Data Analysis Method and System for Smart Kitchen Appliances Based on IoT Cloud Platform

Abstract

This invention belongs to the field of data processing technology, specifically relating to a data analysis method and system for intelligent kitchen appliances based on an IoT cloud platform. The method includes: acquiring the historical state feature sequence of the kitchen appliance at the current evaluation time point; performing dual-path parallel time-series feature analysis on the historical state feature sequence, calculating the time-series context vector, time volatility index, and cumulative severity score respectively; constructing a dynamic risk signature based on the time volatility index and cumulative severity score, and combining the current state and historical context to form an enhanced feature vector; and inputting the enhanced feature vector into a Cox proportional hazards model to calculate real-time fault risk. This invention, by deeply mining the time-series evolution patterns and dynamic dependencies of the equipment's operating state, constructs an enhanced feature set containing multi-dimensional time-series information, significantly improving the accuracy of predicting kitchen appliance fault risks.

Description

Technical Field

[0001] This invention relates to the field of data processing technology. More specifically, this invention relates to a data analysis method and system for smart kitchen appliances based on an Internet of Things (IoT) cloud platform. Background Technology

[0002] With the popularization of IoT technology, smart kitchen appliances have become an important part of modern families. During operation, these devices generate massive amounts of multi-dimensional data through built-in sensors, which are then collected through IoT cloud platforms.

[0003] In the field of equipment life prediction and failure risk assessment, the Cox proportional hazards model is a widely used semi-parametric survival analysis method. This model can effectively establish the relationship between multiple influencing factors and the risk of event occurrence without pre-assuming the basic functional form of risk changing over time, and has good flexibility and interpretability.

[0004] However, existing technologies applying the Cox model typically use a snapshot of the equipment's operating parameters at a specific moment as the model's input features. For example, they directly use static feature values ​​such as current, temperature, and vibration of the equipment at the current moment to construct feature vectors. But this method completely ignores the evolution patterns and dynamic dependencies of the equipment's state features over time. Failures in kitchen appliances are often not instantaneous events, but rather a long-term cumulative process of gradual deterioration. For instance, before a failure, the equipment may experience a transition from a stable operating state to one of violent fluctuations. Due to the static nature of its input, the standard Cox model cannot capture these crucial time-series dynamic information, thus greatly limiting its predictive accuracy and robustness, making it difficult to effectively identify potential failure risks caused by accumulated state changes and dynamic fluctuations. Summary of the Invention

[0005] To address the technical problem of inaccurate prediction of kitchen appliance failure risks in existing technologies, this invention provides solutions in the following aspects.

[0006] In a first aspect, the present invention provides a data analysis method for intelligent kitchen appliances based on an Internet of Things (IoT) cloud platform, comprising: acquiring a historical state feature sequence of the kitchen appliances at the current assessment time point, the historical state feature sequence being composed of multiple feature vectors arranged in chronological order; performing dual-path parallel time-series feature analysis on the historical state feature sequence, the dual-path parallel time-series feature analysis comprising: calculating a time-series context vector and a time volatility index based on the time dynamics of the historical state feature sequence; and calculating a cumulative severity score based on the state severity of the historical state feature sequence; constructing a dynamic risk signature based on the time volatility index and the cumulative severity score, and combining the feature vector at the current assessment time point with the time-series context vector to jointly constitute an enhanced feature vector; and inputting the enhanced feature vector into a preset Cox proportional hazards model to calculate the real-time failure risk of the kitchen appliances, thereby realizing the monitoring of the failure risk of the kitchen appliances.

[0007] This invention uses a dual-path parallel processing module to deeply mine the time-series evolution patterns and dynamic dependencies of equipment operating status, and constructs an enhanced feature set that includes time dynamics and state severity. This enables the risk model to not only perceive the current state, but also understand the evolution history, fluctuation patterns and cumulative damage of the state. As a result, it can identify potential failure risks caused by cumulative changes and dynamic fluctuations of the state earlier, and improve the accuracy and robustness of prediction.

[0008] Preferably, the step of obtaining the historical state feature sequence of the kitchen appliance at the current evaluation time point includes: obtaining the multi-dimensional sensor data stream of the kitchen appliance within a preset monitoring period through an IoT cloud platform; setting a fixed-length time window and sliding step size to perform sliding sampling on the multi-dimensional sensor data stream; calculating statistical features for the data within each time window to generate a feature vector; and arranging all feature vectors generated within the monitoring period in chronological order to form the historical state feature sequence.

[0009] By transforming the raw, continuous sensor data stream into a structured sequence of feature vectors, high-quality, standardized input is provided for subsequent deep time-series analysis, ensuring the effectiveness and comparability of the analysis.

[0010] Preferably, the statistical features include at least one of time-domain features, frequency-domain features, and physical model features.

[0011] By extracting features from multiple dimensions such as the time domain, frequency domain, and physical model, the operating status of the equipment in each time window can be more comprehensively characterized, providing a rich information foundation for accurately capturing fault symptoms.

[0012] Preferably, the method for calculating the temporal context vector is as follows: for each feature vector in the historical state feature sequence, calculate the significance weight based on its time decay effect and the drasticness of local state changes; and perform a weighted average of all feature vectors in the historical state feature sequence to obtain the temporal context vector.

[0013] Preferably, the time volatility index is calculated as follows: the local state change intensity between adjacent feature vectors in the historical state feature sequence is calculated to form a change intensity sequence; the standard deviation of the change intensity sequence is calculated to obtain the time volatility index.

[0014] By introducing a weighted summary vector that can represent the entire historical sequence, we can focus on recent states and key historical turning points, thereby effectively capturing long-term temporal dependencies.

[0015] Preferably, the cumulative severity score is calculated as follows: for each feature vector in the historical state feature sequence, the weighted distance between it and the predefined health state feature vector is calculated as the single-point severity; the single-point severity of all time points in the historical state feature sequence is averaged to obtain the cumulative severity score.

[0016] Preferably, the dynamic risk signature is calculated as follows: the cumulative severity score is used as the basic risk item, and the time volatility index is used as the exponential amplification factor. The dynamic risk signature is obtained by multiplying the basic risk item by the natural exponential value of the exponential amplification factor.

[0017] By nonlinearly fusing the severity and volatility of the state, the risk amplification effect of high-risk scenarios with severe fluctuations under high load is accurately assessed, making the risk assessment results more consistent with physical reality and significantly improving the sensitivity to near-failure states.

[0018] Preferably, the enhanced feature vector is formed by concatenating and combining the feature vector at the current evaluation time point, the temporal context vector, and the dynamic risk signature.

[0019] Secondly, the present invention provides a smart kitchen appliance data analysis system based on an Internet of Things (IoT) cloud platform, comprising a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the aforementioned smart kitchen appliance data analysis method based on an IoT cloud platform is implemented.

[0020] By adopting the above technical solution, a data analysis method for smart kitchen appliances based on an Internet of Things cloud platform is generated into a computer program and stored in a memory for loading and execution by a processor. This allows for the creation of terminal devices based on the memory and processor, making them convenient to use.

[0021] This invention extracts two core dimensions—time dynamics and state severity—from raw time-series data and nonlinearly fuses them into a dynamic risk signature. This enables the model to automatically capture and assess the evolution history, fluctuation patterns, and cumulative damage of equipment states, greatly enriching the information dimensions input into the Cox risk model.

[0022] Furthermore, by constructing enhanced feature vectors rich in temporal information, the Cox model can not only assess the risk of the current state but also gain insights into the future risks predicted by the evolution of historical states. This significantly improves the accuracy and lead time of fault prediction, enabling predictive maintenance to shift from passive response to proactive prevention and enhancing the model's adaptability and generalization ability. Attached Figure Description

[0023] The above and other objects, features, and advantages of exemplary embodiments of the present invention will become readily apparent upon reading the following detailed description with reference to the accompanying drawings. In the drawings, several embodiments of the invention are illustrated by way of example and not limitation, and like or corresponding reference numerals denote like or corresponding parts, wherein:

[0024] Figure 1 This is a flowchart illustrating the data analysis method for smart kitchen appliances based on an Internet of Things cloud platform in this invention;

[0025] Figure 2 This is a schematic diagram illustrating the changes in raw sensor data between healthy and faulty devices throughout their entire lifecycle, according to an embodiment of the present invention; wherein the upper sub-graph is a comparison diagram of current data, and the lower sub-graph is a comparison diagram of vibration data;

[0026] Figure 3 This is a schematic diagram illustrating the evolution of key timing characteristics of a faulty device according to an embodiment of the present invention;

[0027] Figure 4 This is a comparison diagram of dynamic risk signatures provided according to an embodiment of the present invention. Detailed Implementation

[0028] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0029] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0030] This invention discloses a data analysis method for smart kitchen appliances based on an Internet of Things (IoT) cloud platform, referring to... Figure 1 This includes steps S1-S4:

[0031] S1. Obtain the historical state feature sequence of the kitchen appliances at the current evaluation time point. The historical state feature sequence consists of multiple feature vectors arranged in chronological order.

[0032] In an optional embodiment, a multi-dimensional sensor data stream of a specified smart kitchen appliance within a certain monitoring period can be obtained through the data interface of an IoT cloud platform. This data is continuous time-series data; for example, for a smart oven, it may include the core cavity temperature, heating element operating current, and cooling fan vibration signals. Before the data is sent into the processing flow, necessary preprocessing can be performed, such as filling missing points in the data stream with the mean or nearest neighbor values, and using a moving average or low-pass filter to remove high-frequency noise.

[0033] In this optional embodiment, to transform a continuous data stream into a discrete feature sequence, a fixed-length time window and a sliding step size can be set. The length of the time window determines the temporal granularity of the analysis state, while the step size determines the update frequency of the feature sequence. For example, the time window is set to 10 minutes, and the sliding step size is 1 minute.

[0034] Furthermore, for each data segment within a time window, a series of statistical features can be calculated to comprehensively summarize the equipment's operating status during that time period. These features can be categorized into time-domain features, including the mean, variance, peak-to-peak value, and root mean square of the calculated current, the standard deviation of the calculated temperature, and the kurtosis and skewness of the calculated vibration, which reflect the amplitude, dispersion, and waveform morphology of the signal; frequency-domain features, including the extraction of the dominant frequency, total spectral energy, and power spectral entropy from the vibration signal using a fast Fourier transform, which reveal whether the equipment experiences abnormal periodic vibrations; and, for the oven's heating process, the time taken to heat from room temperature to the preset temperature each time can be calculated, or the slope parameter of the fitted heating curve can be used as a physical model feature to reflect the performance of the equipment's core functions.

[0035] In this optional embodiment, all statistical features calculated for each time window are combined into a multidimensional feature vector. For example, in the window at time t, the calculated mean current is 5.1A, the standard deviation of temperature is 0.5℃, and the dominant vibration frequency is 50Hz. Then the feature vector at that time is... Part of it is [5.1, 0.5, 50, ...].

[0036] In this optional embodiment, by continuously sliding the time window, a series of feature vectors arranged in chronological order can be generated throughout the entire monitoring period, forming a historical state feature sequence.

[0037] like Figure 2 The diagram shown is a comparison of the changes in raw sensor data between healthy and faulty devices throughout their entire lifecycle, according to an embodiment of the present invention. The upper subplot is a comparison of current data, and the lower subplot is a comparison of vibration data. It can be seen that the data of the faulty device (orange line) does indeed exhibit abnormalities such as increased fluctuations or spikes after the fault initiation point (t=900). However, these abnormal signals are mixed with normal noise and disturbances. If only a simple threshold method is used for judgment, false alarms or missed alarms are easily generated. Therefore, it is necessary to perform deep feature extraction on these raw time-series data.

[0038] Thus, by windowing and feature extraction of the original multidimensional time series data, unstructured sensor data streams can be transformed into structured feature sequences with higher information density, laying a solid data foundation for subsequent in-depth time series analysis.

[0039] S2. Perform dual-path parallel time series feature analysis on the historical state feature sequence. The dual-path parallel time series feature analysis includes: calculating the time series context vector and time volatility index based on the time dynamics of the historical state feature sequence; and calculating the cumulative severity score based on the state severity of the historical state feature sequence.

[0040] In an optional embodiment, a dual-path parallel temporal feature analysis can be performed on the historical state feature sequence, including calculating the temporal context vector and time volatility index based on the temporal dynamics of the historical state feature sequence; and calculating the cumulative severity score based on the state severity of the historical state feature sequence.

[0041] Specifically, the stability and fluctuation patterns of the equipment during recent operation are analyzed from a temporal dynamics perspective. To capture long-term temporal dependencies, rather than relying solely on features at the current moment, a representative vector that serves as a summary of the entire historical state feature sequence can be constructed as a temporal context vector. The temporal context vector satisfies the following relationship:

[0042]

[0043] in, For time-series context vectors, This is the feature vector at time t in the historical state feature sequence. Here, k represents the current evaluation time point, and k is the sequence length. The significance weight at time t combines the time decay effect with the drasticness of local state changes.

[0044] Satisfying the relation:

[0045]

[0046] in This is the attenuation coefficient, exemplified by a value of 0.1. This is the mutation enhancement coefficient, which, for example, has a value of 1; The drastic change in the local state at time t After normalizing to the [0,1] interval, the drastic change in local state is obtained. That is, the present invention takes the square of the Euclidean distance between the feature vector at time t and the feature vector at time t-1 as the drasticness of the local state change.

[0047] Specifically, significance weight It consists of two parts. This indicates that the further away a moment is from the current time, the less important it is. This means that if a drastic change occurs at a certain historical moment, for example, such as a sudden surge in current, then the weight of that moment will be amplified.

[0048] Next, the time volatility index can be calculated to assess the overall instability of the equipment state. The time volatility index satisfies the following relationship:

[0049]

[0050] in, The time volatility index is k, where k is the sequence length. Let be the degree of drastic change in the local state at time t. For all The average value, This is the current assessment point in time.

[0051] Furthermore, to assess the cumulative degree to which the equipment's condition deviates from the health baseline during recent operation, a health status feature vector can be predefined. ,in This can be determined by analyzing operational data from a large amount of new equipment, for example, .

[0052] Next, a cumulative severity score can be calculated to reflect cumulative damage, representing the average degree to which the equipment condition deviates from the health baseline. The cumulative severity score satisfies the following relationship:

[0053]

[0054] in, To accumulate the severity score, k is the sequence length. Let be the eigenvector at time t. The single-point severity at time t represents... and The weighted distance can be calculated using the weighted Euclidean distance:

[0055]

[0056] in, The importance weight of the j-th feature can be predetermined using feature selection algorithms such as random forests, or it can be specified by a domain expert. Let j be the feature vector of the j-th feature at time t. Let be the health status feature vector corresponding to the j-th feature.

[0057] like Figure 3 The diagram shown is a schematic representation of the evolution of key temporal characteristics of a faulty device according to an embodiment of the present invention. It can be seen that the cumulative severity score in the upper subgraph exhibits a continuous and stable monotonically increasing trend after the fault occurs, indicating the continuous accumulation of equipment damage; the time volatility index in the lower subgraph, in the form of pulse-like spikes, captures each impact event as the equipment's operating state fluctuates from stable to violently.

[0058] Thus, through parallel analysis of dual paths, deep temporal features concerning time dynamics and state severity can be extracted from historical sequences, providing key inputs for building a more comprehensive risk assessment model.

[0059] S3. Based on the time volatility index and the cumulative severity score, construct a dynamic risk signature, and combine it with the feature vector of the current assessment time point and the time series context vector to form an enhanced feature vector.

[0060] In an optional embodiment, to reflect the interaction effect of state volatility and severity, the following can be used: and Nonlinear fusion is performed to construct a dynamic risk signature, which satisfies the following relation:

[0061]

[0062] in, For dynamic risk signature, To accumulate severity scores, This is the time volatility index.

[0063] Next, the feature vector at the current evaluation time point and the temporal context vector can be combined to form an enhanced feature vector, which satisfies the following relationship:

[0064]

[0065] in, To enhance the feature vector, This is the feature vector at the current evaluation time point. For time-series context vectors, Dynamic risk signatures can comprehensively summarize the current status, historical evolution, and overall risk level of the equipment.

[0066] Thus, by integrating the results of dual-path analysis with the current state information, an enhanced feature vector with rich information dimensions and a profound reflection of the device's runtime sequence characteristics can be constructed, providing a strong foundation for subsequent accurate risk modeling.

[0067] S4. Input the enhanced feature vector into the preset Cox proportional risk model to calculate the real-time failure risk of kitchen appliances, so as to realize the monitoring of failure risk of kitchen appliances.

[0068] In an alternative embodiment, a historical dataset containing a large amount of device lifecycle data can be used to train the Cox proportional hazards model, and the resulting enhanced feature vectors can be used. Replace the original covariates in the Cox proportional hazards model The resulting Cox proportional hazards model is as follows:

[0069]

[0070] in, It is the instantaneous failure risk rate. It is the benchmark risk function. These are the regression coefficient vectors that need to be learned from the data, and can be estimated by maximizing the partial likelihood function.

[0071] Furthermore, once the model is trained, it can be deployed online for real-time risk assessment. The system will continuously calculate the latest enhanced feature vectors and substitute them into the trained model to calculate a real-time risk score, which reflects the failure risk multiple of the current device relative to the baseline state.

[0072] In this optional embodiment, when the calculated real-time risk score is greater than a preset warning threshold, an alarm can be issued in a timely manner, thereby achieving accurate and timely predictive maintenance.

[0073] like Figure 4The diagram shows the comparison effect of dynamic risk signatures provided by the embodiment of the present invention. It can be seen that the risk signature of healthy equipment (blue line) always operates stably at a level below the threshold, verifying the stability and low false alarm rate of the present invention. However, the risk signature of faulty equipment (red line) shows a sharp exponential increase after the fault initiation point (t=900) due to the synergistic amplification effect of accumulated severity and time volatility, and crosses the preset warning threshold line (purple dashed line). Moreover, the warning time point (approximately t=980) is much earlier than the end of the equipment life cycle (t=1200), thereby achieving early, accurate and reliable prediction of equipment failure.

[0074] Thus, by inputting enhanced feature vectors rich in time-series information into the Cox model, it is possible to achieve dynamic and accurate assessment of equipment failure risks and establish an effective early warning mechanism, thereby enabling predictive maintenance.

[0075] This invention also discloses a smart kitchen appliance data analysis system based on an Internet of Things (IoT) cloud platform, including a processor and a memory. The memory stores computer program instructions, and when the computer program instructions are executed by the processor, a smart kitchen appliance data analysis method based on an IoT cloud platform according to the present invention is implemented.

[0076] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.

[0077] In the description of this specification, "multiple" or "several" means at least two, such as two, three or more, unless otherwise expressly and specifically defined.

[0078] While this specification has shown and described numerous embodiments of the invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Many modifications, alterations, and alternatives will occur to those skilled in the art without departing from the spirit and essence of the invention. It should be understood that various alternatives to the embodiments of the invention described herein may be employed in the practice of this invention.

Claims

1. A data analysis method for intelligent kitchen appliances based on an Internet of Things (IoT) cloud platform, characterized in that, include: Obtain the historical state feature sequence of kitchen appliances at the current evaluation time point, wherein the historical state feature sequence consists of multiple feature vectors arranged in chronological order; A dual-path parallel time-series feature analysis is performed on the historical state feature sequence. This analysis includes: calculating a time-series context vector and a time volatility index based on the temporal dynamics of the historical state feature sequence; and calculating a cumulative severity score based on the state severity of the historical state feature sequence. The time-series context vector is calculated as follows: for each feature vector in the historical state feature sequence, a significance weight is calculated based on its time decay effect and the intensity of local state changes; a weighted average is performed on all feature vectors in the historical state feature sequence to obtain the time-series context vector. The time volatility index is calculated as follows: the intensity of local state changes between adjacent feature vectors in the historical state feature sequence is calculated to form an intensity change sequence; the standard deviation of the intensity change sequence is calculated to obtain the time volatility index. The cumulative severity score is calculated as follows: for each feature vector in the historical state feature sequence, a weighted distance between it and a predefined healthy state feature vector is calculated as a single-point severity; the single-point severity scores at all time points in the historical state feature sequence are averaged to obtain the cumulative severity score. Based on the time volatility index and the cumulative severity score, a dynamic risk signature is constructed, and combined with the feature vector of the current assessment time point and the time series context vector, an enhanced feature vector is formed. The enhanced feature vector is input into a preset Cox proportional risk model to calculate the real-time failure risk of kitchen appliances, so as to realize the monitoring of failure risk of kitchen appliances.

2. The data analysis method for intelligent kitchen appliances based on an Internet of Things cloud platform according to claim 1, characterized in that, The acquisition of the historical state feature sequence of the kitchen appliances at the current evaluation time point includes: The system acquires multi-dimensional sensor data streams from kitchen appliances within a preset monitoring period through an IoT cloud platform. A fixed-length time window and sliding step size are set to perform sliding sampling on the multidimensional sensor data stream; For the data within each time window, calculate the statistical characteristics and generate a feature vector; All feature vectors generated during the monitoring period are arranged in chronological order to form the historical state feature sequence.

3. The data analysis method for intelligent kitchen appliances based on an Internet of Things cloud platform according to claim 2, characterized in that, The statistical features include at least one of the following: time-domain features, frequency-domain features, and physical model features.

4. The data analysis method for intelligent kitchen appliances based on an Internet of Things cloud platform according to claim 1, characterized in that, The dynamic risk signature is calculated as follows: the cumulative severity score is used as the basic risk item, and the time volatility index is used as the exponential amplification factor. The dynamic risk signature is obtained by multiplying the basic risk item by the natural exponential value of the exponential amplification factor.

5. The data analysis method for intelligent kitchen appliances based on an Internet of Things cloud platform according to claim 1, characterized in that, The enhanced feature vector is formed by concatenating and combining the feature vector at the current evaluation time point, the temporal context vector, and the dynamic risk signature.

6. A data analysis system for intelligent kitchen appliances based on an Internet of Things (IoT) cloud platform, characterized in that, include: A processor and a memory, wherein the memory stores computer program instructions, which, when executed by the processor, implement a data analysis method for intelligent kitchen appliances based on an Internet of Things cloud platform according to any one of claims 1-5.

Images