Campus self-service laundry equipment abnormity identification system based on Internet of Things

Abstract

The invention relates to the technical field of Internet of Things, and discloses an Internet of Things-based campus self-service laundry equipment abnormity identification system, which comprises a user operation behavior alignment module, a dynamic time warping algorithm quantification behavior deviation degree and a first weight factor generation module, the multi-source state feature extraction module is used for extracting an equipment state matrix to generate a second weight factor in combination with vibration spectrum and motor current waveform analysis; the health degree fusion decision module dynamically distributes weights based on an entropy method, and calculates a health degree score to trigger equipment-level anomaly detection; the abnormal propagation topology modeling module constructs a fusion topological graph, dynamically distributes propagation weights in combination with a historical fault rate and a real-time load, and generates an abnormal propagation path probability graph; and the space-time early warning visualization module integrates the equipment position and the abnormal probability, multiple alarm levels are divided, and the risk distribution is dynamically displayed through the thermodynamic diagram layer, so that the detection precision and efficiency of abnormal identification of the campus self-service laundry equipment are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of the Internet of Things, and particularly to an abnormal recognition system for campus self-service laundry equipment based on the Internet of Things. Background Art

[0002] The following problems generally exist in the operation of campus self-service laundry equipment: The traditional system mainly relies on the monitoring of the state of a single device (such as temperature, current), and ignores the impact of user operation behaviors on the health of the device (such as frequently switching modes, using during unconventional periods, etc.), resulting in incomplete abnormal detection. At the same time, there is no effective modeling of the abnormal propagation path of the device, and it is impossible to predict and block cascading failures. For example, when a certain device stops due to mechanical failure, devices physically connected to it (such as sharing a water supply pipeline) or data communicating with it (such as sharing a network node) may fail successively due to the chain reaction, but the prior art cannot analyze such propagation paths, resulting in a lag in maintenance response. In addition, the prior art does not integrate multi-source data for comprehensive analysis, the accuracy of abnormal recognition is insufficient, and there is no intuitive visual warning mechanism, resulting in low operation and maintenance efficiency, prolonged user waiting time, and affecting the use experience. Therefore, there is an urgent need for an intelligent abnormal recognition system to improve the efficiency and accuracy of abnormal recognition of campus self-service laundry equipment. Summary of the Invention

[0003] The present invention provides an abnormal recognition system for campus self-service laundry equipment based on the Internet of Things, and its main purpose is to solve the problem of low efficiency in abnormal recognition of campus self-service laundry equipment.

[0004] To achieve the above purpose, an abnormal recognition system for campus self-service laundry equipment based on the Internet of Things provided by the present invention includes: A user operation behavior alignment module, which is used to generate a user operation time series log of the laundry equipment, align the user operation time series log with a preset standard operation template through a dynamic time warping algorithm, and calculate the operation behavior deviation degree to generate a first weight factor representing the probability of user behavior abnormality; A multi-source state feature extraction module, which is used to construct a device state matrix of the laundry equipment based on real-time obtained multi-source operation data, and extract feature parameters exceeding a preset safety threshold in the device state matrix to generate a second weight factor representing the abnormality of the physical state of the device; A health degree fusion decision module, which is used to perform weighted decision on the first weight factor and the second weight factor to obtain the health degree score of the laundry equipment, and trigger device-level abnormal detection when the health degree score is lower than a first threshold to obtain the device-level abnormal type of the laundry equipment; Anomaly propagation topology modeling module, which is used to construct a fusion topology graph of the physical connection topology and the data communication topology of the laundry equipment, dynamically allocate the anomaly propagation weights of each node in the fusion topology graph based on the historical failure rate and real-time load status of the equipment, and generate an anomaly propagation path probability graph according to the connection strength between the nodes in the fusion topology graph and the anomaly propagation weights; Space-time warning visualization module, which is used to calculate the group anomaly probability of the laundry equipment based on the equipment-level anomaly type and the anomaly propagation path probability graph, divide multiple warning levels according to the probability interval, and generate a visual warning log by fusing the location information of the laundry equipment and the multiple warning levels.

[0005] Optionally, when the user operation behavior alignment module executes to generate the user operation timing log of the laundry equipment, it includes: Collect user operation behavior data, where the user operation behavior data includes: reservation period, washing mode selection, and payment record; Generate a user operation timing log containing the reservation period, washing mode selection sequence, and payment success rate based on the user operation behavior data.

[0006] Optionally, when the user operation behavior alignment module executes to align the user operation timing log with the preset standard operation template by the dynamic time warping algorithm and calculate the operation behavior deviation as the first weight factor of the equipment health, it includes: Extract the reservation period distribution density, mode switching frequency, and payment interruption times in the user operation timing log as key behavior indicators; Calculate the cumulative path distance between each key behavior indicator and the preset standard operation template by the dynamic time warping algorithm; Perform weighted summation on the cumulative path distance to obtain the operation behavior deviation, and use the operation behavior deviation as the first weight factor of the equipment health.

[0007] Optionally, the generation method of the preset standard operation template in the user operation behavior alignment module includes: Collect the user operation data of N laundry equipment within the normal operation and maintenance cycle to form a training set; Perform Gaussian kernel density estimation on the reservation period data in the training set to generate a period distribution probability curve; Extract the optimal path of the washing mode switching as the standard mode sequence; Statistically analyze the normal distribution interval of the payment success rate as the benchmark payment behavior indicator; Generate a preset standard operation template based on the period distribution probability curve, standard mode sequence, and benchmark payment behavior indicator.

[0008] Optionally, when constructing the device state matrix of the laundry device based on the multi-source operation data obtained in real time, the multi-source state feature extraction module includes: Obtain the multi-source operation data of the laundry device in real time, where the multi-source operation data includes: vibration spectrum data and motor current waveform; Perform wavelet packet decomposition on the vibration spectrum data, and extract the energy ratio of the preset frequency band as the mechanical wear feature; After performing fast Fourier transform on the motor current waveform, calculate the amplitude ratio of the fundamental wave to the third harmonic as the electrical anomaly feature; Align the mechanical wear feature and the electrical anomaly feature according to the time stamp to form a two-dimensional state matrix, and determine the two-dimensional state matrix as the device state matrix of the laundry device.

[0009] Optionally, when performing weighted decision on the first weight factor and the second weight factor to obtain the health score of the laundry device, the health degree fusion decision module includes: Dynamically calculate the weight distribution ratio of the first weight factor and the second weight factor based on the entropy method; Calculate the health score of the laundry device by using the weighted summation algorithm and the weight distribution ratio, where the weighted summation algorithm is:

[0010] Where, is the health score of the laundry device, is the weight distribution ratio corresponding to the first weight factor, is the weight distribution ratio corresponding to the second weight factor, is the first weight factor, is the second weight factor.

[0011] Optionally, when dynamically allocating the abnormal propagation weights of the nodes in the fusion topology graph based on the device historical failure rate and the real-time load state, the abnormal propagation topology modeling module includes: Determine the initial static weight of each node according to the betweenness centrality of the laundry device in the fusion topology graph and the device historical failure rate; When the real-time load state of the laundry device exceeds 70% of the rated value, determine the dynamic correction coefficient of the initial static weight, and correct the node weight in the fusion topology graph based on the dynamic correction coefficient; When the laundry device is in the maintenance lock state, force the node weight of the fusion topology graph to zero to block the abnormal propagation.

[0012] Optionally, when the spatio-temporal warning visualization module calculates the population anomaly probability of the laundry equipment based on the equipment-level anomaly type and the anomaly propagation path probability graph, it includes: Input the equipment-level anomaly type and the anomaly propagation path probability graph into the spatio-temporal propagation model to calculate the population anomaly probability of the laundry equipment, where the spatio-temporal propagation model satisfies: Adopt an exponential decay propagation rule based on physical distance for mechanical anomalies; Adopt a linear propagation rule based on network hop count for data anomalies; When the two types of anomalies coexist, give priority to performing mechanical anomaly propagation calculation and superimpose the data anomaly correction term.

[0013] Optionally, the calculation formula for the population anomaly probability in the spatio-temporal warning visualization module is as follows:

[0014] Wherein, is the population anomaly probability of the laundry equipment, is the number of equipment that may malfunction due to the propagation of the initial anomaly equipment , is the number of all laundry equipment in a specific area, is the equipment number of the affected laundry equipment, is the laundry equipment to which the anomaly propagates the final probability, is the equipment number of the laundry equipment in a specific area, is the laundry equipment the probability of independent anomaly occurrence (independent of propagation), is the laundry equipment, is the equipment number of the initial anomaly equipment.

[0015] Optionally, when the spatio-temporal warning visualization module executes the generation of a visual warning log by fusing the location information of the laundry equipment and the multi-level warning levels, it includes: According to the location information of the laundry equipment and the multi-level warning levels, render a three-level heat layer of the abnormal equipment in the laundry equipment on the campus electronic map, where red, yellow, and blue respectively correspond to high risk, medium risk, and low risk; Mark the possible influence range and the estimated arrival time of the abnormal propagation of the abnormal equipment; Automatically generate a maintenance work order containing a spare part replacement list and an operation guide video based on a pre-associated maintenance knowledge base.

[0016] The present invention aligns the user operation timing with the standard template through the dynamic time warping algorithm, quantifies the behavior deviation degree, accurately identifies misoperations or malicious behaviors, avoids equipment failures caused by abnormal user behaviors, and at the same time combines vibration spectrum and motor current waveform analysis to capture mechanical wear and electrical abnormalities in real time, realizes multi-dimensional physical state monitoring, enhances the ability of fault prediction, dynamically allocates the weights of user behaviors and equipment states through the entropy method, generates a comprehensive health score, helps quickly locate high-risk equipment and prioritize the handling of key faults, reduces downtime; predicts the abnormal propagation path based on the physical connection and data communication topology, dynamically adjusts the node weights, blocks the influence range of high-risk equipment, avoids cascading failures, and reduces the operation and maintenance costs; in addition, dynamically displays the abnormal distribution and risk level of campus equipment through a heat map, combines location information to mark the abnormal propagation range and the expected impact time, and assists managers in quickly formulating emergency strategies, so as to comprehensively improve the efficiency of abnormal identification of campus self-service laundry equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 FIG. is a system architecture diagram of an abnormal identification system for campus self-service laundry equipment based on the Internet of Things provided by an embodiment of the present invention.

[0018] The implementation, functional features and advantages of the present invention will be further described with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0019] To make the objectives, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0020] The terms used in the embodiments of the present invention are only for the purpose of describing specific embodiments, and are not intended to limit the present invention. The singular forms of "the" and "said" used in the embodiments of the present invention and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. "Plurality" generally includes at least two.

[0021] Depending on the context, the words "if" or "when" as used herein may be interpreted as "when" or "when...", or "in response to determining" or "in response to detecting". Similarly, depending on the context, the phrases "if determined" or "if detecting (stated condition or event)" may be interpreted as "when determined" or "in response to determining", or "when detecting (stated condition or event)" or "in response to detecting (stated condition or event)".

[0022] In addition, the step sequence in the following method embodiments is only an example and not a strict limitation.

[0023] In practice, the server-side device deployed in an IoT-based campus self-service laundry equipment anomaly identification system may consist of one or more devices. The aforementioned IoT-based campus self-service laundry equipment anomaly identification system can be implemented as a service instance, a virtual machine, or hardware devices. For example, an IoT-based campus self-service laundry equipment anomaly identification system can be implemented as a service instance deployed on one or more devices in a cloud node. Simply put, an IoT-based campus self-service laundry equipment anomaly identification system can be understood as software deployed on a cloud node, providing a IoT-based campus self-service laundry equipment anomaly identification system to each user terminal. Alternatively, an IoT-based campus self-service laundry equipment anomaly identification system can be implemented as a virtual machine deployed on one or more devices in a cloud node. This virtual machine contains application software for managing each user terminal. Alternatively, an IoT-based campus self-service laundry equipment anomaly identification system can be implemented as a server-side device composed of multiple hardware devices of the same or different types, with one or more hardware devices configured to provide a IoT-based campus self-service laundry equipment anomaly identification system to each user terminal.

[0024] In terms of implementation, the IoT-based campus self-service laundry equipment anomaly identification system and the user end are mutually compatible. Specifically, the IoT-based campus self-service laundry equipment anomaly identification system is an application installed on a cloud service platform, and the user end is a client that establishes a communication connection with the application. Alternatively, the IoT-based campus self-service laundry equipment anomaly identification system is implemented as a website, and the user end is implemented as a webpage. Alternatively, the IoT-based campus self-service laundry equipment anomaly identification system is implemented as a cloud service platform, and the user end is implemented as a mini-program within an instant messaging application.

[0025] like Figure 1 FIG. 1 is a system architecture diagram of an abnormality identification system for campus self-service laundry equipment based on the Internet of Things provided by one embodiment of the present invention.

[0026] The abnormal recognition system 100 for campus self-service laundry equipment based on the Internet of Things according to the present invention can be set in a cloud server. In terms of implementation form, it can be used as one or more service devices, or can be used as an application installed on the cloud (such as the server of a mobile service operator, a server cluster, etc.), or can also be developed into a website. According to the functions achieved, the abnormal recognition system 100 for campus self-service laundry equipment based on the Internet of Things can include a user operation behavior alignment module 101, a multi-source state feature extraction module 102, a health degree fusion decision module 103, an abnormal propagation topology modeling module 104, and a spatio-temporal early warning visualization module 105. The modules in the present invention can also be referred to as units, which refer to a series of computer program segments that can be executed by an electronic device processor and can complete fixed functions, and are stored in the memory of the electronic device.

[0027] In an embodiment of the present invention, in an abnormal recognition system for campus self-service laundry equipment based on the Internet of Things, each of the above modules can be independently implemented and called by other modules. Here, the call can be understood as that a certain module can be connected to multiple modules of another type and provide corresponding services for the multiple modules it is connected to. In the abnormal recognition system for campus self-service laundry equipment based on the Internet of Things provided by the embodiment of the present invention, without modifying the program code, the applicable range of the architecture of the abnormal recognition system for campus self-service laundry equipment based on the Internet of Things can be adjusted by adding modules and directly calling, so as to achieve cluster-level horizontal expansion, so as to achieve the purpose of quickly and flexibly expanding the abnormal recognition system for campus self-service laundry equipment based on the Internet of Things. In actual application, the above modules can be set in the same device or different devices, or can also be set in virtual devices, such as service instances in a cloud server.

[0028] The following will be described in detail for each component and specific working process of an abnormal recognition system for campus self-service laundry equipment based on the Internet of Things in combination with specific embodiments: The user operation behavior alignment module 101 is used to generate a user operation time sequence log of the laundry equipment, align the user operation time sequence log with a preset standard operation template through a dynamic time warping algorithm, and calculate the operation behavior deviation degree to generate a first weight factor representing the probability of user behavior abnormality.

[0029] In an embodiment of the present invention, the generation of the user operation time sequence log of the laundry equipment includes: Collecting user operation behavior data, where the user operation behavior data includes: reservation period, washing mode selection, and payment record; Generating a user operation time sequence log including a reservation period, a washing mode selection sequence, and a payment success rate based on the user operation behavior data.

[0030] Specifically, the user operation timing log is the behavioral data of the user using the laundry equipment recorded in chronological order, including information such as the reservation time, the selected washing mode, and the payment status, in order to capture the user behavior pattern and provide a data basis for subsequent anomaly detection.

[0031] Specifically, the traditional system only records the equipment status (such as temperature, current), and ignores the impact of user operations on the equipment health (such as frequent reservation + high water level mode may accelerate equipment wear).

[0032] Specifically, the reservation period refers to the specific time when the user reserves the washing machine (such as 2023-10-05 18:30); the washing mode selection refers to the mode selected by the user (such as "quick wash for 15 minutes", "high temperature sterilization"); the payment record refers to whether the payment is successful and the payment method (such as the number of failed WeChat payments).

[0033] Specifically, the data is sorted by timestamp to generate a timing log. For example, a user reserves the washing machine at 2 am for three consecutive days, selects the "large item washing" mode, but fails to pay 3 times.

[0034] Specifically, the dynamic time warping algorithm is an algorithm used to align time series of different lengths or speeds, and achieves the best match by calculating the minimum cumulative distance, aiming to eliminate the differences in user operation rhythms (such as some people operate fast and some people operate slow), and unify the comparison with the standard template.

[0035] Specifically, directly comparing the original time series may lead to misjudgment due to time offset (such as user A reserves at 10:00 and user B reserves at 10:05, and their actual behavior patterns are the same). Specifically, define the standard operation template: Collect historical data of N devices during the normal operation and maintenance period (such as 1000 records); The reservation period can generate a period distribution curve through Gaussian kernel density estimation (such as the peak period is from 18:00 to 20:00); The mode sequence refers to extracting the common mode switching paths (such as "quick wash → dehydration" is the optimal path); The payment success rate needs to be statistically analyzed in the normal distribution interval (such as the normal payment success rate > 95%).

[0036] In the embodiment of the present invention, when aligning the user operation timing log with the preset standard operation template through the dynamic time warping algorithm and calculating the operation behavior deviation degree as the first weight factor of the equipment health degree, it includes: Extract the reservation period distribution density, mode switching frequency, and payment interruption times in the user operation timing log as key behavior indicators; Calculate the cumulative path distance between each of the key behavior metrics and a preset standard operation template through the dynamic time warping algorithm; Perform a weighted sum on the cumulative path distance to obtain an operation behavior deviation degree, and use the operation behavior deviation degree as the first weight factor of the device health degree.

[0037] For example, after aligning the operation sequence of user A with the template, the cumulative distance is 5 (low deviation); user B operates in the early morning, and the cumulative distance is 20 (high deviation).

[0038] Specifically, the deviation degree is a quantitative value of the difference between the user operation sequence and the standard template, used to characterize the abnormal probability. The purpose is to convert the time series alignment result into a comparable numerical index (the first weight factor F1), because directly using the DTW distance cannot reflect the comprehensive abnormality in multiple dimensions (time period, mode, payment).

[0039] Specifically, the appointment time period distribution density refers to the frequency of users making appointments during off-peak hours (such as in the early morning); the mode switching frequency refers to the frequency of frequently switching modes within a short period of time (such as switching 3 times within 5 minutes); the payment interruption times refer to the proportion of the number of payment failures or cancellations.

[0040] Specifically, set weights (such as the time period distribution accounts for 50%, the mode switching accounts for 30%, and the payment interruption accounts for 20%), and calculate the operation behavior deviation degree according to the following formula:

[0041] For example, the deviation degree calculation of user C: Time period deviation: 0.8 (appointment in the early morning) Mode deviation: 0.6 (frequent switching) Payment deviation: 0.9 (3 failures) Then the comprehensive deviation degree is

[0042] Specifically, the first weight factor F1 is a quantitative value of the user behavior abnormal probability, used for subsequent health degree scoring, mapping the deviation degree to the health degree scoring system, and reflecting the influence weight of the user behavior on the device health.

[0043] Specifically, user misoperations (such as incorrect mode selection) may be related to device failures (such as drainage blockage), and the weights need to be dynamically adjusted.

[0044] Furthermore, set the threshold rules as follows: Low risk: deviation degree < 0.3 → F1 = 0.1 Medium risk: 0.3 ≤ deviation degree < 0.7 → F1 = 0.3 High risk: deviation degree ≥ 0.7 → F1 = 0.6.

[0045] For example, the deviation of user C is 0.76, which belongs to high risk, and F1 = 0.6. The device used by the user will be preferentially checked for abnormalities.

[0046] In the embodiment of the present invention, the method for generating the standard operation template includes: Collecting user operation data of N laundry devices within a normal operation and maintenance cycle to form a training set; Performing Gaussian kernel density estimation on the reservation time period data in the training set to generate a time period distribution probability curve; Extracting the optimal path of washing mode switching as the standard mode sequence; Statistically analyzing the normal distribution interval of the payment success rate as the benchmark payment behavior indicator; Generating a preset standard operation template based on the time period distribution probability curve, the standard mode sequence, and the benchmark payment behavior indicator.

[0047] Specifically, collect user operation data of multiple laundry devices during normal operation and maintenance (such as data of 1000 devices for 3 consecutive months), and establish a normal behavior benchmark for subsequent comparison in anomaly detection.

[0048] Specifically, the data fields are the reservation timestamp, the washing mode sequence, and the payment status (success / failure).

[0049] For example, within the normal cycle of a certain device, the user's reservation time period is concentrated between 18:00 and 20:00, and the common mode is "standard wash → dehydration".

[0050] Specifically, Gaussian kernel density estimation to generate a time period distribution curve is a non-parametric statistical method used to estimate the probability density distribution of the user's reservation time period, which can identify the distribution law of normal time periods (such as peak periods) and detect abnormal time periods (such as high-frequency reservations in the early morning).

[0051] Specifically, generate a time period probability density curve (with a peak between 18:00 and 20:00) based on all reservation timestamps (such as 10,000 records).

[0052] Specifically, the calculation formula of Gaussian kernel density estimation is:

[0053] Among them, is the probability density value estimated at point , is the number of samples, that is, the total number of observed data used for estimation (such as 1000 user reservation time records), is the parameter that controls the smoothing degree of the kernel function, is the summation index, traversing all sample points (from i = 1 to i = N), is the Gaussian kernel function, is the target point (such as a specific time point) where the probability density needs to be estimated, is the th observed data point (such as the reservation time of the th user).

[0054] Specifically, each data point contributes a Gaussian distribution to the density estimate at , and the final density value is the weighted average of all contributions.

[0055] Specifically, analyze the most common path of user mode switching (such as "quick wash → rinse → spin"), because abnormal switching (such as frequent "quick wash → pause → quick wash") may indicate misoperation or equipment failure.

[0056] For example, in the normal mode, the probability that a user switches from "standard wash" to "spin" is 85%, while abnormal users frequently switch between "quick wash" and "pause".

[0057] Specifically, the normal distribution interval of the payment success rate is to determine the range of the normal payment success rate (such as a mean of 95% and a standard deviation of 2%), because an abnormal increase in the payment failure rate may reflect system failure or malicious attack.

[0058] Specifically, calculate the mean and standard deviation of the payment success rate, and set the confidence interval (such as the 95% confidence interval is 91% - 99%) to obtain the normal distribution interval of the payment success rate.

[0059] Furthermore, integrate the time period distribution curve, mode path, and payment interval to form a multi-dimensional benchmark.

[0060] The multi-source state feature extraction module 102 is used to construct the device state matrix of the laundry device based on the multi-source operation data obtained in real time, and extract the characteristic parameters in the device state matrix that exceed the preset safety threshold to generate a second weight factor representing the abnormal physical state of the device.

[0061] In the embodiment of the present invention, when constructing the device state matrix of the laundry device based on the multi-source operation data obtained in real time, it includes: Obtain the multi-source operation data of the laundry device in real time, where the multi-source operation data includes: vibration spectrum data and motor current waveform; Perform wavelet packet decomposition on the vibration spectrum data, and extract the proportion of energy in the preset frequency band as the mechanical wear feature; After performing fast Fourier transform on the motor current waveform, calculate the amplitude ratio of the fundamental wave to the third harmonic as the electrical anomaly feature; Align the mechanical wear characteristics and electrical anomaly characteristics according to the time stamp to form a two-dimensional state matrix, and determine that the two-dimensional state matrix is the device state matrix of the laundry equipment.

[0062] Specifically, the vibration spectrum refers to the vibration signal of the device collected by the acceleration sensor; the motor current waveform refers to the working current of the motor recorded by the current sensor.

[0063] For example, a vibration spectrum diagram (0 - 10 kHz) and a current waveform (50 Hz fundamental wave) are obtained in a certain sampling.

[0064] Specifically, the mechanical wear characteristics are extracted by wavelet packet decomposition by decomposing the vibration signal into different frequency bands and extracting the energy of a specific frequency band (such as 3 - 5 kHz), because mechanical wear (such as bearing damage) will cause an increase in high-frequency vibration energy.

[0065] Specifically, select the Daubechies wavelet basis for 5-layer decomposition, and calculate the energy ratio of the 4th sub-band of the 3rd layer (3 - 5 kHz). For example, the energy ratio of a normal device is 10%, and it rises to 25% during wear.

[0066] Specifically, the current signal is converted into a spectrum by fast Fourier transform (FFT), and the amplitude ratio of the fundamental wave (50 Hz) to the third harmonic (150 Hz) is calculated, because harmonic distortion (such as an increase in the amplitude of the third harmonic) may cause a short circuit in the motor winding.

[0067] Specifically, the amplitude ratio is , for example, R < 0.1 for a normal device, and R > 0.3 during a fault.

[0068] Specifically, constructing the two-dimensional state matrix is to align the mechanical wear characteristics (E_wear) and electrical anomaly characteristics (R) according to the time stamp to form a matrix, in order to comprehensively consider the mechanical and electrical states and comprehensively evaluate the device health.

[0069] Generally speaking, the combination of the user behavior template (time period, mode, payment) and the device state matrix (vibration, current) covers two dimensions of "human operation" and "physical state".

[0070] Specifically, the preset safety threshold is the safety critical value of the preset mechanical wear and electrical anomaly characteristics. Among them, the mechanical wear threshold is that the energy ratio of the 3 - 5 kHz frequency band > 15% (the average of normal devices is 10%), and the electrical anomaly threshold is that the amplitude ratio of the fundamental wave to the third harmonic > 0.25 (the average of normal devices is 0.1).

[0071] Specifically, quantify the abnormal degree of the device physical state to avoid subjective judgment deviation.

[0072] Further, extract the over-limit characteristic parameters. Among them, for mechanical wear detection, if the energy ratio at 3 - 5 kHz in a certain sampling is 20% (exceeding the threshold of 15%), it is marked as abnormal; for electrical abnormality detection, if the current harmonic ratio is 0.3 (exceeding the threshold of 0.25), it is marked as abnormal.

[0073] Specifically, the rule for generating the second weight factor (F2) is as follows: Single-index over-limit: F2 = 0.4 (medium risk).

[0074] Double-index over-limit: F2 = 0.8 (high risk).

[0075] Continuous over-limit: Accumulate over time (e.g., if over-limit occurs continuously 3 times, F2 = 1.0).

[0076] Specifically, the calculation formula for the second weight factor (F2) is as follows:

[0077] Among them, is the second weight factor, is the total number of over-limit characteristic parameters, is the over-limit characteristic parameter identifier, is the index weight (60% for mechanical and 40% for electrical), is the over-limit indication function (over-limit = 1, otherwise = 0).

[0078] For example, in a certain detection, there is a mechanical over-limit (weight 0.6) and no electrical over-limit (weight 0.4), then the second weight factor is 0.6.

[0079] The health degree fusion decision module 103 is used to perform weighted decision on the first weight factor and the second weight factor to obtain the health degree score of the laundry equipment. When the health degree score is lower than the first threshold, it triggers device-level abnormality detection to obtain the device-level abnormality type of the laundry equipment.

[0080] In the embodiment of the present invention, when performing weighted decision on the first weight factor and the second weight factor to obtain the health degree score of the laundry equipment, it includes: Dynamically calculate the weight distribution ratio of the first weight factor and the second weight factor based on the entropy value method; Calculate the health degree score of the laundry equipment using the weighted summation algorithm and the weight distribution ratio, where the weighted summation algorithm is: <x

[0081] Among them, is the health degree score of the laundry equipment, is the weight allocation ratio corresponding to the first weight factor, is the weight allocation ratio corresponding to the second weight factor, is the first weight factor, is the second weight factor.

[0082] Specifically, the weights are dynamically allocated according to the information volume (entropy value) of the data. The higher the information volume of an index, the higher its weight.

[0083] Specifically, the information entropy has the following calculation formula:

[0084] For example, when the information entropy of the weight factors is as follows:

[0085]

[0086] Specifically, the calculation formula for entropy redundancy is: then the total redundancy:

[0087] Specifically, the weight allocation ratio is:

[0088] Specifically, for a certain device, F1 = 0.6 (high risk of user behavior), F2 = 0.8 (high risk of device status), α = 0.625, β = 0.375, then the health score is 0.675.

[0089] Specifically, assume the threshold rule is: H < 0.3: healthy (green); 0.3 ≤ H < 0.6: warning (yellow); H ≥ 0.6: failure (red).

[0090] In the embodiments of the present invention, when the health score is lower than the first threshold, device - level anomaly detection is triggered, and the device - level anomaly types of the laundry device obtained include: Specifically, device - level anomaly detection refers to independently analyzing the health status of a single laundry device, and judging whether it is in an abnormal operating state through multi - dimensional data fusion (user behavior deviation and device physical state anomaly).

[0091] Specifically, the device - level anomaly type refers to the specific fault category determined according to the combination of over - limit characteristic parameters, including but not limited to: mechanical anomalies, electrical anomalies, and user operation anomalies.

[0092] Further, the mechanical anomalies include: bearing wear and belt looseness. Bearing wear means that the proportion of high-frequency energy in the vibration spectrum exceeds a threshold (e.g., the energy in the 3 - 5 kHz frequency band > 15%); belt looseness means that the amplitude of low-frequency vibration increases abnormally. Further, the electrical anomalies include: motor winding short circuit and power module failure. Among them, a motor winding short circuit means that the current harmonic ratio (third harmonic / fundamental wave) exceeds a threshold (e.g., R > 0.25); a power module failure means that the distortion rate of the current waveform is abnormal. Further, the user operation anomalies refer to malicious mode switching and non-period use. Among them, malicious mode switching means frequent switching of the washing mode within a short time (e.g., switching ≥ 3 times within 5 minutes); non-period use means high-frequency reservation of the device during the early morning period (e.g., 00:00 - 05:00).

[0093] Example: If the health score H = 0.2 (lower than the threshold 0.3), and the motor harmonic ratio R = 0.3 (exceeding the limit) is detected, and at the same time the proportion of high-frequency vibration energy is 20%, it is determined as the composite anomaly type of "motor winding short circuit + bearing wear".

[0094] The abnormal propagation topology modeling module 104 is used to construct a fusion topology diagram of the physical connection topology and the data communication topology of the laundry equipment, dynamically allocate the abnormal propagation weights of each node of the fusion topology diagram based on the historical failure rate and real-time load status of the equipment, and generate an abnormal propagation path probability diagram according to the connection strength between the nodes of the fusion topology diagram and the abnormal propagation weights.

[0095] Specifically, the nodes of the physical connection topology are each laundry equipment, the edges are physical connection relationships (such as shared water supply pipelines, circuits), and the weights are connection strengths (such as water pipe diameter, circuit capacity).

[0096] For example, equipment A and equipment B share the main water supply pipe, and the connection strength is 0.8 (full value 1.0).

[0097] Specifically, the nodes of the data communication topology are laundry equipment and network equipment (such as gateways, servers), the edges are data transmission paths (such as Wi-Fi, Ethernet), and the weights are network bandwidth or delay (such as a bandwidth of 100 Mbps corresponding to a strength of 0.9).

[0098] Specifically, if there are both physical and data connections between two nodes, they are merged into one edge, and the total connection strength is the weighted sum. For example, the physical strength is 0.8, the data strength is 0.9, and the weight is 0.6, then the total connection strength is 0.84.

[0099] Specifically, the propagation probability formula is , where in the formula, is the propagation weight of the target node j, is the node From The connection strength is the weight of node to The propagation probability.

[0100] For example, if device A (weight 0.6) is connected to device B (weight 0.8) with a connection strength of 0.72, the propagation probability is 0.576.

[0101] In the embodiments of the present invention, when dynamically allocating the abnormal propagation weights of the nodes in the fusion topology graph based on the historical failure rate and real-time load status of the device, it includes: Determining the initial static weight of each node according to the betweenness centrality and the historical failure rate of the device in the fusion topology graph; When the real-time load status of the device exceeds 70% of the rated value, determining the dynamic correction coefficient of the initial static weight, and correcting the node weights in the fusion topology graph based on the dynamic correction coefficient; When the device is in the maintenance lock state, forcing the node weights in the fusion topology graph to be zero to block abnormal propagation.

[0102] Specifically, betweenness centrality measures the hubness of a node in a graph, and the calculation formula is as follows:

[0103] Among them, is the betweenness centrality, is the node to node The number of shortest paths, is, is the shortest path passing through the node The number of shortest paths, is the node.

[0104] Specifically, the historical failure rate is the ratio of the number of past failures of the device to the total operating time.

[0105] Specifically, the trigger condition for real-time load correction is that the real-time load of the device > 70% of the rated value.

[0106] For example: The real-time load of device B is 85% and the initial weight is 0.4, then the dynamic correction coefficient is:

[0107] Furthermore, .

[0108] Specifically, if the device is in the maintenance lock state, force its propagation weight to be 0. For example, device C is locked for maintenance, and its weight changes from 0.5 to 0, and abnormalities cannot be propagated through C.

[0109] The space-time warning visualization module 105 is configured to calculate the group anomaly probability of the laundry equipment based on the device-level anomaly type and the anomaly propagation path probability graph, divide multiple warning levels according to the probability interval, and generate a visualization warning log by integrating the location information of the laundry equipment and the multiple warning levels.

[0110] In the embodiment of the present invention, when calculating the group anomaly probability of the laundry equipment based on the device-level anomaly type and the anomaly propagation path probability graph, it includes: Inputting the device-level anomaly type and the anomaly propagation path probability graph into a space-time propagation model to calculate the group anomaly probability of the laundry equipment, where the space-time propagation model satisfies: Adopting an exponential decay propagation rule based on physical distance for mechanical anomalies; Adopting a linear propagation rule based on the number of network hops for data anomalies; When the two types of anomalies coexist, preferentially perform mechanical anomaly propagation calculation and superimpose the data anomaly correction term.

[0111] Specifically, the decay factor used in the exponential decay propagation rule can be a time decay factor or a distance decay factor.

[0112] Specifically, the exponential decay propagation rule based on physical distance includes: a physical distance decay factor. Among them, the physical distance decay factor The calculation formula is as follows: [[ID=2"]]

[0113] Among them, is the initial propagation probability (determined by device weight and connection strength), is the distance decay coefficient (such as 0.1 / m), is the physical distance between devices (unit: meter).

[0114] Specifically, the linear propagation rule for data anomalies includes: network hop count decay, where the network hop count decay The calculation formula is as follows:

[0115] Among them, is the network hop count decay, <000"]] is the initial propagation probability (determined by device weight and connection strength), is the decay rate per hop (such as 0.15 / hop), is the number of network hops (the path length from the source device to the target device).

[0116] Specifically, the priority mechanical anomaly propagation refers to the superposition of a data anomaly correction term based on the mechanical anomaly propagation probability Pmech.

[0117] In the embodiments of the present invention, the calculation formula for the population anomaly probability is as follows:

[0118] Wherein, is the population anomaly probability of the laundry equipment, is the number of equipment that may fail due to the propagation of the initial abnormal equipment ; is the number of all laundry equipment in a specific area, is the equipment number of the affected laundry equipment, is the laundry equipment to which the anomaly propagates to the laundry equipment ; is the equipment number of the laundry equipment in a specific area, is the laundry equipment with the probability of independent anomaly occurrence (independent of propagation), is the laundry equipment, is the equipment number of the initial abnormal equipment.

[0119] Specifically, the calculation formula for the population anomaly probability consists of two main parts: one is the probability of cascading failures, and the other is the probability of concurrent anomalies in the area. Then, the total population anomaly probability is calculated by multiplication and complementation.

[0120] Specifically, is the population anomaly probability of the laundry equipment, indicating the likelihood of anomalies occurring in the entire laundry equipment population. The closer the value is to 1, the higher the anomaly risk; is the number of equipment that may fail due to the propagation of the initial abnormal equipment . If the anomaly of equipment propagates to 3 pieces of equipment, then , is the number of all laundry equipment in a specific area (such as a certain dormitory building), is the equipment number of the affected laundry equipment. When , it represents the second affected equipment.

[0121] Specifically, the cascading failure probability represents the probability that all affected equipment does not fail due to the anomaly of equipment ; the concurrent anomaly probability in the area represents the sum of the probabilities of independent anomalies of the equipment in the area, and if it exceeds 1, it takes 1 (probability upper limit constraint).

[0122] For example, assume there are 2 devices in the system, where the probability of propagation anomaly , independent anomaly probability , then .

[0123] Specifically, represents independent individuals or local anomaly judgment units ( is the final anomaly probability of a single unit) all do not have anomalies; the joint probability represents correction terms is the anomaly probability corresponding to the correction term) all do not have anomalies; the joint probability.

[0124] In the embodiments of the present invention, when fusing the location information of the laundry device and the multi-level alarm level to generate a visual alarm log, it includes: According to the location information of the laundry device and the multi-level alarm level, render a three-level heat layer of abnormal devices in the laundry device on the campus electronic map, where red, yellow, and blue respectively correspond to high risk, medium risk, and low risk; Mark the possible influence range and the estimated arrival time of the abnormal propagation of the abnormal device; Automatically generate a maintenance work order including a spare part replacement list and an operation guide video based on a pre-associated maintenance knowledge base.

[0125] Specifically, based on the location information of the laundry device, combined with the multi-level alarm level (high, medium, and low risks), render a heat layer of abnormal devices on the campus electronic map in color differentiation, where red represents high risk, yellow represents medium risk, and blue represents low risk; through the visual heat map, intuitively display the distribution density and risk level of abnormal laundry devices, facilitating quick positioning of high-risk areas and assisting management decisions.

[0126] Specifically, mark the possible influence range of the abnormal propagation of the abnormal device on the electronic map and predict the estimated arrival time; early warning of the abnormal diffusion trend, helping managers plan resource allocation, such as guiding students to avoid the affected area in advance, or making emergency preparations, and the device location information is GPS coordinates or building number + room number.

[0127] Specifically, use a GIS platform (such as ArcGIS, Mapbox) or an open-source library (such as Leaflet.js) to render the map; integrate a real-time data interface (such as MQTT or WebSocket) to dynamically update the device status.

[0128] Specifically, generate a gradient color coverage area according to the device density and the alarm level.

[0129] Example: In a dormitory building, equipment is concentrated and there are multiple high risks, which are shown as dark red blocks; dispersed low-risk equipment is shown as light blue dots.

[0130] Specifically, by associating with a pre-established maintenance knowledge base, maintenance work orders including spare part replacement lists and operation guidance videos are automatically generated; the maintenance process is standardized, and maintenance personnel can directly obtain spare part information and operation guidelines based on the work orders, reducing the maintenance preparation time, improving the efficiency of fault handling, and at the same time reducing the dependence on the experience of maintenance personnel.

[0131] Specifically, the input trigger for the work order generation logic is the equipment alarm level and the abnormal type (such as "mechanical wear"), and the matching rule is: if the abnormal type is "excessive motor harmonic ratio", match the "motor fault" entry in the knowledge base, and extract the spare part list (such as "motor winding") and the video link.

[0132] For those skilled in the art, it is obvious that the present invention is not limited to the details of the above exemplary embodiments, and without departing from the spirit or basic characteristics of the present invention, the present invention can be implemented in other specific forms.

[0133] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Among them, artificial intelligence is a theory, method, technology and application system that uses a digital computer or a machine controlled by a digital computer to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0134] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. An abnormal recognition system for campus self-service laundry equipment based on the Internet of Things, characterized in that, The system includes: A user operation behavior alignment module, which is used to generate a user operation time sequence log of the laundry equipment, align the user operation time sequence log with a preset standard operation template through a dynamic time warping algorithm, and calculate the operation behavior deviation degree to generate a first weight factor representing the probability of user behavior abnormality; A multi-source state feature extraction module, which is used to construct an equipment state matrix of the laundry equipment based on the real-time acquired multi-source operation data, and extract the characteristic parameters in the equipment state matrix that exceed the preset safety threshold to generate a second weight factor representing the abnormality of the equipment physical state; A health degree fusion decision module, which is used to perform weighted decision on the first weight factor and the second weight factor to obtain the health degree score of the laundry equipment. When the health degree score is lower than the first threshold, it triggers equipment-level anomaly detection to obtain the equipment-level anomaly type of the laundry equipment; An anomaly propagation topology modeling module, which is used to construct a fusion topology graph of the physical connection topology and the data communication topology of the laundry equipment, dynamically allocate the anomaly propagation weights of each node of the fusion topology graph based on the equipment historical failure rate and the real-time load state, and generate an anomaly propagation path probability graph according to the connection strength between the nodes of the fusion topology graph and the anomaly propagation weights; A spatio-temporal warning visualization module, which is used to calculate the group anomaly probability of the laundry equipment based on the equipment-level anomaly type and the anomaly propagation path probability graph, divide multiple warning levels according to the probability interval, and fuse the location information of the laundry equipment and the multiple warning levels to generate a visual warning log.

2. The abnormal recognition system of a campus self-service laundry device based on the Internet of Things according to claim 1, characterized in that, When the user operation behavior alignment module executes to generate a user operation time sequence log of the laundry equipment, it includes: Collecting user operation behavior data, where the user operation behavior data includes: reservation period, washing mode selection, and payment record; Generating a user operation time sequence log containing the reservation period, washing mode selection sequence, and payment success rate based on the user operation behavior data.

3. The abnormal recognition system for campus self-service laundry equipment based on the Internet of Things according to claim 1, characterized in that, When the user operation behavior alignment module executes to align the user operation time sequence log with a preset standard operation template through a dynamic time warping algorithm and calculate the operation behavior deviation degree as the first weight factor of the equipment health degree, it includes: Extracting the reservation period distribution density, mode switching frequency, and payment interruption times in the user operation time sequence log as key behavior indicators; Calculating the cumulative path distance between each key behavior indicator and the preset standard operation template through a dynamic time warping algorithm; Performing weighted summation on the cumulative path distance to obtain the operation behavior deviation degree, and using the operation behavior deviation degree as the first weight factor of the equipment health degree.

4. The abnormal recognition system for campus self-service laundry equipment based on the Internet of Things according to claim 3, characterized in that, The generation method of the preset standard operation template in the user operation behavior alignment module includes: Collecting user operation data of N laundry equipment within a normal operation and maintenance cycle to form a training set; Performing Gaussian kernel density estimation on the reservation period data in the training set to generate a period distribution probability curve; Extracting the optimal path of washing mode switching as the standard mode sequence; Statistical normal distribution interval of payment success rate as the benchmark payment behavior indicator; Generate a preset standard operation template based on the time period distribution probability curve, the standard mode sequence, and the benchmark payment behavior metrics.

5. The abnormal recognition system for campus self-service laundry equipment based on the Internet of Things according to claim 1, wherein When the multi-source state feature extraction module constructs the device state matrix of the laundry device based on the real-time acquired multi-source operation data, it includes: Real-time acquire the multi-source operation data of the laundry device, where the multi-source operation data includes: vibration spectrum data and motor current waveform; Perform wavelet packet decomposition on the vibration spectrum data, and extract the energy ratio of the preset frequency band as the mechanical wear feature; After performing fast Fourier transform on the motor current waveform, calculate the amplitude ratio of the fundamental wave to the third harmonic as the electrical anomaly feature; Align the mechanical wear feature and the electrical anomaly feature according to the timestamp to form a two-dimensional state matrix, and determine the two-dimensional state matrix as the device state matrix of the laundry device.

6. The abnormal recognition system of a campus self-service laundry device based on the Internet of Things according to claim 1, characterized in that When the health degree fusion decision module performs weighted decision on the first weight factor and the second weight factor to obtain the health degree score of the laundry device, it includes: Dynamically calculate the weight distribution ratio of the first weight factor and the second weight factor based on the entropy method; Calculate the health degree score of the laundry device using the weighted summation algorithm and the weight distribution ratio, where the weighted summation algorithm is: ； Wherein, is the health score of the laundry equipment, is the weight allocation ratio corresponding to the first weight factor, is the weight allocation ratio corresponding to the second weight factor, is the first weight factor, is the second weight factor.

7. The abnormal recognition system for campus self-service laundry equipment based on the Internet of Things according to claim 1, characterized in that, When the abnormal propagation topology modeling module dynamically assigns the abnormal propagation weights of each node in the fusion topology graph based on the device historical failure rate and the real-time load status, it includes: Determine the initial static weight of each node according to the betweenness centrality of the laundry device in the fusion topology graph and the device historical failure rate; When the real-time load status of the laundry device exceeds 70% of the rated value, determine the dynamic correction coefficient of the initial static weight, and correct the node weights in the fusion topology graph based on the dynamic correction coefficient; When the laundry device is in the maintenance lock state, force the node weights of the fusion topology graph to be zero to block abnormal propagation.

8. The abnormal recognition system for campus self-service laundry equipment based on the Internet of Things according to claim 1, characterized in that, When the spatio-temporal warning visualization module calculates the group abnormal probability of the laundry device based on the device-level abnormal type and the abnormal propagation path probability graph, it includes: Input the device-level abnormal type and the abnormal propagation path probability graph into the spatio-temporal propagation model, and calculate the group abnormal probability of the laundry device, where the spatio-temporal propagation model satisfies: Adopt an exponential decay propagation rule based on physical distance for mechanical anomalies; Adopt a linear propagation rule based on the number of network hops for data anomalies; When the two types of anomalies coexist, preferentially perform mechanical anomaly propagation calculation and superimpose the data anomaly correction term.

9. The abnormal recognition system of a campus self-service laundry device based on the Internet of Things according to claim 8, characterized in that, The calculation formula for the group abnormal probability in the spatio-temporal warning visualization module is as follows: ； Among them, is the group abnormal probability of the laundry equipment, is the number of equipment that may malfunction due to the spread of the initial abnormal equipment and is the number of all laundry equipment in a specific area, is the equipment number of the affected laundry equipment, is the laundry equipment to which the abnormal condition spreads and is the final probability that the abnormal condition spreads to the laundry equipment is the laundry equipment and the probability of independent occurrence of abnormal conditions (regardless of spread), is the laundry equipment, is the equipment number of the initial abnormal equipment.

10. An abnormal recognition system for campus self-service laundry equipment based on the Internet of Things according to any one of claims 1 to 8, characterized in that When the spatio-temporal warning visualization module performs fusion of the location information of the laundry device and the multi-level alarm levels to generate a visual alarm log, it includes: According to the location information of the laundry device and the multi-level alarm levels, render a three-level heat layer of the abnormal devices in the laundry device on the campus electronic map, where red, yellow, and blue correspond to high risk, medium risk, and low risk respectively; Mark the possible influence range and the estimated arrival time of the abnormal propagation of the abnormal devices; Automatically generate a maintenance work order that includes a spare part replacement list and an operation guidance video based on a pre-associated maintenance knowledge base.

Images