A refrigerator production data collection method and system

Abstract

The present invention relates to the technical field of data processing, and in particular to a refrigerator production data collection method and system. The method comprises: preliminarily collecting multiple parameter data of a refrigerator production process, calculating the intrinsic characteristic value of a data point, clustering and denoising a data set by a fuzzy clustering algorithm, and taking the data set as the final collected data set. In each round of clustering, the fuzzy coefficient of the data point in the next round is adaptively adjusted, including: calculating the similarity between the data point and the cluster center according to the difference between the intrinsic characteristic value of the data point and any cluster center, and the difference between the collection time of the data point and the cluster center, and calculating the fuzzy coefficient of the data point in the next round according to the average difference between the similarity of the data point and all the cluster centers and the intrinsic characteristic value. The adaptive adjustment mechanism can flexibly adjust the fuzzy coefficient of the data point according to the characteristics of the data point, so as to effectively distinguish the noise data, improve the quality of the collected data, and ensure the accuracy of the subsequent data analysis.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and in particular to a refrigerator production data collection method and system. Background Art

[0002] With the development of the automobile industry and consumers' pursuit of high-quality life, car refrigerators, as key accessories to improve driving comfort, are continuously releasing their market potential. Condensers, compressors, motors, controllers, etc., which are the core components of car refrigerators, have a direct impact on the product's refrigeration efficiency, energy consumption performance and durability. In the manufacturing process, precise data collection and analysis have become an important link to achieve precise control of product quality, continuous optimization of production processes and significant improvement of production efficiency.

[0003] However, in the actual production and testing process, data noise caused by multiple factors such as equipment status, environmental factors, signal transmission quality and sensor accuracy seriously interferes with the accurate interpretation and effective use of data, thereby affecting the precise control of product quality. Therefore, before analyzing the parameter data, denoising is required to obtain high-quality collected data to ensure the accuracy and reliability of subsequent data analysis. Clustering denoising is a common data processing method, which aims to obtain the clustering cluster where the noise data is located by clustering the collected original data, and then eliminate the noise to obtain high-quality collected data. For example, the Chinese patent document with the existing announcement number CN117131336B proposes a data processing method for an electronic device connector, including: collecting several segments of current data of an electronic device connector to obtain the current data of the current segment; constructing a noise matrix of the current data of the current segment and obtaining the credibility of each column element in the noise matrix; obtaining the noise fitting value and noise probability of each component of the current data of the current segment; obtaining the corrected distance according to the noise probability and clustering analysis; denoising and reconstructing the current data of the current segment according to the clustering result to obtain the processed current data of the current segment.

[0004] Fuzzy clustering algorithm is a commonly used clustering denoising method. It controls the degree to which data points belong to each cluster, i.e., the degree of membership, through the parameter of fuzzy coefficient to achieve intelligent classification of data. However, traditional fuzzy clustering algorithms usually preset a unified fuzzy coefficient, which is shared by all data points. If the fuzzy coefficient is not selected properly, it will not only reduce the quality of clustering results, but also may cause noise data to be mistakenly included in a cluster instead of being identified or excluded as an independent noise cluster. This not only weakens the algorithm's ability to effectively identify noise data and reduces the accuracy of data denoising, but also affects the quality of collected data and the accuracy of production parameter data analysis. Summary of the invention

[0005] In order to solve the problem that when clustering and denoising the raw data collected in the refrigerator production process, improper setting of fuzzy coefficients easily leads to low data denoising accuracy, affects the quality of collected data, and reduces the accuracy of data analysis, the present invention proposes a refrigerator production data collection method and system, which specifically adopts the following scheme:

[0006] In one aspect, the present invention provides a refrigerator production data collection method, comprising:

[0007] Preliminarily collecting multiple parameter data of the refrigerator production process, each parameter data is a data set, and the data set includes multiple data points;

[0008] By the difference between any data point and other data points, the prominence of the data point and the relevance of the data point to the data set are evaluated, and the ratio of the prominence to the relevance is used as the intrinsic characteristic value of the data point;

[0009] The data set is clustered and denoised by a fuzzy clustering algorithm as the final collected data set. In each round of clustering, the fuzzy coefficient of the data point in the next round is adaptively adjusted, including:

[0010] According to the difference between the intrinsic characteristic value of the data point and any cluster center, and the difference between the acquisition time of the data point and the cluster center, the similarity between the data point and the cluster center is calculated, and the similarity is negatively correlated with the sum of the squares of the two differences;

[0011] The fuzzy coefficient of the data point in the next round is calculated according to the average difference of the similarity between the data point and all cluster centers and the intrinsic eigenvalue, and the product of the average difference and the intrinsic eigenvalue is negatively correlated with the fuzzy coefficient of the next round.

[0012] The above technical solution evaluates the intrinsic eigenvalue of the data point by the prominence of the data point and its correlation with the data set. The intrinsic eigenvalue can quantify the abnormality of the data point and effectively identify potential noise points. The fuzzy clustering algorithm is used to denoise the data set. By adaptively adjusting the fuzzy coefficient, the noise data can be more accurately identified and isolated, and the quality of the data set can be improved. By improving the data quality, the accuracy and effectiveness of the production parameter analysis are ensured, and reliable data support is provided for quality control, process optimization and efficiency improvement in the refrigerator production process. The adaptive adjustment process includes: by considering the intrinsic eigenvalue difference and acquisition time difference between the data point and the cluster center, the similarity between the data point and the cluster center is calculated. This calculation method takes into account the static attributes and dynamic time series characteristics of the data point, and improves the accuracy of the clustering results. Based on the average difference and intrinsic eigenvalue of the similarity between the data point and all cluster centers, the fuzzy coefficient of the data point in the next round is dynamically adjusted. This adaptive adjustment mechanism can flexibly change its attribution weight according to the actual characteristics of the data point, avoid the limitations of the traditional fixed fuzzy coefficient, effectively distinguish noise data, and enhance the adaptability of the algorithm to complex data sets.

[0013] Furthermore, the fuzzy coefficient of the data point in the next round satisfies the following relationship:

[0014] ;

[0015] In the formula, For the The data point in The fuzzy coefficient in the round, For the The data point in The fuzzy coefficient in the round, is the natural exponential function, For the The intrinsic eigenvalues ​​of the data points, For the The data point in The average difference in similarity to all cluster centers in a round.

[0016] The above technical solution dynamically adjusts the fuzzy coefficient of each data point in the iteration process by introducing the intrinsic eigenvalue of the data point and the average difference of clustering similarity, thereby realizing the adaptive response of the algorithm to the data characteristics and improving the accuracy and robustness of the clustering results. Through the coupling of intrinsic eigenvalues ​​and similarity differences, the adaptive adjustment of the fuzzy coefficient can more keenly identify and isolate noise data points and avoid their inappropriate influence on the cluster center, thereby improving the purity of the data set and the reliability of the analysis.

[0017] Furthermore, the average difference of the similarity is calculated as follows:

[0018] ;

[0019] In the formula, For the The data point in The average difference in similarity to all cluster centers in a round, For the In round Data points and The similarity of the cluster centers, For the In round Data points and The similarity of the cluster centers, For the The serial numbers of the other cluster centers other than the cluster centers, is the number of cluster centers.

[0020] The above technical solution calculates the average difference in similarity between data points and all cluster centers. This formula ensures that the attribution of data points is not only based on their similarity with a single cluster center, but also comprehensively considers their relationship with all cluster centers, thereby enhancing the comprehensiveness and accuracy of the clustering results. The calculation of the average difference can more finely identify the subtle differences between data points and cluster centers, and for those data points with low similarity with all cluster centers, it can more accurately judge the possibility that they are noise points.

[0021] Furthermore, the step of evaluating the prominence of any data point by comparing the difference between any data point and other data points includes:

[0022] taking the absolute value of the difference between the data point and the average of the data set as the first difference value;

[0023] Calculate the cumulative sum of the differences between the data point and all other data points in the data set, and use the average of the cumulative sums as the second difference value;

[0024] The ratio of the first difference value to the second difference value is used as the prominence degree of the data point.

[0025] The above technical solution can comprehensively evaluate the prominence of data points from both individual and group dimensions by calculating the absolute value of the difference between the data point and the average of the data set (the first difference value), as well as the average of the cumulative sum of the differences between the data point and all other data points in the data set (the second difference value), ensuring the comprehensiveness and accuracy of the evaluation. The first difference value focuses on the degree of deviation of the data point from the overall average level, while the second difference value focuses on the relative difference between the data point and other data points in the set. The ratio of the two as the prominence can effectively amplify the characteristics of abnormal data points and improve the algorithm's sensitivity and recognition rate to noise data.

[0026] Furthermore, the fuzzy coefficient of the data point in the first round is a preset value.

[0027] Furthermore, the similarity between the data point and the cluster center satisfies the following relationship:

[0028] ;

[0029] In the formula, For the In round Data points and The similarity of the cluster centers, For the The intrinsic eigenvalues ​​of the data points, For the The intrinsic eigenvalues ​​of the cluster centers, For the The collection time of data points, For the The collection time of the cluster center points, is a natural exponential function.

[0030] The above technical solution not only takes into account the intrinsic eigenvalue differences between data points and cluster centers, but also incorporates the time differences at the collection moments. Through the natural attenuation characteristics of the exponential function, it can effectively quantify the impact of these two differences on similarity. This formula can more accurately evaluate the similarity between data points and cluster centers, and help identify data points that deviate significantly in features or time.

[0031] Furthermore, the stopping condition of the clustering is:

[0032] If a cluster contains the least number of data points in the current round, and the number of data points contained in the cluster is the same as the number of data points contained in the cluster in the previous round, the cluster is regarded as a cluster with noise and clustering stops.

[0033] The above technical solution monitors the dynamic changes of data point distribution during the clustering process, especially focusing on those clusters with a small number of data points that remain unchanged in consecutive rounds. The stopping condition can effectively identify clusters that may contain noise data, thereby discovering and isolating potential noise sources in the early stage of cluster analysis.

[0034] Furthermore, the cluster denoising also includes: performing a data smoothing operation on data points in the noisy clusters to complete denoising of the data set.

[0035] On the other hand, the present invention also provides a refrigerator production data acquisition system, the data acquisition system comprising a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps of any one of the acquisition methods.

[0036] The present invention has the following effects:

[0037] The present invention clusters and denoises the parameter data in the refrigerator production process through fuzzy clustering, and dynamically adjusts the fuzzy coefficient of the fuzzy clustering algorithm during the clustering process. This adaptive adjustment mechanism can flexibly change the attribution weight of the data point according to its actual characteristics, avoiding the limitations of traditional fixed fuzzy coefficients, and can effectively distinguish noise data, ensuring the quality of collected data, thereby improving the accuracy and reliability of subsequent parameter data analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] By reading the following detailed description with reference to the accompanying drawings, the above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:

[0039] Figure 1 It is a schematic flow chart of the method of the present invention. DETAILED DESCRIPTION

[0040] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0041] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0042] Reference Figure 1 , a refrigerator production data collection method, including steps S1 to S3, specifically as follows:

[0043] S1: Preliminary collection of a data set of multiple parameters of the refrigerator production process.

[0044] In the process of refinement and intelligence in refrigerator manufacturing, the performance and quality of core components such as compressors, motors, controllers, and condensers are directly related to the overall efficiency of the refrigerator. In order to ensure that the production quality of these key parts meets the standards, accurate data collection and analysis technology is required. By analyzing various parameters in the production process, quality control, process optimization, and efficiency improvement can be achieved. However, due to the complexity of the actual production environment and the unique working principles and physical properties of each component, the data collection process is extremely susceptible to various interferences, generating noise data, thereby affecting the accuracy and reliability of data analysis.

[0045] For different key parts, the parameter data that needs to be collected is also different, and the reasons that affect the parameter data and generate noise interference data are also different, such as:

[0046] For compressors, the parameter data that need to be collected include: current, voltage, temperature, pressure and vibration characteristics monitored during operation. These parameter data are subject to electromagnetic interference, mechanical vibration and the instability of refrigerant flow, which will introduce noise;

[0047] For motors, the parameter data that need to be collected include: speed, torque, power, temperature and current waveform. These parameter data are affected by power grid fluctuations, electromagnetic interference, mechanical vibration and temperature changes in the working environment, which will cause abnormal fluctuations in the data and the existence of noisy data.

[0048] For the controller, the parameter data that needs to be collected include: recording input / output signals, control instructions, feedback signals, operating status and fault codes. These parameter data are easily affected by signal transmission interference, software instability and external environment changes, which can generate noise data.

[0049] For the condenser, the parameter data that need to be collected include: temperature, pressure, heat exchange efficiency and fluid flow rate. The noise of these parameter data comes from fluid vibration, unstable heat exchange, ambient temperature fluctuation and sensor accuracy limitation.

[0050] In this embodiment, the parameter data of the above-mentioned key parts in the refrigerator production process can be processed according to the subsequent methods. This embodiment only takes the collected temperature data of the condenser as an example for explanation. In order to accurately monitor the operating status of the vehicle refrigerator condenser and ensure its efficient and stable working performance, a temperature sensor is installed at a key position of the vehicle refrigerator condenser to monitor its temperature changes during operation in real time. The data acquisition frequency of the sensor is set to 1Hz, that is, the temperature data is collected once per second to capture the slight fluctuations in the condenser temperature data. The temperature data collected every 15 minutes is aggregated into a data set, and then the data set is subsequently analyzed and processed.

[0051] When the present embodiment uses the fuzzy clustering algorithm to perform the denoising task on the initially collected data set, it mainly identifies and removes those abnormal clusters formed in the clustering process. These abnormal clusters usually show two situations: one is a small cluster whose scale is significantly smaller than other clusters, and the other is an outlier that is isolated in position and keeps a large distance from other clusters. The high-quality denoising effect directly depends on the accuracy of clustering, that is, each data point is correctly assigned to the cluster that is most similar to it. In order to achieve this goal, the clustering algorithm needs to accurately measure the similarity between data points, which is usually achieved by calculating the distance or similarity between points. Ideally, data points in the same cluster should have a high degree of similarity, while data points between different clusters should show obvious differences. Therefore, small clusters or clusters far away from the main clusters often represent noise or outliers in the data set, which may be caused by measurement errors, data entry errors or abnormal events.

[0052] Noise data usually appears as isolated points that deviate significantly from the mainstream pattern in the data set. They are difficult to clearly attribute to any known cluster center. In the fuzzy clustering algorithm, the membership of a data point is determined by the fuzzy coefficient, which quantifies the possibility of a data point belonging to different clusters. When a data point has a relatively low membership in all clusters, it often indicates that it may be a noise point. However, if the same fuzzy coefficient is uniformly applied to all data points, the identification of noise data may become unclear, because the fixed fuzzy coefficient cannot fully reflect the inherent differences and complexity between data points, which will not only weaken the sensitivity of the clustering algorithm to noise, but also confuse the boundary between normal data and noise.

[0053] In view of this, the characteristic performance of data points becomes a key consideration. In order to ensure the accuracy and rationality of the clustering results, the selection of fuzzy coefficients should be closely aligned with the characteristic performance of data points, that is, the fuzzy coefficients should be dynamically adjusted according to the specificity of each data point and the environment in which it is located.

[0054] Based on the above analysis, this embodiment designs an effective noise data identification strategy: by analyzing the characteristic performance of each data point, a fuzzy coefficient is tailored for it. For data points suspected of being noise, a smaller fuzzy coefficient can be set to limit their degree of affiliation to any cluster center. In this way, these data points will naturally tend to cluster together to form independent clusters because they are difficult to meet any clear clustering requirements, thereby achieving effective separation and identification of noise data.

[0055] S2: Calculate the intrinsic eigenvalues ​​for each data point in the dataset.

[0056] The intrinsic eigenvalues ​​of a data point are composed of two aspects: one is the distribution characteristics of the data point itself, and the other is the relationship between the data point and other members in the data set. These two together shape the uniqueness and environment of the data point, which is crucial to revealing the essential structure of the data. The distribution characteristics of data points are an important part of their intrinsic eigenvalues ​​and provide a basis for understanding and modeling data.

[0057] Therefore, this step evaluates the prominence of any data point and the correlation between the data point and the data set through the difference between the data point and other data points, and takes the ratio of the prominence to the correlation as the intrinsic characteristic value of the data point.

[0058] The method for evaluating the prominence of any data point by the difference between any data point and other data points is as follows:

[0059] The absolute value of the difference between the data point and the average of the data set is used as the first difference value; the cumulative sum of the differences between the data point and all other data points in the data set is calculated, and the average of the cumulative sum is used as the second difference value; and the ratio of the first difference value to the second difference value is used as the prominence of the data point.

[0060] The formula is:

[0061]

[0062] In the formula, For the The prominence of a data point, is the total number of data points contained in the dataset, For the The value of the data point, Indicates that except for the first The serial numbers of other data points other than data points, Indicates the other data points The value of the data point, Indicates that except for the first The total number of data points other than data points, The average value of all data points contained in the data set. In this formula, is the first difference value mentioned above, is the second difference value mentioned above.

[0063] The correlation between the data point and the data set is calculated using an existing DTW (Dynamic Time Warping) algorithm. When calculating the correlation between the data point and the entire data set, the DTW algorithm may be performed in the following steps:

[0064] Initialization: treat all data points in the dataset as a time series;

[0065] Calculate the distance matrix: For any data point, use the DTW algorithm to calculate the distance between the data point and each data point in the data set. This step will result in a distance matrix, in which each element represents the DTW distance between the data point and a data point in the data set.

[0066] Distance matrix aggregation: All DTW distances in the distance matrix are aggregated and analyzed. There are two commonly used methods to evaluate the relevance of a data point to a data set: one is the average DTW distance, which is calculated by taking the average of all DTW distances as the relevance of a data point to a data set. This provides a measure of the average similarity between a data point and a data set. The smaller the average DTW distance, the more similar the data point is to other data points in the data set. The other is the median DTW distance, which is calculated by taking the median of all DTW distances as the relevance of a data point to a data set. Compared to the average, the median is more resistant to the influence of outliers. Therefore, when there is noise or outliers in the data set, it is more appropriate to use the median DTW distance.

[0067] A smaller DTW distance average or median indicates that the data point has a higher similarity with most data points in the data set, so it can be considered that the data point is highly correlated with the data set. A larger DTW distance average or median indicates that the data point may be an outlier or noise point, and the data point does not match other data points in the data set, that is, the data point is less correlated with the data set.

[0068] Among them, the intrinsic characteristic value of the data point is expressed by the formula:

[0069]

[0070] In the formula, For the The intrinsic eigenvalues ​​of the data points, For the The prominence of a data point, For the The degree of relevance of a data point to the dataset.

[0071] The significance of this formula is that it can simultaneously reflect the abnormality of a data point and its representativeness in the data set. The larger the intrinsic eigenvalue, the more likely the data point is an abnormal data point. Specifically:

[0072] When the prominence of a data point is large and the correlation between the data point and its data set is small, this indicates that the data point has both significant abnormality and is not correlated with other data points in the data set. The data point may be noise or an outlier. Conversely, if the prominence of a data point is small and the correlation between the data point and its data set is large, this means that the data point is similar to other data points in the data set and is more likely to be an ordinary normal data point.

[0073] S3: clustering and denoising the data set using a fuzzy clustering algorithm. In each round of clustering, the fuzzy coefficient of the data point in the next round is adaptively adjusted based on the intrinsic eigenvalue of the data point.

[0074] In the fuzzy clustering algorithm, the fuzzy coefficient plays a vital role. Its function is to control the fuzziness of the clustering results, that is, the degree to which the data points belong to multiple clusters. A lower fuzzy coefficient will cause the data points to be more inclined to belong to only one cluster, while a higher fuzzy coefficient allows the data points to be distributed in multiple clusters with a more uniform probability.

[0075] As an iterative optimization process, the core of fuzzy clustering is to continuously adjust the position of the cluster center and the membership of the data point to the cluster until a stable state is reached. This process not only involves the update of the cluster center, but also the fine adjustment of the membership of the data point. The membership of the cluster center and the data point changes in each round of iteration.

[0076] Therefore, the degree of association between the data points and the cluster centers in each round of clustering will affect the selection of the fuzzy coefficient in the next round of clustering. The greater the difference in the degree of association between the data points and the cluster centers in the current round, the clearer the cluster to which the data points belong. Therefore, a smaller fuzzy coefficient is required in the next round. In other words, a suitable fuzzy coefficient can help the clustering algorithm converge to a stable result faster.

[0077] Since traditional fuzzy clustering algorithms use the same fixed fuzzy coefficient for all data points during the clustering process, the differences in data point characteristics are not fully considered. In complex data sets, the characteristics of different data points may vary greatly, including their distances from different cluster centers, the relationships between data points, and the density distribution of data points. The use of a fixed fuzzy coefficient ignores these differences, causing the algorithm to perform poorly when processing data subsets with different intrinsic structures, and it is easy to confuse noise with normal data. That is, if all data points share the same fuzzy coefficient, noise data may be incorrectly assigned to a cluster instead of being regarded as an independent noise cluster or excluded. This is because the fixed fuzzy coefficient may not be sufficient to distinguish the accurate attribution of noise data and normal data points.

[0078] In view of the above, this embodiment designs a strategy of adaptive fuzzy coefficient, which allows the algorithm to dynamically adjust the fuzzy coefficient of each data point during the clustering process according to the characteristics of the data point. In this way, the algorithm can more accurately evaluate the similarity between the data point and the cluster center, effectively identify and isolate noise data, and enhance the accuracy and adaptability of the clustering results. It provides high-quality clustering results in complex data sets and obtains a high-quality data set as the final collected data set, which provides a reliable and accurate data foundation for subsequent data analysis.

[0079] Specifically, this step first initializes the clustering parameters and sets the number of cluster centers. (Experience value), the fuzzy coefficient of each data point in the first round is a preset value, such as The fuzziness coefficient of the first round of data points (initial fuzziness coefficient) is set to , the cluster centers are randomly selected.

[0080] Then, in each round of clustering, the fuzzy coefficient of the data point in the next round is adaptively adjusted, including:

[0081] S31: Calculate the similarity between the data point and the cluster center according to the difference between the intrinsic characteristic value of the data point and any cluster center, and the difference between the acquisition time of the data point and the cluster center, wherein the similarity is negatively correlated with the sum of the squares of the two differences.

[0082] The similarity between the data point and the cluster center satisfies the following relationship:

[0083]

[0084] In the formula, For the In round Data points and The similarity of the cluster centers, For the The intrinsic eigenvalues ​​of the data points, For the The intrinsic eigenvalues ​​of the cluster centers, For the The collection time of data points, For the The collection time of the cluster center points, is a natural exponential function. For the data points and The difference in the intrinsic eigenvalues ​​of the cluster centers, For the data points and The difference in the collection time of the cluster center points, represents the sum of the squares of these two differences.

[0085] This formula takes into account both the intrinsic eigenvalue differences and the acquisition time differences between the data points and the cluster centers, which makes the similarity calculation more comprehensive. The intrinsic eigenvalue differences reflect the proximity of the data points' own characteristics to the cluster centers, while the acquisition time differences take into account the synchronization of the data points and the cluster centers in the time dimension. This dual consideration ensures the comprehensiveness and accuracy of the similarity evaluation. Similarity is negatively correlated with the sum of the squares of the intrinsic eigenvalue differences and the acquisition time differences, and is calculated using a natural exponential function. This exponential decay characteristic means that when the difference between the data points and the cluster centers in intrinsic eigenvalues ​​or acquisition time increases, their similarity decreases rapidly, and vice versa, which ensures that the algorithm can sensitively respond to the proximity of the data points and the cluster centers in terms of attributes and time.

[0086] S32: Calculate the fuzzy coefficient of the data point in the next round according to the average difference of the similarity between the data point and all cluster centers and the intrinsic eigenvalue, and the product of the average difference and the intrinsic eigenvalue is negatively correlated with the fuzzy coefficient of the next round.

[0087] Among them, the average difference of similarity is calculated as:

[0088]

[0089] In the formula, Indicates The data point in The average difference in similarity to all cluster centers in a round, Indicates In round Data points and The similarity of the cluster centers, Indicates In round Data points and The similarity of the cluster centers, For the The serial numbers of the other cluster centers other than the cluster centers, is the number of cluster centers. The data point in The larger the average difference in similarity with all cluster centers in the round, the The more likely the data point is noise data.

[0090] Here, by calculating the average difference between the similarity of the data point and all cluster centers, the fuzzy coefficient adjustment of the data point does not only rely on its similarity with the nearest or most relevant cluster center, but takes into account the influence of all cluster centers, which ensures that the fuzzy coefficient adjustment of the data point is more comprehensive and objective.

[0091] Among them, the fuzzy coefficient of the data point in the next round satisfies the following relationship:

[0092]

[0093] In the formula, For the The data point in The fuzzy coefficient in the round, For the The data point in The fuzzy coefficient in the round, is the natural exponential function, For the The intrinsic eigenvalues ​​of the data points, For the The data point in The average difference in similarity to all cluster centers in a round.

[0094] The formula combines the intrinsic eigenvalues ​​of data points and the average difference of similarity to achieve adaptive adjustment of the fuzzy coefficient. The adjustment of the fuzzy coefficient helps data points gradually find their most appropriate cluster affiliation and reduces the frequent conversion of data points between multiple clusters, thereby improving the stability of clustering results. A smaller fuzzy coefficient means that data points are more likely to belong to a certain cluster clearly, which helps to form clusters with clear boundaries and stable structures.

[0095] Among them, the intrinsic eigenvalue reflects the degree of abnormality of the data point, while the similarity difference reflects the degree of matching between the data point and the cluster center. This comprehensive consideration ensures that the fuzzy coefficient can be based on the characteristics of the data point and its dynamic changes in the clustering process. By reducing the fuzzy coefficient of data points with large intrinsic eigenvalues ​​or large similarity differences with the cluster center, the algorithm can more quickly assign these data points to one or a few clusters, thereby accelerating the convergence of the algorithm. This mechanism is conducive to processing data sets containing noise or outliers. The fuzzy coefficient of data points with high intrinsic eigenvalues ​​and large similarity differences will be significantly reduced, which helps to isolate noise data points or outliers from the main clusters to form independent noise clusters. This mechanism can effectively improve the purity of the data set and provide higher quality data support for subsequent data analysis and decision-making.

[0096] In summary, this step is to more finely control the fuzzy coefficient of the data point in the next round of clustering by comprehensively considering the intrinsic eigenvalue of the data point and the average similarity difference between the data point and the cluster center during the clustering process, thereby controlling the membership distribution of the data point, thereby accelerating the convergence of the algorithm and improving the clustering quality. Specifically:

[0097] The larger the intrinsic eigenvalue of a data point is, the more prominent it is in the data set and the less relevant it is to the overall data set. This suggests that the data point may be noise or an outlier, so in the next round of clustering, a smaller fuzzy factor should be given to it to prompt the algorithm to more clearly classify these data points into a cluster, or treat them as independent noise clusters to avoid them being ambiguously distributed among multiple clusters.

[0098] The average similarity difference between data points and cluster centers reflects the changing characteristics of data points in the clustering process and the attribution tendency of data points. A larger average similarity difference (changing characteristic) means that the similarity between data points and a specific cluster center is significantly higher than that of other clusters, which indicates that data points tend to belong to a clear cluster. Therefore, in the next round of clustering, these data points should be given a smaller fuzzy factor to accelerate their clear attribution to the cluster they belong to and ensure that the algorithm quickly converges to a stable clustering result.

[0099] S33: Set the stopping condition of clustering: if a cluster in the current round contains the least number of data points, and the number of data points contained in the cluster is the same as the number of data points contained in the cluster in the previous round, the cluster is regarded as a cluster with noise, and the clustering stops to obtain the final clustering result, that is, multiple clusters are obtained, and the clusters with noise data have been screened out.

[0100] Here, by observing the changes in the number of data points in the clusters, the stopping condition can quickly identify those clusters with the smallest number of data points that remain unchanged in consecutive rounds. These clusters are likely to contain noise data points or outliers because they fail to attract new data points and cannot retain the original data points, indicating that these data points have a very low correlation with the cluster center or are significantly different from other data points.

[0101] Furthermore, after obtaining the noisy clusters, a data smoothing operation is performed on the data points in the noisy clusters to denoise the data set, obtain the final collected data set, and perform subsequent data analysis based on the final collected data set.

[0102] Among them, data smoothing, as an effective denoising technology, is widely used in signal processing and data preprocessing stages to improve data quality and reliability. When a data point is identified as a member of a noise cluster, that is, it is considered to be an outlier or an atypical data point caused by external interference, it is processed by a time series-based smoothing method, which can significantly reduce the impact of noise on subsequent data analysis and decision-making. The smoothing method uses the data of the adjacent moments before and after the data point is collected for smoothing. Specifically, if the data collected at the 5th second is determined to be noise data, then the data collected at the 4th and 6th seconds can be selected, and the mean of these two data can be calculated as the smoothed value of the data collected at the 5th second. This time series-based smoothing method essentially eliminates or weakens the impact of noise data points through local averaging while retaining the basic trends and characteristics of the data. By replacing noise data points with the average values ​​of adjacent data points, the impact of random noise can be effectively suppressed and the purity of the data can be improved. Moreover, since adjacent data points usually have a high degree of correlation, using their mean as a substitute value can better reflect the continuous changes in the data. While removing noise, the original trend of the data is maintained as much as possible. It is particularly suitable for real-time data processing and online monitoring systems, and can quickly respond to and process noise points in the data.

[0103] By processing the noise clusters in this step, high-quality final collected data after denoising is obtained. Based on the final collected data, accurate data analysis can be performed to obtain reliable analysis results.

[0104] The present invention also provides a refrigerator production data acquisition system, the data acquisition system includes a memory and a processor, the memory stores a computer program, and the processor executes the computer program to implement the steps of any one of the detection methods.

[0105] In the description of this specification, "plurality" or "several" means at least two, such as two, three or more, etc., unless otherwise clearly and specifically defined.

[0106] Although this specification has shown and described a number of embodiments of the present invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art will conceive of many modifications, changes and alternatives without departing from the ideas and spirit of the present invention. It should be understood that in the practice of the present invention, various alternatives to the embodiments of the present invention described herein may be employed.

Claims

1. A refrigerator production data collection method, characterized in that: include: Preliminarily collecting multiple parameter data of the refrigerator production process, each parameter data is a data set, and the data set includes multiple data points; By the difference between any data point and other data points, the prominence of the data point and the relevance of the data point to the data set are evaluated, and the ratio of the prominence to the relevance is used as the intrinsic characteristic value of the data point; The data set is clustered and denoised by a fuzzy clustering algorithm as the final collected data set. In each round of clustering, the fuzzy coefficient of the data point in the next round is adaptively adjusted, including: According to the difference between the intrinsic characteristic value of the data point and any cluster center, and the difference between the acquisition time of the data point and the cluster center, the similarity between the data point and the cluster center is calculated, and the similarity is negatively correlated with the sum of the squares of the two differences; The fuzzy coefficient of the data point in the next round is calculated based on the average difference of the similarity between the data point and all cluster centers and the intrinsic eigenvalue. For data points suspected of noise, a smaller fuzzy coefficient is set to limit the degree of membership of the data points suspected of noise to any cluster center, so that the data points suspected of noise tend to gather together to form independent clusters, thereby achieving effective separation and identification of noise data; the product of the average difference and the intrinsic eigenvalue is negatively correlated with the fuzzy coefficient of the next round.

2. The refrigerator production data collection method according to claim 1, characterized in that: The fuzzy coefficient of the data point in the next round satisfies the following relationship: ； In the formula, For the The data point in The fuzzy coefficient in the round, For the The data point in The fuzzy coefficient in the round, is the natural exponential function, For the The intrinsic eigenvalues ​​of the data points, For the The data point in The average difference in similarity to all cluster centers in a round.

3. The refrigerator production data collection method according to claim 2, characterized in that: The average difference of the similarity is calculated as: ； In the formula, For the The data point in The average difference in similarity to all cluster centers in a round, For the In round Data points and The similarity of the cluster centers, For the In round Data points and The similarity of the cluster centers, For the The serial numbers of the other cluster centers other than the cluster centers, is the number of cluster centers.

4. The refrigerator production data collection method according to claim 2, characterized in that: The evaluation of the prominence of any data point by the difference between any data point and other data points includes: taking the absolute value of the difference between the data point and the average of the data set as the first difference value; Calculate the cumulative sum of the differences between the data point and all other data points in the data set, and use the average of the cumulative sums as the second difference value; The ratio of the first difference value to the second difference value is used as the prominence degree of the data point.

5. The refrigerator production data collection method according to claim 2, characterized in that: The fuzzy coefficient of the data point in the first round is a preset value.

6. The refrigerator production data collection method according to claim 3, characterized in that: The similarity between the data point and the cluster center satisfies the following relationship: ； In the formula, For the In round Data points and The similarity of the cluster centers, For the The intrinsic eigenvalues ​​of the data points, For the The intrinsic eigenvalues ​​of the cluster centers, For the The collection time of data points, For the The collection time of the cluster center points, is a natural exponential function.

7. The refrigerator production data collection method according to claim 1, characterized in that: The stopping condition of the clustering is: If a cluster contains the least number of data points in the current round, and the number of data points contained in the cluster is the same as the number of data points contained in the cluster in the previous round, the cluster is regarded as a cluster with noise and clustering stops.

8. The refrigerator production data collection method according to claim 7, characterized in that: The cluster denoising further includes: performing a data smoothing operation on data points in the noisy clusters to complete the denoising of the data set.

9. A refrigerator production data collection system, characterized in that: The data acquisition system comprises a memory and a processor, wherein a computer program is stored in the memory, and the processor executes the computer program to implement the steps of the acquisition method according to any one of claims 1 to 8.

Images