Kitchen waste treatment fine slag conveying path optimization system based on time sequence characteristic analysis

Abstract

The invention relates to the technical field of intelligent conveying, in particular to a kitchen waste treatment fine slag conveying path optimization system based on time sequence characteristic analysis. The edge layer performs local processing and fluctuation analysis on the real-time operation data of the kitchen waste fine residues, and adjusts the acquisition configuration of a sensor network; the cloud layer calculates a reconstruction error through a VAE model to form a real-time enhanced feature flow; the LSTM model generates dynamic transmission job scheduling parameters; the PLC receives the dynamic conveying operation scheduling parameters and adjusts the running frequency of the spiral conveyor and the deflection angle of a baffle at a discharge port in real time; and trend prediction is conducted on the abrasion state of the bottom of the spiral conveyor, maintenance early warning is generated, and a standardized three-cloth five-oil repairing instruction is triggered. At present, an improved conveying system capable of solving the problems of splashing and leakage in fine slag conveying is needed in the field. Therefore, the kitchen waste treatment fine slag conveying path optimization system based on time sequence characteristic analysis is provided.

Description

Technical Field

[0001] This invention relates to the field of intelligent conveying technology, specifically to a system for optimizing the conveying path of fine residue from food waste treatment based on time-series feature analysis. Background Art

[0002] In the food waste treatment process, fine residue is a common intermediate product, characterized by high water content and rich organic matter. In the material handling stage, screw conveyors, as a standardized asset, are widely used to transport various materials, including fine residue.

[0003] In existing operation management models, screw conveyors are typically scheduled using fixed operating parameters, and regular or reactive maintenance plans are implemented. This model, based on preset parameters and reactive responses, is a common equipment scheduling and asset maintenance strategy in current material handling operations. However, the operational efficiency of this conventional operation scheduling and asset management strategy can be improved under certain conditions. When the conveying operation fails to dynamically match the real-time status of the materials, unexpected events such as material splashing and equipment leakage may occur. These events directly lead to additional human resource consumption and increased maintenance costs, and may affect the overall production scheduling efficiency due to unplanned downtime.

[0004] Therefore, there is a need in this field for an improved conveying system that can solve the problems of splashing and leakage during the conveying of fine slag, thereby optimizing operating costs, improving equipment resource utilization, and ensuring the execution of production plans. To this end, a fine slag conveying path optimization system based on time-series characteristic analysis for food waste treatment is proposed. Summary of the Invention

[0005] The purpose of this invention is to provide a system for optimizing the conveying path of fine food waste residue based on time-series feature analysis. The edge layer performs local processing and fluctuation analysis on real-time operational data of fine food waste residue and adjusts the sensor network's acquisition configuration. The cloud layer calculates reconstruction errors using a VAE model, forming a real-time enhanced feature stream. An LSTM model generates dynamic conveying operation scheduling parameters. The PLC controller receives these parameters and adjusts the operating frequency of the screw conveyor and the deflection angle of the discharge port baffle in real time. It also predicts the wear status of the screw conveyor bottom, generates maintenance warnings, and triggers standardized "three-cloth, five-oil" repair instructions. Currently, there is a need in the art for an improved conveying system that can solve the problems of splashing and leakage during fine residue conveying. Therefore, this system for optimizing the conveying path of fine food waste residue based on time-series feature analysis is proposed.

[0006] To achieve the above objectives, the present invention provides the following technical solution: The real-time operating condition perception module at the edge layer performs local processing and fluctuation analysis on the real-time operating data of kitchen waste fine residue, and dynamically adjusts the acquisition configuration of the sensor network based on the data fluctuation analysis report; the cloud layer calculates and analyzes the reconstruction error of the preprocessed data through the VAE model, generates operating condition analysis feature data, and combines the preprocessed data to form a real-time enhanced feature stream. The dynamic scheduling decision module uses an LSTM model to infer the real-time enhanced feature stream and generate dynamic delivery job scheduling parameters. The closed-loop control module, PLC controller receives the dynamic conveying operation scheduling parameters, and adjusts the operating frequency of the screw conveyor and the deflection angle of the discharge port baffle in real time by driving the frequency converter and servo motor; quantifies the discharge port splash data, forms process feedback data and sends it back to the LSTM model; The predictive maintenance module uses a time series prediction algorithm to predict the wear status of the bottom of the screw conveyor, generate maintenance warnings, and trigger standardized "three-cloth five-oil" repair instructions.

[0007] Furthermore, the process of dynamically adjusting the acquisition configuration of the sensor network is as follows: The sensor network's acquisition configuration is dynamically adjusted based on the data fluctuation analysis report. When the data variance in the data fluctuation analysis report exceeds a first preset threshold, the sampling frequency of the sensor network is increased; when the data variance is lower than a second preset threshold, the sampling frequency of the sensor network is decreased and the communication routing between wireless sensor nodes is adjusted.

[0008] Furthermore, the process of forming the real-time enhanced feature stream is as follows: The cloud layer encodes and decodes the preprocessed data using a VAE model to obtain reconstructed data; it calculates the Euclidean distance between the preprocessed data and the reconstructed data to obtain a quantified reconstruction error value; the reconstruction error value is used as the working condition analysis feature data and concatenated with the preprocessed data to form a real-time enhanced feature stream.

[0009] Furthermore, the process for generating the dynamic transport operation scheduling parameters is as follows: By loading pre-trained LSTM model weights, the system receives real-time augmented feature streams and performs matrix operations and activation function processing to output dynamic delivery scheduling parameters including the target running frequency and target deflection angle. The target running frequency is generated according to a set of preset scheduling rules. These rules include: when the humidity value in the real-time augmented feature stream is higher than a first humidity threshold and the particle size value is lower than a first size threshold, the target running frequency is set to 35Hz; when the humidity value in the real-time augmented feature stream is lower than a second humidity threshold and the particle size value is higher than a second size threshold, the target running frequency is set to a value higher than 35Hz.

[0010] Furthermore, the process of adjusting the operating frequency of the screw conveyor in real time is as follows: The PLC controller receives and parses the dynamic conveying operation scheduling parameters to obtain the target operating frequency value. The PLC controller sends a control message to the frequency converter via the Modbus communication protocol. The control message contains the target operating frequency value. The frequency converter receives and parses the control message and adjusts the frequency of the three-phase AC power output to the screw conveyor drive motor so that the actual operating frequency of the drive motor matches the target operating frequency value.

[0011] Furthermore, the process of adjusting the deflection angle of the discharge port baffle in real time is as follows: The PLC controller receives and parses the dynamic conveying operation scheduling parameters to obtain the target deflection angle value; the PLC controller sends pulse sequence instructions to the servo driver, the number of which is proportional to the target deflection angle value; the servo driver drives the servo motor to rotate, and drives the discharge port baffle to deflect through the mechanical linkage mechanism; the encoder built into the servo motor feeds back the actual rotation angle to the servo driver, forming a position closed-loop control.

[0012] Furthermore, the process of driving the discharge port baffle to deflect via the mechanical linkage mechanism includes: A path optimization unit is configured at the slag discharge port of the screw conveyor. The path optimization unit integrates a flow guide pipe and an outlet baffle to achieve precise control over the material's falling path and height.

[0013] Furthermore, the process of transmitting the feedback data back is as follows: Video streams of the splashing process are acquired by image sensors deployed at the discharge port; image processing algorithms are used to identify and count the number of splashing particles per unit time in the video stream, forming quantified splashing data; the quantified splashing data is associated with the dynamic conveying operation scheduling parameters at the corresponding time, and packaged into structured process feedback data; the LSTM model receives the process feedback data, calculates the loss value between the predicted splashing result generated by the dynamic conveying operation scheduling parameters and the quantified splashing data, and uses the gradient descent algorithm to update the network weights of the LSTM model online to complete strategy fine-tuning.

[0014] Furthermore, the predictive maintenance module includes: Health status monitoring unit: Collects vibration acceleration time-series data and surface temperature time-series data of the screw conveyor during operation through vibration sensors and temperature sensors; Trend prediction unit: Inputs the vibration acceleration time series data and the surface temperature time series data into a preset time series prediction algorithm model; The time series prediction algorithm model decomposes the data into trend items, seasonal items and holiday items, and predicts the wear index within a specific future time window, generating a wear state prediction report; Early warning and scheduling unit: compares the wear index in the wear status prediction report with the preset maintenance threshold; when the wear index exceeds the maintenance threshold, a maintenance early warning is generated; the maintenance early warning automatically triggers a standardized "three-cloth five-oil" repair instruction, which includes the operation steps of using resin glue and curing agent mixed in a certain proportion, combined with glass fiber cloth, and repairing the leakage point according to the "three-cloth five-oil" process, and pushes the instruction to the maintenance work order module in the enterprise resource planning system.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. By generating a pulse sequence precisely corresponding to the target angle through a PLC controller, the servo system is directly driven, achieving digital and high-precision control of the baffle angle. Simultaneously, the encoder built into the servo motor feeds back the actual rotation angle to the driver, forming a hardware-level, real-time position closed loop. This design not only fundamentally overcomes the positioning errors caused by uncertainties such as pressure fluctuations and oil temperature changes common in traditional pneumatic or hydraulic actuators, ensuring the baffle maintains extremely high stability and repeatability even under high-speed dynamic adjustments, but more importantly, it guarantees that the refined scheduling parameters calculated by the upper-level AI model can be accurately physically executed. This allows the entire intelligent anti-splash strategy to be precisely implemented, significantly improving the system's dynamic adaptability to complex and variable materials. This achieves the goals of optimizing operating costs, improving equipment resource utilization, and ensuring the execution of production plans.

[0016] 2. By deploying image sensors and employing image processing algorithms, this solution successfully transforms the originally vague and difficult-to-quantify "splashing" phenomenon in the physical world into structured data of "the number of splashing particles per unit time," which can be precisely processed by computers. This provides an objective and reliable feedback standard for the self-optimization of AI models. Based on this, by correlating this quantified splashing data with scheduling parameters and using gradient descent to update the network weights of the LSTM model online, we construct a data-driven, self-evolving strategy fine-tuning closed loop. This means the system is no longer a static model that remains unchanged, but a dynamic system capable of continuous learning. It can proactively adapt to equipment aging, environmental changes, and even long-term drift in material properties, thereby maintaining and continuously improving the accuracy and robustness of its conveying path optimization during long-term operation. This achieves the goals of optimizing operating costs, improving equipment resource utilization, and ensuring the execution of production plans.

[0017] 3. By automatically converting the technical predictions of equipment health status—i.e., maintenance alerts—using time-series forecasting algorithms into standardized "three-cloth, five-oil" repair instructions, and seamlessly pushing these instructions to the maintenance work order module of the Enterprise Resource Planning (ERP) system, a complete information chain is successfully established from the underlying equipment status perception to the upper-level enterprise operation management. This design not only elevates equipment maintenance from reactive response to proactive planning, effectively avoiding unplanned downtime caused by sudden failures, but more importantly, it elegantly integrates complex technical decisions into the enterprise's standardized operating processes. This achieves optimal scheduling of maintenance resources and standardized management of the maintenance process, ultimately fundamentally improving the management level of the entire equipment lifecycle and the overall operational efficiency of the enterprise. It achieves the goals of optimizing operating costs, improving equipment resource utilization, and ensuring the execution of production plans. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the process for optimizing the conveying path of fine residue in food waste treatment based on time-series feature analysis, provided in an embodiment of the present invention. Figure 2 This is a schematic diagram of the structure of the food waste fine residue conveying path optimization system based on time-series feature analysis provided in an embodiment of the present invention; Figure 3 This is a schematic diagram illustrating the process of real-time adjustment of the deflection angle of the discharge port baffle provided in an embodiment of the present invention. Detailed Implementation

[0019] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0020] Example 1 To address the issues of splashing and equipment leakage during the conveying of fine slag in a large-scale food waste recycling center, this invention proposes a food waste fine slag conveying path optimization system based on time-series feature analysis. The system's process and structure are as follows: Figure 1 and Figure 2 As shown, it mainly includes: Real-time operating condition perception module: dynamic scheduling decision module, closed-loop control module, predictive maintenance module; Real-time operating condition perception module: The edge layer performs local processing and fluctuation analysis on the real-time operating data of kitchen waste fine residue, and dynamically adjusts the acquisition configuration of the sensor network according to the data fluctuation analysis report; The cloud layer calculates and analyzes the reconstruction error of the preprocessed data through the VAE model, generates operating condition analysis feature data, and combines the preprocessed data to form a real-time enhanced feature stream.

[0021] Furthermore, the process of dynamically adjusting the acquisition configuration of the sensor network is as follows: When a batch of wet waste with extremely high moisture content enters the system, the data variance in the data fluctuation analysis report will quickly exceed the first preset threshold. The system will immediately increase the sampling frequency of the sensor network to accurately capture its characteristic changes. Conversely, when processing a batch of materials from a large canteen that contains a lot of solids such as rice, the data variance will be lower than the second preset threshold. The system will then reduce the sampling frequency of the sensor network and adjust the communication routing between wireless sensor nodes to save energy.

[0022] By dynamically adjusting the acquisition configuration based on data variance, the optimal allocation of sensor network resources is achieved. This reduces energy consumption and extends equipment lifespan when data is stable, while instantly increasing acquisition density during periods of drastic data fluctuation, ensuring high-fidelity capture of critical operating condition changes and improving the intelligence and efficiency of the system's sensing capabilities.

[0023] In the above scheme, the judgment criterion for adjusting the acquisition configuration is only the time-domain statistic of data variance. This has limitations in identifying periodic stationary signals containing important information (whose variance may be small), and may result in the loss of key structural features due to inappropriate reduction of the sampling frequency. Therefore, frequency domain feature analysis can be introduced into the generation of data fluctuation analysis reports to improve the scheme. While performing time-domain statistical analysis on real-time operational data, the edge layer performs a Fast Fourier Transform on the data to extract spectral features. When the energy of a specific frequency band in the spectral features exceeds a preset energy threshold, the system prioritizes increasing the sampling frequency of the sensor network instead of prioritizing responses to states where the data variance is below a second preset threshold. This scheme achieves an upgrade from single amplitude analysis to joint time-frequency analysis.

[0024] Furthermore, the process of forming the real-time enhanced feature stream is as follows: The cloud layer encodes and decodes the preprocessed data using a VAE model to obtain reconstructed data; it calculates the Euclidean distance between the preprocessed data and the reconstructed data to obtain a quantified reconstruction error value; the reconstruction error value is used as the working condition analysis feature data and concatenated with the preprocessed data to form a real-time enhanced feature stream.

[0025] Dynamic scheduling decision module: The LSTM model infers the real-time enhanced feature stream and generates dynamic delivery job scheduling parameters.

[0026] Furthermore, the process for generating the dynamic transport operation scheduling parameters is as follows: By loading pre-trained LSTM model weights, the system receives real-time augmented feature streams, performs matrix operations and activation function processing, and outputs dynamic delivery scheduling parameters including the target operating frequency and target deflection angle. This decision-making process is completed in milliseconds, far exceeding the speed of human reaction. The target operating frequency is ultimately verified according to a set of preset scheduling rules. These rules include: when the humidity value in the real-time augmented feature stream is higher than a first humidity threshold and the particle size value is lower than a first size threshold, the target operating frequency is forcibly set to a safe value of 35Hz; when the humidity value in the real-time augmented feature stream is lower than a second humidity threshold and the particle size value is higher than a second size threshold, the target operating frequency is set to a value higher than 35Hz, maximizing the delivery efficiency of the processing center while ensuring safety.

[0027] Closed-loop control module: The PLC controller receives the dynamic conveying operation scheduling parameters, and adjusts the operating frequency of the screw conveyor and the deflection angle of the discharge port baffle in real time by driving the frequency converter and servo motor; it quantifies the discharge port splash data and forms process feedback data to be sent back to the LSTM model.

[0028] Furthermore, the specific process for adjusting the operating frequency of the screw conveyor in real time is as follows: Figure 3 As shown: The PLC controller receives and parses the dynamic conveying operation scheduling parameters to obtain the target operating frequency value. The PLC controller sends a control message to the frequency converter via the Modbus communication protocol. The control message contains the target operating frequency value. The frequency converter receives and parses the control message and adjusts the frequency of the three-phase AC power output to the screw conveyor drive motor so that the actual operating frequency of the drive motor matches the target operating frequency value.

[0029] By clearly defining the Modbus communication protocol and control messages between the PLC and the frequency converter, it is ensured that the upper-level AI decision commands can be accurately and reliably translated into the physical actions of the underlying drive motors. This standardized digital communication method, compared to analog control, has higher anti-interference capabilities and control precision, improving the stability of system execution.

[0030] The frequency conversion control process in the above scheme is an open-loop control, meaning the PLC only sends commands in one direction without verifying the actual operating state of the drive motor. When there are drastic load changes or unexpected resistance in the transmission chain, the actual operating frequency of the motor may deviate from the target frequency, affecting the accuracy of the control. Therefore, a dual-mode control strategy combining LSTM-based predictive "feedforward control" and PID feedback control can be introduced to improve the scheme. The LSTM model in the dynamic scheduling decision module outputs a predictive load disturbance feedforward value in addition to the target operating frequency. This feedforward value represents the model's prediction of the load impact the screw conveyor will encounter in a very short time (e.g., within 0.5-1 second). The PLC controller superimposes this feedforward value with the output of the traditional PID feedback controller to form the final composite control command, which is then sent to the frequency converter. This improvement upgrades motor control from "error-driven lag compensation" to "predictive-driven advance compensation," significantly improving the system's dynamic response speed and operational smoothness in the face of drastic changes in operating conditions.

[0031] Furthermore, the process of adjusting the deflection angle of the discharge port baffle in real time is as follows: The PLC controller receives and parses the dynamic conveying operation scheduling parameters to obtain the target deflection angle value; the PLC controller sends pulse sequence instructions to the servo driver, the number of which is proportional to the target deflection angle value; the servo driver drives the servo motor to rotate, and drives the discharge port baffle to deflect through the mechanical linkage mechanism; the encoder built into the servo motor feeds back the actual rotation angle to the servo driver, forming a position closed-loop control.

[0032] By employing a servo motor position closed-loop control method, high-precision and high-dynamic-response adjustment of the discharge port baffle deflection angle is achieved. The encoder's real-time feedback mechanism ensures a high degree of consistency between the commanded angle and the actual angle, thereby enabling more precise matching of material flow patterns and further enhancing the anti-splash effect and system control accuracy.

[0033] In the above scheme, the control logic for adjusting the baffle angle relies solely on angle commands from the upper-level model, which is a passive execution. This scheme fails to utilize the physical force information generated during the interaction between the baffle and the material flow, limiting the system's ability to perceive the instantaneous impact characteristics of the material. Therefore, a torque feedback and adaptive adjustment mechanism can be introduced into the control logic of the servo motor. The servo driver operates in torque monitoring mode, acquiring the output torque value driving the servo motor deflection in real time. This torque value is fed back to the LSTM model of the dynamic scheduling decision module as a new physical feature; furthermore, when the torque value exceeds a preset impact threshold, the PLC controller can execute a preset avoidance strategy, autonomously fine-tuning the baffle angle to mitigate the impact. This scheme enables the system to perceive the physical impact characteristics of the material in real time and make proactive adaptive adjustments.

[0034] Furthermore, the process of driving the discharge port baffle to deflect via the mechanical linkage mechanism includes: A path optimization unit is configured at the slag discharge port of the screw conveyor. The path optimization unit integrates a flow guide pipe and an outlet baffle to achieve precise control over the material's falling path and height.

[0035] The above solution enables refined management of the final stage of the fine slag conveying process, suppressing dust and splashing during the conveying process from the source and improving the on-site working environment.

[0036] Furthermore, the process of transmitting the feedback data back is as follows: Video streams of the splashing process are acquired by image sensors deployed at the discharge port; image processing algorithms are used to identify and count the number of splashing particles per unit time in the video stream, forming quantified splashing data; the quantified splashing data is associated with the dynamic conveying operation scheduling parameters at the corresponding time, and packaged into structured process feedback data; the LSTM model receives the process feedback data, calculates the loss value between the predicted splashing result generated by the dynamic conveying operation scheduling parameters and the quantified splashing data, and uses the gradient descent algorithm to update the network weights of the LSTM model online to complete strategy fine-tuning.

[0037] The feedback data transmission process enables the LSTM model to continuously learn. The LSTM model can continuously adjust its strategy based on the feedback data to cope with equipment aging and achieve a high level of operational efficiency.

[0038] Furthermore, the predictive maintenance module includes: Health status monitoring unit: Collects vibration acceleration time-series data and surface temperature time-series data of the screw conveyor during operation through vibration sensors and temperature sensors; Trend prediction unit: Inputs the vibration acceleration time series data and the surface temperature time series data into a preset time series prediction algorithm model; The time series prediction algorithm model decomposes the data into trend items, seasonal items and holiday items, and predicts the wear index within a specific future time window, generating a wear state prediction report; Early warning and scheduling unit: compares the wear index in the wear status prediction report with the preset maintenance threshold; when the wear index exceeds the maintenance threshold, a maintenance early warning is generated; the maintenance early warning automatically triggers a standardized "three-cloth five-oil" repair instruction, which includes the operation steps of using resin glue and curing agent mixed in a certain proportion, combined with glass fiber cloth, and repairing the leakage point according to the three-cloth five-oil process, and pushes the instruction to the maintenance work order module in the enterprise resource planning system.

[0039] Furthermore, according to the patent specification

[0049] of publication number CN114150840B, the "three-cloth-five-oil" process specifically refers to "first applying one coat of epoxy resin primer to the surface of the mortar layer, then applying a second coat of epoxy resin primer, then attaching a first layer of glass cloth on top, then applying one coat of epoxy resin topcoat, then attaching a second layer of glass cloth on top, then applying a second coat of epoxy resin topcoat, then attaching a third layer of glass cloth on top, and finally applying a third coat of epoxy resin topcoat, followed by final acceptance." Specifically, in the implementation of this invention, when there are leakage areas in the screw conveyor trough, a "three-cloth, five-oil" composite material process can be used for on-site repair. The steps are as follows: First, the leakage or wear area is surface-treated by thoroughly grinding the area with tools such as an angle grinder to remove rust and oil stains until the metal substrate is exposed and a rough surface is formed. Then, it is wiped clean with a cleaning solvent and ensured to be completely dry. Next, epoxy resin and curing agent are accurately weighed and mixed evenly at a weight ratio of 2:1 to obtain a prepared resin, which is then laminated for repair. A layer of the prepared resin is evenly brushed onto the treated surface as a base coat (first oil). While the resin is still wet, the first layer of fiberglass cloth (first cloth) is quickly and smoothly laid on it, and immediately brushed with resin as the first impregnation layer (second oil), pressing with a scraper to ensure that the cloth is completely saturated. On the first impregnated fiberglass cloth, a second layer of fiberglass cloth (second cloth) is then laid, and similarly brushed with resin as the second impregnation layer (third oil). Subsequently, a third layer of fiberglass cloth (third cloth) is laid and coated with resin as the third impregnation layer (fourth oil). Finally, after all cloth layers have been laid and impregnated, a layer of resin is evenly applied to the entire repair area as a cover layer (fifth oil) to make the surface smooth and flat. After construction is completed, the repaired area is allowed to cure at room temperature to form a high-strength, corrosion-resistant, and completely sealed composite material repair layer.

[0040] By pushing predictive maintenance instructions to the maintenance work order module in the enterprise resource planning system, not only is the scheduling of maintenance automated, but the allocation of maintenance resources is also optimized, reducing the impact of unplanned downtime on production.

[0041] Example 2 This embodiment describes the application of the present invention in a distributed food waste treatment station deployed underground in a large urban commercial complex. The treatment station mainly processes waste from dozens of catering businesses within the complex, and its material source is relatively singular. However, the treatment station is usually unmanned or minimally staffed, thus requiring extremely high levels of long-term operational stability, remote monitoring, and automated operation and maintenance capabilities.

[0042] In this scenario, the proposed system for optimizing the conveying path of fine residue from food waste treatment based on time-series feature analysis aims to achieve efficient, clean operation and intensive remote management without human intervention. The system's process and structure are as follows: Figure 1 and Figure 2 As shown, it mainly includes: The system includes a real-time operating condition perception module, a dynamic scheduling decision module, a closed-loop control module, and a predictive maintenance module.

[0043] Real-time operating condition perception module: The edge layer performs local processing and fluctuation analysis on the real-time operating data of kitchen waste fine residue, and dynamically adjusts the acquisition configuration of the sensor network according to the data fluctuation analysis report; The cloud layer calculates and analyzes the reconstruction error of the preprocessed data through the VAE model, generates operating condition analysis feature data, and combines the preprocessed data to form a real-time enhanced feature stream.

[0044] Furthermore, due to the relatively stable properties of the materials, the process of dynamically adjusting the sensor network's acquisition configuration is primarily focused on energy saving and extending equipment lifespan. Most of the time, the data variance in the data fluctuation analysis report is below the second preset threshold, and the system operates in low-power mode, automatically reducing the sampling frequency and optimizing communication routing. Only in a few special circumstances, such as when a hot pot restaurant dumps a large amount of high-oil, high-moisture waste, will the system briefly increase the sampling frequency.

[0045] Furthermore, the process of forming the real-time enhanced feature stream is as follows: the cloud layer encodes and decodes the preprocessed data using a VAE model to obtain reconstructed data; the Euclidean distance between the preprocessed data and the reconstructed data is calculated to obtain a quantified reconstruction error value. In this scenario, a suddenly increased reconstruction error value, in addition to serving as a feature for operational condition analysis, often indicates whether a merchant has mistakenly disposed of non-food waste and can trigger a remote alarm.

[0046] Dynamic scheduling decision module: The LSTM model infers the real-time enhanced feature stream and generates dynamic delivery job scheduling parameters.

[0047] Furthermore, the generation process of the dynamic delivery scheduling parameters is as follows: by loading pre-trained LSTM model weights, receiving real-time enhanced feature streams, and performing matrix operations and activation function processing, the dynamic delivery scheduling parameters, including the target running frequency and target deflection angle, are output. In this application, the model's scheduling strategy gradually converges and stabilizes within a small optimization range, ensuring that the system can still make the most appropriate fine-tuning of the material characteristics of different restaurants, such as noodle shops and beverage bars, even without human intervention.

[0048] Closed-loop control module: The PLC controller receives the dynamic conveying operation scheduling parameters, and adjusts the operating frequency of the screw conveyor and the deflection angle of the discharge port baffle in real time by driving the frequency converter and servo motor; it quantifies the discharge port splash data and forms process feedback data to be sent back to the LSTM model.

[0049] Furthermore, the specific process for adjusting the operating frequency of the screw conveyor in real time is as follows: Figure 3As shown: The PLC controller receives and parses the dynamic conveying operation scheduling parameters to obtain the target operating frequency value; the PLC controller sends a control message to the frequency converter through the Modbus communication protocol, and the control message contains the target operating frequency value; the frequency converter receives and parses the control message, and adjusts the frequency of the three-phase AC power output to the screw conveyor drive motor so that the actual operating frequency of the drive motor matches the target operating frequency value.

[0050] Furthermore, the process of adjusting the deflection angle of the discharge port baffle in real time is as follows: the PLC controller receives and parses the dynamic conveying operation scheduling parameters to obtain the target deflection angle value; the PLC controller sends pulse sequence instructions to the servo driver, the number of pulse sequence instructions being proportional to the target deflection angle value; the servo driver drives the servo motor to rotate, and drives the discharge port baffle to deflect through the mechanical linkage mechanism; the encoder built into the servo motor feeds back the actual rotation angle to the servo driver, forming a position closed-loop control.

[0051] Furthermore, the process of transmitting feedback data is particularly important for unattended stations, as it enables the system to self-correct. By acquiring video streams of the splashing process and quantifying the splash data through image sensors deployed at the discharge port, the LSTM model can update its network weights online. This mechanism allows the system to slowly self-adjust to adapt to the slight performance degradation caused by long-term operation, ensuring the long-term stability of the system.

[0052] Furthermore, the predictive maintenance module includes: Health status monitoring unit: Collects vibration acceleration time-series data and surface temperature time-series data of the screw conveyor during operation through vibration sensors and temperature sensors; Trend prediction unit: Inputs the vibration acceleration time series data and the surface temperature time series data into a preset time series prediction algorithm model; The time series prediction algorithm model decomposes the data into trend items, seasonal items and holiday items, and predicts the wear index within a specific future time window, generating a wear state prediction report; Early warning and scheduling unit: compares the wear index in the wear status prediction report with the preset maintenance threshold; when the wear index exceeds the maintenance threshold, a maintenance early warning is generated; the maintenance early warning automatically triggers a standardized "three-cloth five-oil" repair instruction, which includes the operation steps of using resin glue and curing agent mixed in a certain proportion, combined with glass fiber cloth, and repairing the leakage point according to the "three-cloth five-oil" process, and pushes the instruction to the maintenance work order module in the enterprise resource planning system.

[0053] Specifically, when there are leaks in the screw conveyor trough, a "three-layer cloth and five-layer oil" composite material process can be used for on-site repair. The steps are as follows: First, the leaking or worn area is surface-treated by thoroughly grinding the area with tools such as an angle grinder to remove rust and oil until the metal substrate is exposed and a rough surface is formed. Then, it is wiped clean with a cleaning solvent and ensured to be completely dry. Next, epoxy resin and curing agent are precisely weighed and mixed evenly at a weight ratio of 2:1 to obtain a prepared resin, which is then laminated for repair. A layer of the prepared resin is evenly applied to the treated surface as a base coat (first oil). While the resin is still wet, the first layer of fiberglass cloth (first cloth) is quickly and smoothly laid on it, and immediately coated with resin as the first impregnation layer (second oil), pressing with a scraper to ensure that the cloth is completely saturated. On top of the first impregnated fiberglass cloth, a second layer of fiberglass cloth (second cloth) is then laid, and similarly coated with resin as the second impregnation layer (third oil). Subsequently, a third layer of fiberglass cloth (third cloth) is laid and coated with resin as the third impregnation layer (fourth oil). Finally, after all cloth layers have been laid and impregnated, a layer of resin is evenly applied to the entire repair area as a cover layer (fifth oil) to make the surface smooth and flat. After construction is completed, the repaired area is allowed to cure at room temperature to form a high-strength, corrosion-resistant, and completely sealed composite material repair layer.

[0054] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A system for optimizing the conveying path of fine residue in food waste treatment based on time-series feature analysis, characterized in that, include: The real-time operating condition sensing module at the edge layer performs local processing and fluctuation analysis on the real-time operating data of kitchen waste fine residue, and dynamically adjusts the acquisition configuration of the sensor network based on the data fluctuation analysis report. The cloud layer calculates and analyzes the reconstruction error of the preprocessed data through the VAE model, generates working condition analysis feature data, and combines the preprocessed data to form a real-time enhanced feature stream; The dynamic scheduling decision module uses an LSTM model to infer the real-time enhanced feature stream and generate dynamic delivery job scheduling parameters. The closed-loop control module, PLC controller receives the dynamic conveying operation scheduling parameters, and adjusts the operating frequency of the screw conveyor and the deflection angle of the discharge port baffle in real time by driving the frequency converter and servo motor; quantifies the discharge port splash data, forms process feedback data and sends it back to the LSTM model; The predictive maintenance module uses a time series prediction algorithm to predict the wear status of the bottom of the screw conveyor, generate maintenance warnings, and trigger standardized "three-cloth five-oil" repair instructions.

2. The food waste fine residue conveying path optimization system based on time-series feature analysis according to claim 1, characterized in that, The process of dynamically adjusting the data acquisition configuration of the sensor network is as follows: The sensor network's acquisition configuration is dynamically adjusted based on the data fluctuation analysis report. When the data variance in the data fluctuation analysis report exceeds a first preset threshold, the sampling frequency of the sensor network is increased; when the data variance is lower than a second preset threshold, the sampling frequency of the sensor network is decreased and the communication routing between wireless sensor nodes is adjusted.

3. The food waste fine residue conveying path optimization system based on time-series feature analysis according to claim 1, characterized in that, The process of adjusting the operating frequency of the screw conveyor in real time is as follows: The PLC controller receives and parses the dynamic conveying operation scheduling parameters to obtain the target operating frequency value. The PLC controller sends a control message to the frequency converter via the Modbus communication protocol. The control message contains the target operating frequency value. The frequency converter receives and parses the control message and adjusts the frequency of the three-phase AC power output to the screw conveyor drive motor so that the actual operating frequency of the drive motor matches the target operating frequency value.

4. The food waste fine residue conveying path optimization system based on time-series feature analysis according to claim 1, characterized in that, The process of adjusting the deflection angle of the discharge port baffle in real time is as follows: The PLC controller receives and parses the dynamic conveying operation scheduling parameters to obtain the target deflection angle value; the PLC controller sends pulse sequence instructions to the servo driver, the number of which is proportional to the target deflection angle value; the servo driver drives the servo motor to rotate, and drives the discharge port baffle to deflect through the mechanical linkage mechanism; the encoder built into the servo motor feeds back the actual rotation angle to the servo driver, forming a position closed-loop control.

5. The food waste fine residue conveying path optimization system based on time-series feature analysis according to claim 4, characterized in that, The process of driving the discharge port baffle to deflect via the mechanical linkage mechanism includes: A path optimization unit is configured at the slag discharge port of the screw conveyor. The path optimization unit integrates a flow guide pipe and an outlet baffle to achieve precise control over the material's falling path and height.

6. The food waste fine residue conveying path optimization system based on time-series feature analysis according to claim 1, characterized in that, The process of transmitting the feedback data back is as follows: Video streams of the splashing process are acquired by image sensors deployed at the discharge port; image processing algorithms are used to identify and count the number of splashing particles per unit time in the video stream, forming quantified splashing data; the quantified splashing data is associated with the dynamic conveying operation scheduling parameters at the corresponding time, and packaged into structured process feedback data; the LSTM model receives the process feedback data, calculates the loss value between the predicted splashing result generated by the dynamic conveying operation scheduling parameters and the quantified splashing data, and uses the gradient descent algorithm to update the network weights of the LSTM model online to complete strategy fine-tuning.

7. The food waste fine residue conveying path optimization system based on time-series feature analysis according to claim 1, characterized in that, The predictive maintenance module includes: Health status monitoring unit, trend prediction unit, and early warning and dispatch unit; Health status monitoring unit: Collects vibration acceleration time-series data and surface temperature time-series data of the screw conveyor during operation through vibration sensors and temperature sensors; Trend prediction unit: Inputs the vibration acceleration time series data and the surface temperature time series data into a preset time series prediction algorithm model; The time series prediction algorithm model decomposes the data into trend items, seasonal items and holiday items, and predicts the wear index within a specific future time window, generating a wear state prediction report; Early warning and scheduling unit: compares the wear index in the wear status prediction report with the preset maintenance threshold; when the wear index exceeds the maintenance threshold, a maintenance early warning is generated; the maintenance early warning automatically triggers a standardized "three-cloth five-oil" repair instruction, which includes the operation steps of using resin glue and curing agent mixed in proportion, combined with glass fiber cloth, and repairing the leakage point according to the "three-cloth five-oil" process, and pushes the instruction to the maintenance work order module in the enterprise resource planning system.

Images