eccentricity model are trained according to the first feature, the second feature and the third feature respectively.

[0052] Among them, the eccentricity probability is not directly used as a feature, but the eccentricity probability is converted into a probability feature and then brought into the model training process.

[0053] In the above scheme, the second feature used to train the second eccentricity model contains part of the first feature, and the third feature used to train the third eccentricity model contains part of the second feature, thereby realizing the association between the three eccentricity models. Each eccentricity model not only judges the dehydration curve within the time window, but also draws on the characteristics of the dehydration curve before entering the corresponding time window, thereby improving the accuracy of eccentricity detection.

[0054] It should be noted here that the statistical domain features in this embodiment include: the mean of the eccentricity value, the standard deviation of the eccentricity value, the skewness and kurtosis of the eccentricity curve; the spectral domain features include: the spectral distance of the eccentricity curve and the spectral fundamental frequency of the eccentricity curve; the time domain features include: the distance between the peak and the trough in the eccentricity curve, and the number of peaks; in the above scheme, the richness of the features is improved by diversifying the feature types obtained, and the judgment accuracy of the eccentricity model is further improved by fully analyzing the eccentricity curve.

[0055] Specifically, the number of features used to train the eccentricity model in the first i-th time window is k i , k i Satisfies the following relationship:

[0056] k i =M i +c i

[0057] Among them, M i represents the number of features extracted from the i-th time window; c i represents the number of features introduced from the i-1th time window; c i Is a positive integer.

[0058] The following is an example of three time windows:

[0059] By analyzing the eccentricity curves in the first time window, the second time window and the third time window respectively, M1 features corresponding to the first time window, M2 features corresponding to the second time window and M3 features corresponding to the third time window are obtained; correspondingly, the number of features k1 used in training the first eccentricity model is k1=M1, that is, the first eccentricity model is obtained by training M1 features; the number of features k2 used in training the second eccentricity model is k2=M2+c2, where c2 is the number of features selected from the features included in M1; the number of features k3 used in training the third eccentricity model is k3=M3+c3, where c3 is the number of features selected from the features included in M1 and M2.

[0060] It should be noted that the three time windows can be set at intervals or partially overlapped, and the missed detection rate and the false positive rate need to be adjusted during the training of each eccentricity model, where a missed detection refers to judging eccentricity as non-eccentricity, and a false positive refers to judging non-eccentricity as eccentricity. By adjusting the missed detection rate and the false positive rate of each eccentricity model, the false positive rate of several eccentricity models arranged in sequence along the dehydration time gradually increases, and the missed detection rate gradually decreases.

[0061] In the above scheme, by controlling the missed judgment rate and false judgment rate of each eccentricity model, the judgment of eccentricity in the early stage of dehydration is relatively loose, and as the dehydration progresses, the judgment of eccentricity by the subsequent eccentricity model is gradually tightened, taking into full consideration that a small eccentricity in the early stage may be automatically eliminated as the water content decreases, thereby avoiding the dehydration being stopped due to a small eccentricity in the early stage of dehydration, thereby improving the user experience; as the dehydration progresses, the water content is lower and the rotation speed is higher in the later stage of dehydration. At this time, by tightening the judgment standard for eccentricity, the occurrence of eccentricity can be more accurately reflected, and a prompt can be issued to the user when the eccentricity may affect the normal progress of dehydration, or the dehydration can be directly stopped to redistribute the laundry, and then the dehydration can be restarted, thereby improving the intelligence level of the washing machine.

[0062] This embodiment uses three eccentric models as an example to illustrate the technical solution of the present invention, but does not limit the number of eccentric models. Conventional improvements made by technical personnel in this field on the number of eccentric models, the size of the time window and their relative size relationship based on their familiarity with the technical principles of the present invention also fall within the protection scope of the present invention.

[0063] Embodiment 2

[0064] As another embodiment of the present invention, this embodiment further makes the following improvements on the basis of the first embodiment.

[0065] In this embodiment, the importance of features used in training each eccentricity model is ranked, and features with higher importance are screened from the previous time window and introduced into the training process of the subsequent eccentricity model.

[0066] Specifically, the M1 features, M2 features, and M3 features are sorted according to the importance of each feature, and the c2 features obtained in the previous time window introduced when training the i-th eccentric model satisfy the following relationship:

[0067]

[0068] Where n is the number of preset time windows.

[0069] Specifically in this embodiment: when training the second eccentric model, at least two features are selected according to importance for training the second eccentric model; when training the third eccentric model, at least one feature is selected from M1 according to importance, and at least two features are selected from M2, and the selected features are used together for training the third eccentric model.

[0070] In the above scheme, the selection of features is not random, but the number of features selected from each time window increases as it approaches the current time window, satisfying the change in the strength of the correlation between each eccentric model. The closer the eccentric models are, the higher the correlation between them, further improving the correlation between adjacent eccentric models.

[0071] The present invention also provides a method for detecting eccentricity of a washing machine, which uses an eccentricity model trained by the method described in the above embodiment, as follows.

[0072] Embodiment 3

[0073] As another embodiment of the present invention, this embodiment provides a method for detecting eccentricity of a washing machine, which is specifically as follows.

[0074] In this embodiment, if Figure 3 As shown, after the dehydration program starts running, the dehydration time and eccentricity are recorded in real time, and a dehydration curve with eccentricity and dehydration time as coordinate axes is drawn. During the drawing process, various eccentricity models are used to judge the eccentricity at different stages of dehydration.

[0075] Specifically, as the dehydration program proceeds, a dehydration curve is drawn in real time, and after the dehydration time traverses the first time window and forms a complete eccentricity curve in the first time window, the eccentricity curve in the first time window is analyzed to obtain a corresponding first feature, and the first feature is brought into the first eccentricity model to determine that no eccentricity has occurred, dehydration is continued, and the dehydration time traverses the second time window. After the complete eccentricity curve is formed in the second time window, the eccentricity curve in the second time window is analyzed to obtain a corresponding second feature, and some of the features with the highest importance obtained in the first time window and all of the second features are brought into the second eccentricity model together. When the second eccentricity model determines that no eccentricity has occurred, dehydration is continued until the dehydration time traverses the third time window, and a complete eccentricity curve is formed in the third time window. The eccentricity curve in the third time window is analyzed to obtain a third feature, and some of the features with the highest importance obtained in the first time window and all of the third features are brought into the third eccentricity model together; if it is still determined that no eccentricity has occurred, it is considered that eccentricity will not occur even if the speed is continued to be increased.

[0076] On the contrary, if in the above process, any eccentricity model determines that eccentricity occurs, dehydration is stopped, and the clothes are redistributed and then dehydrated again; or the rotation speed is reduced when eccentricity is determined to occur; and when eccentricity is determined to occur, an alarm is sent to the user through the washing machine or a mobile terminal or cloud connected to the washing machine to prompt the user to take action.

[0077] It should be noted here that the above-mentioned eccentricity detection does not cover the entire process of the dehydration program, but an eccentricity detection stage is set in the early stage of dehydration to perform detection. That is, if each eccentricity model determines that no eccentricity occurs, the washing machine continues to increase the speed to the highest dehydration speed for dehydration.

[0078] Furthermore, since the dehydration process usually includes multiple speed-up and speed-down stages, in order to fully judge the eccentricity, the above-mentioned judgment process is performed in each speed-up stage until the dehydration is completed.

[0079] The above description is only a preferred embodiment of the present invention, and does not limit the present invention in any form. Although the present invention has been disclosed as a preferred embodiment as above, it is not used to limit the present invention. Any technician familiar with this patent can make some changes or modifications to equivalent embodiments of equivalent changes by using the technical content suggested above without departing from the scope of the technical solution of the present invention. The implementation scheme in the above embodiment can also be further combined or replaced. However, any simple modification, equivalent change and modification made to the above embodiment based on the technical essence of the present invention without departing from the content of the technical solution of the present invention still falls within the scope of the solution of the present invention.

Claims

1. A washing machine eccentricity model training method, characterized in that: Several time windows are preset along the dehydration time sequence, the eccentricity curve corresponding to each time window is analyzed, and each time window is associated and trained to obtain an eccentricity model corresponding to each dehydration duration.

2. The washing machine eccentricity model training method according to claim 1, characterized in that: Multiple time windows arranged along the dehydration time sequence are analyzed in sequence to obtain several features corresponding to each time window, and at least part of the features obtained from the analysis of the previous time window are brought into the subsequent time window as features for training the eccentricity model corresponding to the subsequent time window.

3. The washing machine eccentric mold training method according to claim 2, characterized in that: Extract the statistical domain, spectral domain and time domain features of the eccentricity curve in each time window, analyze the statistical domain, spectral domain and time domain features to obtain the eccentricity probability in the corresponding time window, and use the statistical domain, spectral domain, time domain features and eccentricity probability of the eccentricity curve in each time window to train and obtain the eccentricity model; At least the eccentricity probability in the previous time window is brought into the subsequent time window as a feature for training the eccentricity model corresponding to the subsequent time window.

4. The washing machine eccentricity model training method according to claim 3, characterized in that: The statistical domain features include: mean of eccentricity values, standard deviation of eccentricity values, skewness and kurtosis of eccentricity curves; The spectrum domain features include: the spectrum distance of the eccentricity curve and the spectrum fundamental frequency of the eccentricity curve; The time domain features include: the distance between the crest and the trough in the eccentricity curve, and the number of crests.

5. The washing machine eccentricity model training method according to claim 3, characterized in that: The number of time windows preset along the dehydration time sequence is defined as n, and the number of features extracted from each time window is defined as m, where m>n.

6. The washing machine eccentricity model training method according to claim 5, characterized in that: The number of features used to train the eccentricity model in the first i-th time window is k i , k i Satisfies the following relationship: k i =M i +c i Among them, M i represents the number of features extracted from the i-th time window; c i Represents the number of features introduced from the features obtained in the previous time window.

7. The washing machine eccentricity model training method according to claim 6, characterized in that: The features obtained in each time window are sorted according to their importance, and the features with the highest importance ranking are selected from the features obtained in the previous time window, totaling c i Features are introduced into the i-th time window; Preferably, Where n is the number of preset time windows.

8. The washing machine eccentricity model training method according to any one of claims 1 to 7, characterized in that: Several time windows preset along the dehydration time sequence partially overlap or do not overlap each other.

9. The washing machine eccentricity model training method according to any one of claims 1 to 7, characterized in that: In the process of training the eccentric model, the missed judgment rate of several eccentric models set along the dehydration time sequence is gradually reduced, and the misjudgment rate is gradually increased.

10. A method for detecting eccentricity of a washing machine, characterized in that: An eccentric model obtained by using the eccentric model training method as described in any one of claims 1 to 9; After the dehydration program starts, a dehydration curve is drawn with the eccentricity and dehydration time as coordinate axes; When the dehydration time passes through the time window corresponding to any eccentricity model, the corresponding eccentricity model is used to judge whether eccentricity occurs. If it is judged that eccentricity does not occur, dehydration continues.

Images