Fault early warning method and system for candy production equipment based on data processing

Abstract

The invention belongs to the technical field of equipment fault identification, and particularly relates to a candy production equipment fault early warning method and system based on data processing, and the method comprises the steps: obtaining sensor window data of candy production equipment; performing frequency domain analysis on the window data to evaluate the operation signal regularity of the window data, and constructing an adaptive truncation distance based on the operation signal regularity; performing density peak clustering analysis on the window data based on the self-adaptive truncation distance so as to judge the running state of the candy production equipment; and dynamically triggering fault early warning according to the operation state so as to realize fault early warning of the candy production equipment. According to the method, normal operation fluctuation and random abnormal fault symptoms can be intelligently distinguished, the problems of missing detection and false alarm caused by fixed parameters are solved, and the detection accuracy is remarkably improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of equipment fault identification. More particularly, the present application relates to a data processing-based confectionery production equipment fault early warning method and system. BACKGROUND

[0002] The confectionery production process involves key equipment such as high-speed running mixers, molding machines, cutting machines, and packaging machines. The continuous and stable operation of these equipment is the core to ensure product quality and production efficiency. Any unexpected downtime can lead to raw material waste, production plan interruption, and high maintenance costs. Therefore, comprehensive and high-precision online monitoring of the running state of these equipment and effective early warning before failure occur are indispensable key links to ensure stable production in modern food manufacturing processes.

[0003] Currently, to meet the above early warning needs, the commonly used method in the prior art is to use the density peak clustering (DPC) algorithm. By installing vibration or temperature sensors on the equipment and setting a sliding window to analyze the collected data, the DPC algorithm is used to calculate the local density and relative distance of the data points to identify the distribution pattern of the data points. This method attempts to distinguish abnormal signals from continuous data streams to some extent to achieve fault early warning.

[0004] However, the performance of the DPC algorithm, especially the accuracy of its local density calculation, is highly dependent on the cutoff distance, which is a global fixed value set by engineers in the prior art. However, this fixed setting cannot adapt to the differences in normal running vibration and temperature characteristics between different equipment and different working conditions. If the value is set larger to tolerate normal fluctuations, minor fault signs may be ignored, causing missed detection. If the value is set smaller to capture minor signs, normal process fluctuations may be misjudged as abnormal, leading to frequent false alarms and unnecessary downtime for maintenance. SUMMARY

[0005] To solve the technical problem that the existing density peak clustering technology cannot adapt to the running characteristics of different confectionery production equipment, the present application provides solutions in the following aspects.

[0006] In a first aspect, the present application provides a data processing-based confectionery production equipment fault early warning method, comprising: acquiring sensor window data of a confectionery production equipment; performing frequency domain analysis on the window data to evaluate the running signal regularity of the window data and constructing an adaptive cutoff distance based on the running signal regularity; performing density peak clustering analysis on the window data based on the adaptive cutoff distance to determine the running state of the confectionery production equipment; and dynamically triggering a fault early warning according to the running state to achieve fault early warning for the confectionery production equipment.

[0007] The application can deeply understand the internal structure of the sensor signal by introducing the frequency domain analysis into the parameter self-adaptation of the DPC algorithm, and dynamically construct a self-adaptive cut-off distance most suitable for the current local working condition in real time, which enables the algorithm to intelligently distinguish between the normal operation signal with periodic rules and the fault symptom signal with random mutations, effectively solves the inaccurate situation between missed detection and false alarm of the existing fixed parameter technology, and significantly improves the comprehensive accuracy of fault warning.

[0008] Preferably, the sensor window data of the confectionery production equipment is obtained by continuously measuring the confectionery production equipment at a reference frequency through a sensor to generate an original data stream, applying a first-in-first-out sliding analysis window to the original data stream, and performing maximum-minimum normalization processing on the data in the sliding analysis window to obtain the window data.

[0009] Preferably, the frequency domain analysis of the window data includes performing Fourier transform on the window data to obtain a complex frequency spectrum, and calculating a one-sided power spectrum based on the complex frequency spectrum, wherein the one-sided power spectrum represents the energy distribution of the window data at different frequencies.

[0010] Preferably, the operation signal regularity satisfies the following relationship: ; wherein, is the operation signal regularity; is the power spectral density value at the frequency is the peak frequency in the preset characteristic frequency band; is the frequency range of the preset characteristic frequency band; is the neighborhood range of the peak frequency; is a normal number to prevent the denominator from being zero.

[0011] Preferably, the adaptive cut-off distance is constructed in the following manner: ; wherein, is the adaptive cut-off distance, and are the minimum value and the maximum value of the preset cut-off distance, respectively, is the operation signal regularity, is the regularity influence factor.

[0012] An intelligent mechanism for automatically switching the analysis scale is realized, and the sensitivity of the algorithm is proportional to the regularity of the background signal. When the background signal regularity is high during normal operation of the equipment, the adaptive cut-off distance is large and can tolerate large but regular fluctuations. When a fault symptom background signal with low regularity appears, the minimum value is automatically taken to ensure high capture sensitivity to small random symptoms.

[0013] ​Preferably, the process of performing density peak clustering analysis on the window data based on the adaptive truncation distance involves using the adaptive truncation distance as the neighborhood radius for each data point within the window data and calculating its local density.

[0014] Preferably, determining the operating status of the candy production equipment includes: identifying the number of density peaks based on the local density and relative distance of each data point; if the number of density peaks is not greater than 1, it is determined to be a normal state; if the number of density peaks is greater than 1, it is determined to be an abnormal state.

[0015] By introducing frequency domain analysis, we can gain insight into and distinguish the intrinsic structure of signals, thereby determining whether data fluctuations stem from regular normal operation or random abnormal fault symptoms.

[0016] Preferably, the step of dynamically triggering fault warning based on the operating status includes: maintaining normal monitoring when the operating status is normal; and issuing a fault warning signal when the operating status is abnormal.

[0017] Preferably, the density peak clustering analysis further includes: for each data point within the window data, calculating the distance between it and all data points with higher local density than itself, and taking the minimum of these distances as the relative distance of the data point.

[0018] Secondly, the present invention provides a data processing-based fault early warning system for candy production equipment, including 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 data processing-based fault early warning method for candy production equipment is implemented.

[0019] By adopting the above technical solution, a computer program is generated from the above-mentioned data processing-based candy production equipment fault early warning method and stored in the memory so that it can be loaded and executed by the processor. In this way, a terminal device can be made based on the memory and the processor for convenient use.

[0020] This invention can deeply understand the internal structure of the sensor signals of equipment, thereby accurately distinguishing between regular normal operation signals and random abnormal fault symptoms. It fundamentally solves the contradiction between missed detection and false alarm due to fixed parameters in traditional technology, and greatly improves the accuracy and reliability of fault early warning.

[0021] Furthermore, this invention achieves complete self-adaptation of key parameters, with the cutoff distance dynamically generated by real-time data features, eliminating the need for manual parameter adjustment. This enables the system to automatically adapt to the characteristics of different equipment models or different production conditions, exhibiting high robustness and significantly reducing deployment and maintenance costs in industrial settings. While ensuring high-precision early warning, it avoids invalid data analysis, improves the overall efficiency of fault early warning, and enables it to truly operate stably in complex production environments. Attached Figure Description

[0022] The above and other objects, features, and advantages of exemplary embodiments of the present invention will become readily apparent from the following detailed description taken in conjunction with 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: Figure 1 This is a flowchart illustrating a data processing-based fault early warning method for candy production equipment according to the present invention; Figure 2 This is a schematic diagram illustrating the application effect verification provided in an embodiment of the present invention. Detailed Implementation

[0023] 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.

[0024] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0025] This invention discloses a fault early warning method for candy production equipment based on data processing, referring to... Figure 1 This includes steps S1-S4: S1. Obtain sensor window data from the candy production equipment.

[0026] In an alternative embodiment, a vibration sensor may be used as a monitoring sensor, mounted on the bearing housing of the candy mixer, to continuously measure the mixer running at a relatively low reference frequency, exemplarily, such as 1 kHz, thereby generating a one-dimensional raw data stream of vibration signals.

[0027] Furthermore, in order to perform real-time analysis, a length of [missing information] is applied to the raw data stream. First-in-first-out sliding analysis window, window length It needs to be large enough to contain meaningful runtime state information, but not so large that it loses locality. For example, it can be set to... Data points.

[0028] Specifically, at each sampling time, the sliding analysis window slides forward one data point, always containing the data set from the most recent period. To eliminate signal baseline drift caused by changes in ambient temperature or long-term sensor operation, as well as numerical differences caused by the production equipment itself under different operating conditions, the data within the sliding analysis window needs to undergo max-min normalization processing. This involves mapping all data points within the window to a linear transformation. Within the interval, window data is obtained.

[0029] In this way, through windowing and normalization, a standardized local data sample representing the minute fluctuations in the local operating status of the candy production equipment can be obtained, laying the foundation for subsequent accurate frequency domain analysis.

[0030] S2. Perform frequency domain analysis on the window data to evaluate the operational signal regularity of the window data, and construct an adaptive truncation distance based on the operational signal regularity.

[0031] In an optional embodiment, a fast Fourier transform is performed on the normalized window data sequence to convert it from the time domain to the frequency domain, resulting in a complex spectrum. Then, the square of the magnitude of each frequency component of the complex spectrum is calculated to obtain a one-sided power spectrum, which clearly shows the distribution of signal energy at different frequencies.

[0032] In this optional embodiment, the operational signal regularity of the window data can be further extracted to quantitatively distinguish between regular normal operating signals and random abnormal fluctuations. The operational signal regularity satisfies the following relationship:

[0033] in, For the regularity of the operating signal; In frequency The power spectral density value at that location; The peak frequency within the preset characteristic frequency band; The frequency range of the preset characteristic frequency band; The neighborhood range of the peak frequency; To prevent positive numbers with a denominator of zero.

[0034] Specifically, a preset characteristic frequency band is used to reflect the normal operating characteristics of the equipment. This is usually related to the motor speed and its harmonics. For example, if the motor speed is 3000 rpm, the characteristic frequency range can be set to [40Hz, 60Hz]. The peak point of the power spectrum is found within this frequency band, and its corresponding peak frequency is the main operating frequency. The value is 2Hz. Values .

[0035] Furthermore, an adaptive truncation distance can be constructed using the obtained operational signal regularity. The adaptive truncation distance is constructed as follows:

[0036] in, For adaptive truncation distance, and These are the minimum and maximum preset cutoff distances, respectively, for example. and The values ​​are 0.02 and 0.5 respectively. For the sake of signal regularity, The regularity influence factor is 0.5, for example.

[0037] For example, in scenario A, the mixer is stably mixing a high-viscosity syrup. The sensor signal exhibits highly regular periodic vibrations. After frequency domain analysis, assuming the total energy within the characteristic frequency band [40Hz, 60Hz] is 100 units, and due to stable operation, the energy is highly concentrated around the power frequency of 50Hz, the energy in the peak frequency neighborhood [48Hz, 52Hz] reaches 81 units, then the calculated energy is... This value is close to 1, indicating that the signal is highly regular; further calculations can be performed. The value is 0.452, which shows that a value close to 0.452 was generated. The relatively large cutoff distance allows subsequent analysis to tolerate such large but regular fluctuations, avoiding false alarms.

[0038] For example, in scenario B, an early crack appears in the mixer bearing. The sensor signal, on top of the normal 50Hz vibration, is superimposed with a random high-frequency impact signal caused by the crack. After frequency domain analysis, the signal energy becomes diffuse. The total energy in the characteristic frequency band [40Hz, 60Hz] may drop to 60 units, while the energy in the peak neighborhood [48Hz, 52Hz] is only 15 units. Therefore, it can be calculated that... The value is 0.25, which is low, indicating poor signal regularity; further calculations can be performed... With a value of 0.26, we can see that a small cutoff distance is automatically generated, which has high sensitivity and can identify local data points caused by minor defects as outliers, ensuring that no outliers are missed.

[0039] Thus, through in-depth analysis based on frequency domain features, a cutoff distance suitable for the current equipment operating conditions is provided in real time for the subsequent DPC algorithm.

[0040] S3. Perform density peak clustering analysis on the window data based on adaptive truncation distance to determine the operating status of candy production equipment.

[0041] In an optional embodiment, the calculation of the local density of each data point within the window data depends entirely on the current window generation. The value is used as the neighborhood radius. After calculating the local density of all points, the relative distance of each point is calculated. For each data point within the window data, the distance between it and all data points with higher local density than itself is calculated, and the minimum of these distances is used as the relative distance of that data point.

[0042] Furthermore, by analyzing the obtained local density and relative distance decision maps, the number of density peaks is identified. If the number of density peaks is no greater than 1, the current window region can be judged as being in a normal state; if the number of density peaks is greater than 1, the current window region can be judged as being in an abnormal state. Because... Intelligent adjustment allows the number of density peaks to more accurately reflect the true state of the region. For example, in high-load operating areas, large... This will cause all points to belong to a single neighborhood, thus forming only one density peak.

[0043] Thus, by embedding a dynamic adaptive cutoff distance, the discrimination capability of the DPC algorithm is effectively improved, enabling it to more reliably distinguish between normal and abnormal states.

[0044] S4. Dynamically trigger fault warnings based on operating status to provide early warnings of faults in candy production equipment.

[0045] In an optional embodiment, when the operating state is normal, the system maintains normal monitoring and does not issue alarms; when the operating state is abnormal, the system immediately triggers a fault warning, for example, by popping up an alarm window through the industrial control computer interface (HMI), illuminating the sound and light alarm, or sending a warning notification to the mobile device of the maintenance personnel to prompt maintenance.

[0046] like Figure 2The diagram shown is a schematic diagram of the application effect verification provided by the embodiment of the present invention. Sub-diagram A shows a vibration signal of a candy production equipment. The signal in the stable operation area with data points of 0-400 is generally stable and contains only weak background noise, representing a high-quality healthy state. The signal in the high-load operation area with data points of 400-900 shows clear and regular periodic fluctuations. In the fault symptom area with data points of 900-1300, a non-periodic impact defect appears on the stable background signal.

[0047] Subgraph B illustrates the dynamic changes in the adaptive cutoff distance calculated by this invention, in both the stable and fault regions. Both are at an extremely low level because the algorithm, through frequency domain analysis, determines that the background signals in both regions are non-periodic. The algorithm then uses the most sensitive detection scale to ensure that subsequent DPC analysis can capture the smallest anomalous signals. However, in the high-load operating region, when the analysis window enters that region... The signal rapidly improved to a very high level because the algorithm identified strong periodic components and determined them to be regular patterns. To prevent the DPC algorithm from misclassifying each normal peak and trough as an independent density peak, the algorithm actively and significantly increased the density of the signal. The value was increased to the analytical scale to tolerate such normal fluctuations.

[0048] Subgraph C shows the final fault warning result. In the stable and high-load operation regions, the algorithm correctly classifies them as normal states, and no warning is triggered, which is represented by the blue area, thus avoiding false alarms. However, in the fault symptom region, the algorithm selects a very small value for this region. Subsequent DPC analysis can identify data points that are caused by the impact and deviate from the background, and judge them as anomalies. Therefore, the system immediately triggers an early warning, which is shown as a red area, ensuring accurate early warning of real faults.

[0049] In this way, by dynamically triggering early warnings based on real-time and accurate status judgment results, while ensuring coverage of all potential fault signs, false alarms under normal operating conditions are greatly reduced, achieving an intelligent balance between early warning accuracy and production efficiency.

[0050] This invention also discloses a data processing-based fault early warning system for candy production equipment, including a processor and a memory. The memory stores computer program instructions, and when the computer program instructions are executed by the processor, a data processing-based fault early warning method for candy production equipment according to this invention is implemented.

[0051] 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.

[0052] 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.

[0053] 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 method for early warning of faults in confectionery production equipment based on data processing, characterized in that, include: Acquire sensor window data from the candy production equipment; Frequency domain analysis is performed on the window data to evaluate the operational signal regularity of the window data, and an adaptive truncation distance is constructed based on the operational signal regularity. Density peak clustering analysis is performed on the window data based on the adaptive cutoff distance to determine the operating status of the candy production equipment; The system dynamically triggers fault warnings based on the operating status, thereby enabling early warning of faults in the candy production equipment.

2. The method for early warning of faults in confectionery production equipment based on data processing according to claim 1, characterized in that, The acquisition of sensor window data from the candy production equipment includes: The candy production equipment is continuously measured by sensors at a reference frequency to generate a raw data stream; A first-in-first-out sliding analysis window is applied to the original data stream, and the data within the sliding analysis window is subjected to max-min normalization to obtain the window data.

3. The method for early warning of faults in confectionery production equipment based on data processing according to claim 1, characterized in that, The frequency domain analysis of the window data includes: Perform a Fourier transform on the window data to obtain a complex spectrum; The one-sided power spectrum is calculated based on the complex spectrum, and the one-sided power spectrum characterizes the energy distribution of the window data at different frequencies.

4. The method for early warning of faults in confectionery production equipment based on data processing according to claim 1, characterized in that, The regularity of the operating signal satisfies the following relationship: in, For the regularity of the operating signal; In frequency The power spectral density value at that location; The peak frequency within the preset characteristic frequency band; The frequency range of the preset characteristic frequency band; The neighborhood range of the peak frequency; To prevent positive numbers with a denominator of zero.

5. The method for early warning of faults in confectionery production equipment based on data processing according to claim 1, characterized in that, The adaptive cutoff distance is constructed as follows: in, For adaptive truncation distance, and These are the minimum and maximum preset cutoff distances, respectively. For the sake of signal regularity, This represents the factor influencing regularity.

6. The method for early warning of faults in confectionery production equipment based on data processing according to claim 1, characterized in that, The process of performing density peak clustering analysis on the window data based on the adaptive truncation distance is as follows: for each data point in the window data, the adaptive truncation distance is used as the neighborhood radius to calculate its local density.

7. A method for early warning of faults in confectionery production equipment based on data processing, as described in claim 1 or 6, characterized in that, The determination of the operating status of the candy production equipment includes: Identify the number of density peaks based on the local density and relative distance of each data point; If the number of density peaks is not greater than 1, it is judged as a normal state; if the number of density peaks is greater than 1, it is judged as an abnormal state.

8. The method for early warning of faults in confectionery production equipment based on data processing according to claim 1, characterized in that, The method of dynamically triggering fault warnings based on the operating status includes: When the operating status is normal, normal monitoring is maintained; When the operating state is abnormal, a fault warning signal is issued.

9. A method for early warning of faults in confectionery production equipment based on data processing according to claim 6, characterized in that, The density peak clustering analysis also includes: For each data point within the window data, calculate its distance to all data points with higher local density than itself, and take the minimum of these distances as the relative distance of that data point.

10. A fault early warning system for confectionery production equipment based on data processing, characterized in that, include: A processor and a memory, wherein the memory stores computer program instructions that, when executed by the processor, implement a data processing-based fault early warning method for confectionery production equipment according to any one of claims 1-9.

Images