A digital management system for recovering resources from linoleic acid waste residues

Abstract

The application discloses a kind of linoleic acid waste residue recycling resource scheduling digital management system, it is related to industrial waste residue recycling processing technical field, its technical points are: the physical correlation of waste residue component and equipment multi-order vibration is established, the vibration characteristic spectrum of its processing whole process is mapped by extracting the multidimensional component characteristics of each batch of waste residue, the technical problem that waste residue is regarded as homogenization material when scheduling, vibration risk cannot be predicted is solved;Afterwards, based on the cognition of the vibration characteristics of each batch, instead of passively suffering from the damage of equipment caused by vibration superposition, the core target of full-band vibration energy cancellation is taken, the phase conjugate of each order vibration mode between different batches is calculated, the low resonance batch combination sequence is constructed, and the vibration energy that originally enhances each other is mutually cancelled in the equipment. From the beginning of scheduling decision, prevent the generation of non-steady-state vibration.

Description

Technical Field

[0001] This invention relates to the field of industrial waste recycling and treatment technology, and in particular to a digital management system for resource scheduling of linoleic acid waste recycling. Background Technology

[0002] Linoleic acid is an essential unsaturated fatty acid for the human body, widely used in functional foods, pharmaceutical preparations, and cosmetic raw materials, with a global annual demand exceeding 3 million tons. Linoleic acid is mainly produced through vegetable oil refining processes. The production of 1 ton of linoleic acid generates approximately 0.3 to 0.5 tons of industrial waste residue, including soapstock, waste bleaching clay, distillation residue, deodorized distillate, and hydrolysis residue. These waste residues contain 15% to 30% recyclable crude oil, which has high resource value. However, improper handling can cause soil and water pollution. Therefore, the industry generally needs to carry out efficient recycling and treatment of waste residues.

[0003] Currently, the digital scheduling systems in the linoleic acid waste residue recycling industry all adopt a homogenized approach, with the following specific steps: First, basic resource data such as the amount of waste generated and the storage location of each production process, as well as the load capacity of transport vehicles, storage capacity, and rated processing capacity of processing equipment are collected. Then, with the optimization goals of the lowest transportation cost, the fastest storage turnover, and the highest equipment utilization rate, resource allocation schemes are generated through traditional scheduling algorithms such as genetic algorithms and particle swarm algorithms to clarify the transportation time, storage location, and processing equipment for each batch of waste. Finally, scheduling instructions are sent to each execution terminal to monitor the task execution progress and equipment operating status in real time. When equipment failure occurs, an alarm is triggered and manual intervention is performed.

[0004] However, during the implementation of the above technical solution, at least the following technical problems were discovered: Firstly, existing scheduling methods treat all linoleic acid waste as homogeneous materials to be processed, basically only collecting information on the amount and location of waste generated, without considering the differences in composition between different batches of waste. However, in actual use, differences in composition parameters such as crude oil content, solid impurity ratio, and moisture content will cause the waste to generate drastically different vibration excitations on the equipment during the processing. As a result, optimization is only focused on resource efficiency and cost objectives, without considering the cumulative effect of vibration excitations from different batches of waste. This leads to the continuous accumulation of vibration energy when multiple batches of waste are processed continuously, and the equipment is in a non-steady vibration state for a long time. Secondly, the management of equipment vibration problems is always done in a reactive mode. When the equipment vibration exceeds the safety threshold, the vibration can only be passively attenuated by hardware modifications such as shutdown for maintenance, installation of rubber damping pads, and pouring concrete to reinforce the foundation. This method cannot address the root cause of the vibration, namely the phase superposition of vibration excitation from different batches of waste residue. Not only are the modification costs high and the construction period long, but it can only reduce the vibration amplitude and cannot fundamentally eliminate the cumulative damage of vibration to the equipment. Therefore, we propose a digital management system for the resource scheduling of linoleic acid waste residue recycling. Summary of the Invention

[0005] (a) Technical problems to be solved To address the shortcomings of existing technologies, this invention provides a digital management system for resource scheduling in linoleic acid waste residue recycling, which solves the technical problem of equipment damage caused by vibration in existing linoleic acid waste residue recycling technologies.

[0006] (II) Technical Solution To achieve the above objectives, the present invention provides the following technical solution: A digital management system for resource scheduling in the recycling of linoleic acid waste residue, the management system comprising: The basic data acquisition module is used to acquire waste residue generation data, basic data of recycled resources, production plan data, and basic parameters of multi-order natural vibration modes of the processing equipment throughout the entire linoleic acid production process. The feature mapping module is used to extract the multidimensional component feature vector of each batch of linoleic acid waste residue, establish the physical relationship between the component features and the vibration excitation of each order of the equipment, and map the multidimensional component feature vector to the multi-order natural vibration feature spectrum of the entire process of the batch of waste residue treatment. The phase conjugate matching scheduling module is used to calculate the phase conjugate degree of each order of the multi-order natural vibration characteristic spectrum of the batch to be processed, and constructs a low resonance batch combination sequence with the goal of maximizing the vibration energy cancellation rate across the entire frequency band. The real-time phase modulation module is used to calculate the phase offset of each vibration mode based on the high-frequency real-time vibration data of the processing equipment, and to perform phase compensation of each vibration mode by dynamically adjusting the particle size of the feeding batch and the feeding time interval. The scheduling scheme generation module is used to generate an initial resource scheduling scheme by combining basic data on recycled resources, forecast data on waste residue recycling demand, and low-resonance batch combination sequences. The constraint verification and execution module is used to perform multi-dimensional constraint verification on the initial resource scheduling plan, including equipment load, transportation capacity, storage capacity, and environmental emissions. It outputs an optimized resource scheduling plan and converts it into execution instructions for distribution. The real-time monitoring module is used to synchronously collect the operating status of each execution terminal, process multi-level vibration data, real-time load data and temperature data of the equipment; The effect evaluation module is used to generate a multi-mode comprehensive cancellation rate based on the real-time vibration data of the processing equipment, and to generate a comprehensive evaluation value of resource scheduling effect by combining the scheduling execution data. Then, based on the multi-mode comprehensive cancellation rate, waste generation data and equipment operating data, the low resonance batch combination sequence and phase modulation parameters are adjusted synchronously.

[0007] Preferably, the basic data acquisition module specifically acquires the following: The production workshop of linoleic acid produces five types of waste residues: soap residue, waste clay, distillation residue, deodorization distillate, and hydrolysis residue. These waste residues include the amount, time of generation, storage location, and batch number of each type. The workshop also includes the reaction temperature, pressure, feed rate, catalyst addition, reaction time, raw material origin, raw material acid value, and moisture content of each corresponding process. The load capacity, mileage, maintenance status, average speed, loading and unloading time and leakage prevention level of transport vehicles; the storage capacity, location, temperature and humidity control capabilities, remaining capacity, leakage prevention level and ventilation capacity of storage facilities; the rated processing capacity, continuous load, operating time, maintenance cycle, historical vibration and fault data of processing equipment; the minimum feeding interval of automatic feeding system; and the job position, skill level, working hours and operational proficiency of operators. Enterprise medium- and long-term production plans and temporary production adjustment data, including daily product output, process scheduling, raw material batch switching, equipment maintenance plans, as well as emergency orders, raw material quality abnormalities, production failures, environmental inspections, and extreme weather warnings; The first to fifth natural vibration frequencies, modal damping ratios, mode shapes, safe acceleration thresholds, and energy percentages of the processing equipment were obtained through hammer impact modal testing.

[0008] Preferably, the execution steps of the feature mapping module are as follows: The rapid detection data of each batch of waste residue is integrated with the process parameters of the corresponding production process to generate an eight-dimensional component feature vector containing crude oil content, free fatty acid ratio, solid impurity content, saponification content, moisture content, ash content, pigment content, and phospholipid content, and a unique 32-bit hexadecimal component code is generated for each batch. Establish the physical correspondence between component characteristics and equipment vibration excitation. Among them, crude oil and free fatty acids mainly affect the first-order vibration amplitude, solid impurities and ash mainly affect the second-order vibration amplitude, moisture and saponification mainly affect the third-order vibration amplitude, and pigments and phospholipids mainly affect the fourth and fifth-order vibration amplitudes. Based on historical data from valid batches, a component fingerprint-multi-order vibration feature database is constructed, and a multi-order vibration prediction model is trained and generated. The eight-dimensional component feature vector of the batch to be processed is input into the model to generate the multi-order natural vibration feature spectrum of the batch on the corresponding processing equipment from feeding to processing end, including the frequency, amplitude, initial phase, duration and energy ratio of each order of vibration.

[0009] Preferably, the data in the component fingerprint-multi-order vibration feature database are divided into a training set, a validation set, and a test set in a ratio of 8:1:1; Empirical mode decomposition is performed on the original vibration data to separate the first to fifth order independent natural mode functions and filter out noise components; Using an eight-dimensional component feature vector as input and the frequency, amplitude, and initial phase of each order of intrinsic mode function as output, gradient boosting sub-models of orders one to five are trained respectively. The outputs of each sub-model are fused to generate a complete multi-order natural vibration characteristic spectrum. The error is corrected by test set data so that the amplitude prediction error of each vibration order does not exceed 8% and the phase prediction error does not exceed 15°.

[0010] Preferably, the execution steps of the phase conjugate matching scheduling module are as follows: The time axis of the multi-order natural vibration characteristic spectrum of all batches to be processed is standardized to be the longest processing time among all batches. Calculate the phase conjugate degree of each vibration mode between any two batches to be processed, where the phase conjugate degree = 1 - |phase difference - 180 degrees| / 180 degrees, and the value ranges from 0 to 1; Based on the energy proportion weight of each order of vibration, the overall cancellation rate of the full-frequency vibration of the two batches is calculated, where the overall cancellation rate of the full-frequency vibration is Σ (phase conjugation degree from the first to the fifth order × energy proportion of each order). Batch combinations with a comprehensive vibration cancellation rate of not less than 85% across the entire frequency band were selected and arranged in order of cancellation rate from high to low and total processing time from short to long to generate a low resonance batch combination sequence.

[0011] Preferably, the phase conjugate matching scheduling module further includes: If the cancellation rate of two batches does not reach 85%, the third and fourth batches are introduced in sequence for multi-stage phase matching until the cancellation rate reaches the target. The energy proportion weight of each vibration is dynamically adjusted based on the historical fault data of the processing equipment. For high-order vibrations that cause more than 3 faults, their weight is increased by 50%. When a single batch belongs to multiple qualified combinations, the combination with the highest offset rate and lowest energy consumption is selected first; if the energy consumption is the same, the combination with the shortest total processing time is selected. The total throughput of each low-resonance batch combination is controlled between 90% and 110% of the equipment's rated continuous load.

[0012] Preferably, the execution steps of the real-time phase modulation module are as follows: Empirical mode decomposition is performed on the real-time vibration data of the processing equipment to extract the first to fifth order actual vibration signals of the current batch combination and filter out environmental noise. The actual vibration signal is compared with the predicted vibration signal to calculate the phase shift and amplitude deviation of each order of vibration. The next complete batch to be processed is split into multiple micro-batches of equal weight, wherein the weight of a single micro-batch does not exceed 5% of the complete batch and is not less than the minimum feeding amount of the automatic feeding system. The feeding time interval for each micro-batch is calculated based on the phase shift and amplitude deviation of each vibration.

[0013] Preferably, the real-time phase modulation module further includes: The nonlinear coupling effect between vibration modes of each order is calculated, a coupling coefficient matrix is ​​established, and the time interval of micro-batch feeding is corrected, with the correction amplitude not exceeding 20% ​​of the original modulation amount; The feeding interval between adjacent micro-batches shall not be less than the minimum interval of the automatic feeding system, and shall not be greater than 1 / 20 of the processing time of a complete batch. After determining the feeding interval, pre-evaluate the offset rate after modulation. If it does not reach 85% or the energy consumption increases by more than 5%, readjust the number of micro-batch splits and the feeding interval. When any vibration acceleration exceeds 120% of the safety threshold of that order, feeding is immediately suspended and an audible and visual alarm is triggered. Once the vibration returns to a safe range, the modulation parameters are recalculated and feeding is resumed.

[0014] Preferably, the execution steps of the effect evaluation module are as follows: The cancellation rate of vibration modes from first to fifth order, the comprehensive cancellation rate of vibration across the entire frequency band, the effective value of equipment vibration acceleration, and the effective value of vibration velocity are calculated based on real-time vibration data. Among them, the cancellation rate of each order mode = 1 - actual vibration amplitude / predicted vibration amplitude. Calculate the average load rate, load fluctuation rate, and load peak-to-valley difference based on real-time load data of the equipment; The on-time completion rate of recycling tasks, average utilization rate of transport vehicles, average turnover rate of storage facilities, comprehensive energy consumption per unit of waste residue recycling, and cost per unit of waste residue treatment are calculated based on the scheduling and execution data. Determine the weight coefficients for each evaluation indicator, among which the weight of the comprehensive vibration cancellation rate across the entire frequency band is not less than 30%, the weight of the effective value of equipment vibration acceleration is not less than 20%, and the sum of all weight coefficients is 1; After normalizing the actual values ​​of each indicator, multiply them by their corresponding weights and sum them to generate a comprehensive evaluation value for resource scheduling effectiveness.

[0015] Preferably, the management system also includes a data management module, and the specific execution steps are as follows: Encrypt and store all historical waste generation data, compositional data, vibration characteristic spectrum data, batch combination data, phase modulation parameters, scheduling schemes, execution status, equipment operating conditions, and evaluation data. By using waste residue batch number, component code, transport vehicle number or processing equipment number, the entire process information of waste residue from generation, component detection, vibration prediction, phase matching, micro-batch modulation, transportation, storage to processing can be traced, as well as the vibration and operating condition data of the corresponding equipment at all times. Generate daily, monthly, and annual statistical reports, including total recovery volume, resource utilization rate, cost analysis, energy consumption analysis, vibration mitigation effect, equipment health status, and statistics on abnormal events; Each quarter, incremental training is performed on the component-vibration mapping model, phase matching model, and phase modulation model based on newly added historical data, ensuring that the prediction error of the updated model is not higher than that of the original model. Based on historical vibration data and vibration mitigation effects, a model for predicting the remaining life of equipment is established to generate preventative maintenance reminders.

[0016] (III) Beneficial Effects 1. Establish the physical correlation between waste residue composition and multi-order vibration of equipment. By extracting the multi-dimensional composition characteristics of each batch of waste residue, the vibration characteristic spectrum of its entire processing process is mapped, solving the technical problem of treating waste residue as homogeneous material and being unable to predict vibration risks during scheduling. Then, based on the understanding of the vibration characteristics of each batch, the equipment is no longer passively subjected to the damage caused by the superposition of vibrations. Instead, the core objective is to cancel the vibration energy across the entire frequency band. By calculating the phase conjugation degree of each vibration mode between different batches, a low-resonance batch combination sequence is constructed, so that the vibration energy that would originally reinforce each other cancels each other out inside the equipment, preventing the generation of unsteady vibrations from the scheduling decision.

[0017] 2. By employing micro-batch splitting and phase modulation, a complete batch is divided into multiple independently controllable feeding units. The feeding time interval of each micro-batch is dynamically adjusted to compensate for phase deviations in vibration at different levels in real time, forming a mutual adjustment mechanism. No modifications to existing equipment are required, thus enabling active suppression of vibration across the entire frequency band without affecting normal production. This method keeps the equipment in a stable operating state for extended periods, reducing cumulative fatigue damage to core components such as bearings, gears, and agitators. It also lowers the frequency of sudden equipment failures, substantially extends the effective service life, stabilizes process parameters in waste residue treatment, and ensures more complete extraction of recyclable resources.

[0018] 3. By constructing a fully digitalized process from waste generation to final treatment, all data in all stages are fully encrypted, stored, and traceable. Based on long-term accumulated vibration and operational data, the core prediction model can be continuously incrementally trained to improve the accuracy of vibration prediction and mitigation. At the same time, the system can predict the changing trend of equipment health status in advance based on historical vibration data and mitigation effects, transforming traditional post-fault maintenance into pre-emptive preventive maintenance, further reducing the risk of production interruption. Attached Figure Description

[0019] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, the preferred embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0020] Figure 1 This is an overall structural diagram of the system in an embodiment of the present invention; Figure 2 This is a system flowchart in an embodiment of the present invention. Detailed Implementation

[0021] This application provides a digital management system for resource scheduling in the recycling of linoleic acid waste residue, which solves the technical problem of equipment damage caused by vibration in existing linoleic acid waste residue recycling technologies.

[0022] like Figure 2 As shown, taking the actual implementation process of a linoleic acid production enterprise as an example, it uses two continuous waste residue treatment devices of the same model and a scheduling mode based on the waste residue generation time. The main failure modes affected by equipment use are bearing burnout and agitator fatigue fracture. Traditional mechanical repair requires equipment shutdown and has a long repair time, which will cause a lot of production losses for the enterprise. Based on this, we propose a digital management system for linoleic acid waste residue recycling resource scheduling to actively offset vibration at the scheduling level. Since the vibration amplitude of the equipment is positively correlated with the operating energy consumption, it is also necessary to consider a two-way coupling optimization mechanism of vibration and energy consumption to achieve a synergistic balance between vibration reduction and energy saving.

[0023] Example 1: I. Calibration of inherent equipment characteristics and construction of basic dataset: The hammer impact modal testing method is an existing engineering testing method for obtaining the inherent vibration characteristics of equipment. It mainly involves using a hammer to strike the equipment structure to induce free vibration, and then collecting the response signals to calculate parameters. During testing, a dynamic signal analyzer was used, and piezoelectric accelerometers were placed at eight measuring points: the bearing housing, the frame beam, the drive motor end cover, and the middle of the stirring shaft. Five impacts were made in each of the X, Y, and Z orthogonal directions, and the average response was used to calculate the modal parameters. The bearing housing, frame beam, motor end cover, and stirring shaft were chosen as measuring points because these locations are the most concentrated areas of vibration and the most prone to damage. Furthermore, averaging multiple impacts in multiple directions eliminates single-test errors and ensures parameter accuracy. This method yields the basic parameters of the equipment's first to fifth order inherent vibration modes. Since two processing devices were used, and the same model of equipment may exhibit parameter differences due to maintenance and wear, sharing a model would lead to prediction bias. Independent modeling is necessary to match the true characteristics of each individual device. Therefore, independent vibration prediction models were established for each device, and universal parameters were not used; thus, targeted model construction was performed. The fundamental parameters of the first to fifth natural vibration modes, which are the core vibration characteristics of the equipment, include five dimensions: natural vibration frequency, modal damping ratio, mode shape, safe acceleration threshold, and the proportion of vibration energy at each order. The reason for collecting these data is that the natural vibration frequency is the frequency at which the equipment resonates; the modal damping ratio reflects the rate of vibration decay; the mode shape is the relative displacement pattern of each point during equipment vibration; the safe acceleration threshold is the maximum vibration acceleration for long-term safe operation of the equipment; and the proportion of vibration energy at each order is the proportion of energy carried by different orders of vibration to the total vibration energy. These are all key information reflecting the vibration of the equipment.

[0024] Export the entire production process data. Since there is invalid data in the data, it is necessary to clean the data to remove abnormal records such as sensor fault zero values, invalid data of equipment shutdown, and more than 3 missing component detections. Each valid record includes waste residue component detection data, process parameters of the corresponding production process, vibration data of the equipment sampled for 10ms throughout the process, crude oil recovery rate after treatment, real-time energy consumption data of the equipment, and fault records.

[0025] Once usable data is obtained, a multi-order vibration prediction model can be trained. This model is used to predict the vibration characteristics of equipment during the processing of waste residue based on its composition. During training, the dataset is divided into training, validation, and test sets in an 8:1:1 ratio. To address the nonlinear and non-stationary characteristics of industrial vibration signals, empirical mode decomposition (EMD) is used to decompose the original vibration signal into five intrinsic mode functions (IMFs), filtering out 50Hz power frequency, 100Hz harmonics, and random noise components. EMD is used because it can decompose complex non-stationary signals into a superposition of multiple simple intrinsic mode functions, which facilitates subsequent analysis. Gradient Boosting Tree (GBDT) training is performed in stages, with each stage corresponding to an independent sub-model. The maximum tree depth is 6, the learning rate is 0.1, and the number of iterations is 100. The sub-models output the frequency, amplitude, and initial phase of each vibration stage. Initially, a backpropagation neural network was used for single-model training with 3 hidden layers. After 1000 iterations, the validation set showed that its generalization ability was insufficient. The reason for choosing gradient boosting trees is that they improve prediction accuracy through iterative training of multiple decision trees. Furthermore, physical constraint regularization terms are added during training to ensure that the output strictly fits the inherent modal range of the equipment. When constructing the GBDT model, the waste residue composition and process characteristics are first screened and normalized, and then grouped and assigned values ​​according to physical correlation. Then, independent GBDT sub-models are constructed for vibrations from the first to the fifth order, forming a five-in-one sub-model cluster. Subsequently, the inherent vibration parameters of the equipment are added to the loss function as physical constraints, and training is carried out with a unified hyperparameter of a maximum depth of 6, a learning rate of 0.1, and 100 iterations. After training, error verification is performed on each sub-model.

[0026] The multi-order vibration prediction model is constructed, and the specific construction process is as follows: A dedicated training dataset was constructed, with each sample containing eight-dimensional components of waste residue, process parameters of five steps, inherent vibration baselines of the equipment, vibration time series labels for each order, and energy consumption auxiliary labels, divided in an 8:1:1 ratio. Empirical mode decomposition and noise removal were then performed on the original vibration signals, unifying the time series format and correcting outliers. Subsequently, the input features were normalized, and feature grouping and weight allocation were completed according to the physical correlation between components and vibrations. First- to fifth-order vibration sub-models were then trained, using components and process features as input and corresponding vibration parameters as output. The outputs of the five sub-models were then concatenated into a complete vibration feature spectrum. A test set was used to verify amplitude and phase errors, ensuring that the amplitude prediction error for each order of vibration did not exceed 8% and the phase prediction error did not exceed 15°. If these requirements were not met, the hyperparameters were adjusted retrospectively. Finally, a quarterly incremental update mechanism was established, updating only leaf nodes and performing error verification to ensure long-term stable operation of the model.

[0027] The system interfaces with the enterprise's existing MES, WMS, TMS, and equipment PLC systems via the OPC UA protocol. This is an existing industrial communication standard, enabling interoperability and secure data transmission between devices from different manufacturers without requiring modifications to the existing hardware architecture. The vibration sensor sampling rate was ultimately determined to be 10ms, and data transmission uses the AES-128 encryption algorithm. The collected data is divided into three categories, as detailed below: Firstly, the waste residue production data covers five processes: saponification, decolorization, distillation, deodorization, and hydrolysis. It collects the production amount, production time, storage tank location, and unique batch number of soapstock, waste clay, distillation residue, deodorization distillate, and hydrolysis residue. Simultaneously, it collects 17 parameters for the corresponding process, including reaction temperature, pressure, feed rate, catalyst addition, reaction time, raw material origin, raw material acid value, and moisture content. However, due to the large error in vibration prediction caused by the composition differences between different batches, near-infrared spectroscopy rapid detection and process parameter collection are added. Secondly, it collects resource data, including the load, real-time location, mileage, maintenance status, average speed, and loading / unloading time of transport vehicles; the remaining capacity, temperature and humidity, ventilation status, and leakage detection data of storage facilities; the operating time, real-time load, historical fault records, minimum 10-second feeding interval, and real-time power of processing equipment; and the job position, skill level, and shift schedule of operators, totaling 22 parameters. Thirdly, planning and environmental data: collecting data on daily product output, process scheduling, raw material batch switching plans for the next 30 days, and eight parameters including equipment maintenance, emergency orders, raw material quality anomalies, environmental inspections, and extreme weather warnings within 72 hours.

[0028] All collected data are preprocessed at the edge gateway to avoid outliers, specifically by using the 3σ criterion to remove outliers exceeding the mean ± 3 standard deviations. Data with a missing rate ≤ 30% is completed using Kriging interpolation based on causal relationships; data with a missing rate > 30% is completed using the mean of the previous and next frames and an anomaly log is recorded. Data from different sampling frequencies are unified to a 10ms time axis through cubic spline interpolation to achieve time alignment. The Kriging interpolation method used here is an improved version of the existing interpolation method, which uses real-time values ​​of other parameters that are causally related to the missing parameters as auxiliary variables on the basis of traditional Kriging interpolation.

[0029] II. Multi-dimensional feature mapping: After each batch of waste residue is generated, near-infrared spectroscopy results are rapidly obtained and fused with the corresponding process parameters to generate an eight-dimensional component feature vector, which is the core feature set of the waste residue. This vector includes eight dimensions: crude oil content, free fatty acid ratio, solid impurity content, saponification content, moisture content, ash content, pigment content, and phospholipid content. These eight components are collected because they have a significant impact on equipment vibration and can be adjusted as needed and in accordance with actual conditions. In addition, for easy traceability, a unique 32-bit hexadecimal component code is generated for each batch, which is a full-process traceability identifier and is stored in a MySQL 8.0 database bound to the production batch number. This code allows for the traceability of the entire process of waste residue from generation to treatment.

[0030] A component-vibration physical correlation was established. Crude oil and free fatty acids mainly affected the amplitude of the first-order vibration, solid impurities and ash affected the second-order vibration, moisture and saponification affected the third-order vibration, and pigments and phospholipids affected the fourth and fifth-order vibrations. The eight-dimensional feature vector was input into the pre-trained GBDT model, which output the multi-order intrinsic vibration feature spectrum of the batch sampled for 10ms from feeding to the end of processing. It includes time series data of five dimensions: frequency, amplitude, initial phase, duration and energy ratio of each vibration. At the same time, the model outputs the predicted energy consumption curve of the batch processing for subsequent multi-objective optimization.

[0031] After obtaining the vibration characteristic spectra of all batches to be processed that day, time axis standardization is first performed, stretching the characteristic spectra of different processing times to a maximum of 4 hours to ensure consistency in phase difference calculation; the core phase conjugate degree calculation initially uses the cosine similarity formula cos(Δφ) ij,k Phase conjugation is a quantitative index of vibration complementarity, used to measure the degree to which the phases of two batches of vibration signals cancel each other out. However, when the phase difference is close to 180°, the function gradient is steep, and small phase errors can cause large fluctuations in the results. Furthermore, negative values ​​are not conducive to subsequent weighted summation. An improved linear difference formula is proposed: C ij,k =1-(|Δφ ij,k -180°|) / 180°,where C ij,k Let Δφ be the phase conjugate of the k-th order vibrations of batch i and batch j. ij,k C represents the phase difference between the k-th order vibrations of the two batches, ranging from 0° to 360°. ij,k ∈[0, 1], the closer the value is to 1, the closer the phase is to 180°, and the better the cancellation effect; in the implementation process, first extract the phase value at the same moment in the vibration characteristic spectrum of the two batches, calculate the phase difference time by time, and then take the average value to obtain the overall phase difference of the vibration of that order. Substitute it into the formula to calculate the phase conjugate degree. For example, the average value of the phase difference time by time between the first-order vibrations of batches A and B is 172°, and substituting it into the formula gives C ij,1 =1-|172-180| / 180≈0.956, the average second-order phase difference is 165°, so C ij,2 ≈0.917. Similarly, the values ​​for the third to fifth orders are 0.883, 0.852, and 0.821, respectively.

[0032] The overall cancellation rate across the entire frequency band, i.e., the comprehensive evaluation index of the batch combined vibration cancellation effect, is calculated using energy weighting: In the formula, R ij W represents the overall cancellation rate across the entire frequency band for combinations of batches i and j, where n is the number of batches. k The energy proportion weight of the k-th order vibration, ∑W k=1; The energy proportion of each vibration order of the equipment is obtained by hammer impact modal testing as the initial weight, and then dynamically adjusted according to the equipment's historical fault data: if a certain vibration order has caused ≥3 faults in the past, its weight is increased by 50%; for example, the second-order vibration of equipment No. 2 has caused 3 bearing faults, and the weight is adjusted from the initial 25% to 37.5%; among them, because the energy proportion of the first-order vibration is relatively large and causes the most damage to the equipment, after weighting according to the actual energy proportion, the equipment failure rate is further reduced by 12%, which can be adjusted according to the actual situation.

[0033] A vibration-energy consumption multi-objective optimization function, namely a batch combination comprehensive evaluation function, is introduced to perform synergistic optimization of vibration reduction and energy saving. The comprehensive score = 0.7 × comprehensive offset rate + 0.3 × (1 - energy consumption ratio), where the energy consumption ratio is the ratio of the total energy consumption of the combined batch to the total energy consumption of the two batches processed individually. Here, the weight coefficients are set to 0.7 and 0.3, which can be adjusted according to the actual situation. For example, if the comprehensive offset rate of batches A and B is 91.2% and the energy consumption ratio is 0.92, the comprehensive score = 0.7 × 0.912 + 0.3 × (1 - 0.92) = 0.6624, which is better than another combination with an offset rate of 90% but an energy consumption ratio of 1.0, whose score is 0.63.

[0034] If the offset rate of two batches does not reach the 85% threshold, the third and fourth batches are introduced in sequence for matching. For example, if the offset rate of batches C and D is 82%, after adding batch E, the vibration waveforms of the three batches are superimposed according to the feeding order. After calculating the total vibration amplitude, it is compared with the average amplitude of a single batch to obtain an offset rate of 87% and an energy consumption ratio of 0.94. The combination of these three batches is then adopted.

[0035] During use, it was discovered that each batch has a chaotic sensitive area, which is a critical range for vibration characteristics. This refers to the time interval during which a ±5% fluctuation in parameters during the curing process will cause a ≥20% change in vibration amplitude. The phase complementarity within the sensitive area contributes up to 60% to the overall cancellation effect. Therefore, a priority matching strategy for chaotic sensitive areas is implemented, as follows: Parametric perturbation analysis is performed on the vibration characteristic spectrum of each batch. The parameter perturbation is applied at a step size of 10ms with ±5% every hour. The rate of change of vibration amplitude before and after the perturbation is calculated. The time interval with a rate of change ≥20% is marked as the chaotic sensitive region. Then, batches with overlapping time in the sensitive region are matched first, so that the vibrations in the sensitive region cancel each other out, which can improve the overall cancellation rate.

[0036] The final batch combination must meet the requirement that the total processing capacity is 90% to 110% of the equipment's rated load to avoid overload or underload. The batch combination sequence of the day is generated by sorting the comprehensive scores from high to low and the processing time from short to long. This is the optimal processing sequence, which refers to the batch sequence after combination with a comprehensive vibration cancellation rate of more than 85% across the entire frequency band. Here, low resonance refers to an engineering safety state, that is, the vibration amplitude is reduced to below the equipment's safety threshold, rather than absolute zero vibration in a physical sense.

[0037] III. Scheduling Scheme Generation and Verification: Based on the low-resonance batch sequence, combined with the status of recycled resources and 72-hour demand forecast, an initial scheduling plan is generated, which clarifies the execution time of each task, transport vehicles, driving routes, storage tank locations, allocation of processing equipment, and feeding process accurate to 100 milliseconds.

[0038] To improve the validity of the scheduling scheme, a five-dimensional constraint verification is performed on the initial scheme, as follows: (1) Transportation constraints: vehicle load ≤ rated load, daily driving time ≤ 8 hours, and driving routes should avoid residential areas and environmentally sensitive areas; (2) Storage constraints: the amount of waste residue entering the warehouse is less than or equal to the remaining storage capacity; different types of waste residue are stored in separate areas; and the temperature and humidity meet the storage requirements. (3) Equipment constraints: the real-time load of the processing equipment is ≤110% of the rated load, the continuous operation time is ≤24 hours, and the equipment maintenance window is avoided; (4) Personnel constraints: The skill level of the operators matches the job requirements, and the working hours comply with the provisions of the Labor Law; (5) Environmental constraints: There should be no leakage during the transportation of waste residue, and the exhaust gas emissions during the treatment process should meet the relevant standards.

[0039] Tasks that fail the verification are corrected by changing vehicles, adjusting execution times, or changing storage locations until all constraints are met, generating a final optimized scheduling plan. For example, if a transport vehicle has been scheduled for 3 trips with a total travel time of 9 hours, the 3rd trip will be rescheduled for the next day to be performed by an idle vehicle.

[0040] After obtaining the scheduling plan, it is broken down into executable instructions for each terminal. For example, the transportation terminal receives the departure time, loading and unloading location, and driving route; the warehousing terminal receives the warehousing time and storage tank position; the equipment PLC receives the start time, processing volume, and micro-batch feeding timestamp; and the operator terminal receives the division of labor and working time.

[0041] The specific executable instructions are sent through the 4G / 5G network using a send-confirmation mechanism. After receiving the instruction, the terminal returns a confirmation frame within 10 seconds. If no confirmation is received within 30 seconds, the instruction is resent, with a maximum of 3 resentments. If the resentment fails 3 times, an alarm is sent to the administrator via SMS and APP push notification, requiring manual intervention. In cases where signal blind spots in the workshop caused the terminal to fail to receive instructions, this was resolved by adding a signal amplifier.

[0042] During the task execution, the status of each terminal is monitored in real time, and the location and speed of the transport vehicles are tracked via GPS; the warehousing system updates inventory data in real time; the processing equipment collects real-time load, temperature, vibration and energy consumption data, and all data is displayed visually on the large screen in the central control room.

[0043] IV. Real-time Phase Modulation and Protection: To address raw material composition fluctuations and model prediction errors, a real-time phase modulation mechanism was designed for dynamic vibration compensation. This involves EMD decomposition of real-time vibration data sampled for 10ms, extracting actual vibration signals of each order, filtering out 50Hz power frequency, 100Hz harmonics, and random environmental noise, and comparing the phase shift and amplitude deviation with the predicted signal. The next complete batch is then divided into 10 to 20 micro-batches, with each micro-batch weighing ≤5% of the total weight and ≥0.1 tons of minimum feed amount. By adjusting the feed time of each micro-batch, the vibration generated by the micro-batch is precisely offset to compensate for the remaining vibration deviation.

[0044] Considering the nonlinear coupling effect between vibrations of different orders—that is, the mutual influence between vibrations of different orders, where a change in one order can trigger synchronous changes in other orders—a 5×5 coupling coefficient matrix is ​​constructed based on historical data. This matrix is ​​used to correct the feeding time, with the correction amplitude ≤ 20% of the original modulation amount to avoid over-correction that could induce new vibrations. The coupling coefficient matrix serves as a vibration coupling quantization tool, K... ij =0.7×K ij0 +0.3×K ijs Element K ij K is the influence coefficient of the i-th order vibration on the j-th order vibration, taking values ​​from 0 to 1. ij0 K represents the initial coupling coefficient obtained from a single-order vibration-specific excitation experiment. ijsThe statistical coupling coefficients are obtained from historical operating data. Initial coupling coefficients are obtained primarily through single-order vibration excitation experiments, where the equipment's natural frequencies are excited and vibration response data is collected. Correlation analysis is then performed based on historical operating data to correct the initial coefficients. Finally, the coefficients are weighted and fused at a 7:3 ratio to obtain the final coefficients, constructing a complete coupling coefficient matrix. The matrix is ​​dynamically calibrated every six months based on new operating data to ensure the accuracy of the coupling characteristics. This matrix is ​​used to correct the feeding time, calculate the additional deviations caused by other orders of vibration, and finally obtain the optimized feeding adjustment amount. The correction range does not exceed 20% of the original modulation amount to avoid over-correction that could trigger new vibrations.

[0045] The feeding interval between adjacent micro-batches must meet the requirement of being ≥10 seconds of the minimum feeding interval of the equipment and ≤1 / 20 of the total processing time. After determining the feeding time, pre-evaluate the offset rate and energy consumption after modulation. If the offset rate is <85% or the energy consumption increases by >5%, then readjust the number of micro-batches and the feeding interval.

[0046] When any first-order vibration acceleration exceeds 120% of the safety threshold, immediately cut off the power to the feeding system, trigger an audible and visual alarm, and record the current vibration data and equipment status. After the vibration drops below the safety threshold, recalculate the modulation parameters and resume feeding. In the past, a sudden change in the raw material batch caused the second-order vibration to exceed the standard, so it was necessary to stop the machine in time to avoid bearing damage.

[0047] V. Equipment Remaining Life Prediction: After each batch is processed, a comprehensive evaluation index is calculated, as follows: (1) Vibration index, a total of 4 items, including mode cancellation rate, overall cancellation rate of the whole frequency band, effective value of vibration acceleration, and effective value of vibration velocity. Among them, mode cancellation rate = 1 - actual vibration amplitude / predicted vibration amplitude. When the mode cancellation rate > 0, it means that the vibration is cancelled. The larger the value, the better the cancellation effect. When the mode cancellation rate = 0, it means that the vibration is not cancelled and is the same as the predicted value. When the mode cancellation rate < 0, it means that the vibration is enhanced and the cancellation fails. The modulation parameters need to be adjusted immediately. (2) Energy consumption indicators, a total of 2 items, including energy consumption per unit of waste residue treatment and energy consumption reduction rate; (3) Scheduling indicators, a total of 7 items, including on-time completion rate of tasks, average vehicle utilization rate, average warehouse turnover rate, crude oil recovery rate, unit recovery energy consumption, unit processing cost, and average equipment load rate.

[0048] The weights of the indicators are set according to the needs of the enterprise, namely, the comprehensive cancellation rate of the whole frequency band is ≥30%, the effective value of vibration acceleration is ≥20%, the energy consumption reduction rate is 10%, and the remaining indicators are allocated according to their importance, with a weight sum of 1; after normalizing each indicator, the weighted sum is obtained to obtain the comprehensive evaluation value.

[0049] When the comprehensive evaluation value is less than 0.7, the offset rate is less than 85%, the equipment load is greater than 110%, or the energy consumption exceeds the standard by more than 5%, the closed-loop adjustment mechanism is triggered to recalculate the vibration characteristic spectrum and phase difference of the batch to be processed, update the low resonance batch sequence, adjust the phase modulation parameters, and generate a new scheduling scheme for execution.

[0050] To facilitate data traceability, all historical data is stored using AES-256 encryption. This allows for the tracing of the entire process information through multiple dimensions such as batch number, component code, and equipment number. Daily reports are generated daily, and comprehensive statistical reports are generated monthly, quarterly, and annually, including total recovery volume, resource utilization rate, cost analysis, vibration mitigation effect, and equipment health status.

[0051] Incremental training of the model is performed once per quarter, updating only the leaf nodes of the GBDT model while preserving the original tree structure to avoid catastrophic forgetting. The newly added 3 months of valid data are used, and the model error after training must be ≤ the original model; otherwise, it will be rolled back to the previous version.

[0052] The remaining equipment life prediction uses a two-parameter Weibull distribution model to describe equipment failure patterns. Input parameters include the equipment's cumulative operating time, the monthly average of historical effective vibration acceleration values, the monthly average of the comprehensive vibration cancellation rate across the entire frequency band, the equipment's monthly average energy consumption, and the number and type of historical failures. Output parameters include the equipment's remaining service life, bearing's remaining service life, impeller's remaining service life, and recommended maintenance time. When the remaining life is less than 30 days, a preventative maintenance reminder is generated, specifying the components to be inspected, the recommended maintenance time, and the methods. The equipment failure probability density function is used in this model. In the formula, β is the shape parameter, which determines the shape of the failure curve. β < 1 represents the early failure period, where the failure rate decreases over time; β = 1 represents the random failure period, where the failure rate is constant; β > 1 represents the wear-out failure period, where the failure rate increases over time, reflecting the equipment failure mode. η is the scale parameter, reflecting the average lifespan of the equipment, and t is the cumulative operating time of the equipment in days. For example, the No. 2 processing equipment has been running for 4.2 years, with an average historical effective value of 0.35 mm / s², an average comprehensive vibration cancellation rate of 88.5% across the entire frequency band, an average energy consumption of 120 kWh / ton, and 2 historical failures. After inputting into the model, the calculated shape parameter β = 2.3, the scale parameter η = 8.1 years, and the predicted remaining service life of 3.9 years, requiring no immediate maintenance.

[0053] Example 2: like Figure 1 As shown, an edge-cloud collaborative three-layer architecture is adopted, which is a commonly used deployment architecture in the existing industrial internet. It mainly consists of 8 modules, which communicate with each other through a standardized OPC UA industrial interface; the specific details are as follows: The data acquisition and preprocessing module is the data foundation of the system. It is installed at the edge acquisition layer and consists of an edge gateway, multiple types of sensors, a PLC controller, a synchronous trigger controller, and a data preprocessing unit. Among them, the multiple types of sensors include accelerometers, temperature sensors, diffused silicon pressure sensors, laser displacement sensors, near-infrared spectrometers, and industrial cameras. The synchronous trigger controller is used to perform millisecond-level synchronous triggering of all acquisition devices.

[0054] The data preprocessing unit incorporates sub-units for outlier detection, missing value completion, time-series alignment, and format conversion. The outlier detection sub-unit uses the 3σ criterion to remove erroneous data that exceeds a reasonable range. The missing value completion sub-unit selects either Kriging interpolation or mean-based completion based on the missing value rate. The time-series alignment sub-unit unifies data with different sampling frequencies to a 10ms time axis through cubic spline interpolation. The format conversion sub-unit converts various types of raw data into the system's unified Parquet columnar storage format, improving storage and query efficiency.

[0055] The module takes raw data from various sensors and control systems as input and outputs a standardized preprocessed dataset. It also receives differentiated detection cycle information from the performance evolution prediction module and adjusts the sampling frequency and acquisition points of each acquisition device in real time.

[0056] The mapping module, built on the gradient boosting tree algorithm, includes a feature fusion subunit, a multi-order vibration prediction subunit, and an energy consumption prediction subunit. The feature fusion subunit fuses the waste residue composition data obtained from near-infrared spectroscopy with the process parameters of the corresponding production process to generate an eight-dimensional composition feature vector. The multi-order vibration prediction subunit contains five independent GBDT sub-models, corresponding to the frequency, amplitude, and initial phase prediction of first to fifth order vibrations, respectively. The energy consumption prediction subunit predicts the real-time energy consumption curve during the batch processing based on the composition features and equipment operating parameters.

[0057] The module takes a standardized preprocessed dataset as input and outputs a multi-order natural vibration characteristic spectrum and a predicted energy consumption curve, providing basic data for subsequent phase conjugate matching scheduling.

[0058] The phase conjugate matching scheduling module includes a phase conjugate degree calculation subunit, a comprehensive cancellation rate calculation subunit, a multi-objective optimization subunit, and a chaotic sensitive area matching subunit. Specifically, the phase conjugate degree calculation subunit calculates the phase conjugate degree of each order of vibration between any two batches based on the vibration characteristic spectrum of each batch; the comprehensive cancellation rate calculation subunit calculates the full-band comprehensive cancellation rate of batch combinations by combining the energy proportion weights of each order of vibration; the multi-objective optimization subunit calculates the comprehensive score of batch combinations based on the vibration-energy consumption multi-objective optimization function; and the chaotic sensitive area matching subunit locates the chaotic sensitive areas of each batch and prioritizes matching batch combinations with overlapping sensitive areas.

[0059] The module takes vibration characteristic spectra and energy consumption curves of all batches to be processed as input and outputs a low-resonance batch combination sequence, providing a basis for generating scheduling schemes.

[0060] The real-time phase modulation module includes a vibration signal decomposition subunit, a deviation calculation subunit, a micro-batch splitting subunit, and a feeding time correction subunit. The vibration signal decomposition subunit performs EMD decomposition on the real-time acquired vibration data to extract the actual vibration signals of each order. The deviation calculation subunit compares the actual vibration signals with the predicted signals to calculate the phase offset and amplitude deviation. The micro-batch splitting subunit splits the next complete batch into multiple micro-batches of equal weight. The feeding time correction subunit, in conjunction with the coupling coefficient matrix, calculates the optimal feeding time for each micro-batch to achieve real-time dynamic compensation of vibration.

[0061] The module takes real-time vibration data and predicted vibration characteristic spectrum as input, and outputs micro-batch feeding time adjustment instructions, which are sent to the equipment PLC for execution.

[0062] The scheduling scheme generation and constraint verification module includes an initial scheme generation subunit, a multi-dimensional constraint verification subunit, and a scheme optimization subunit. The initial scheme generation subunit generates an initial scheduling scheme by combining the low-resonance batch sequence and the status of recycled resources. The multi-dimensional constraint verification subunit verifies the legality of the initial scheme from five dimensions: transportation, warehousing, equipment, personnel, and environmental protection. The scheme optimization subunit adjusts the tasks that fail the verification and generates the final optimized scheduling scheme.

[0063] The module takes low-resonance batch combination sequence and real-time status data of recycled resources as input, and outputs the final optimized scheduling scheme.

[0064] The instruction issuance and execution monitoring module includes an instruction decomposition subunit, a communication transmission subunit, a status monitoring subunit, and an anomaly alarm subunit. The instruction decomposition subunit breaks down the scheduling scheme into specific instructions executable by each terminal. The communication transmission subunit sends instructions to the corresponding terminals via 4G / 5G networks for confirmation, ensuring reliable instruction transmission. The status monitoring subunit collects real-time operational status data from each terminal and displays it visually on the central control room's large screen. The anomaly alarm subunit triggers audible and visual alarms and notifies management personnel when it detects equipment failures, network interruptions, or other abnormal situations.

[0065] The module takes the final optimized scheduling scheme as input and outputs the execution instructions and running status monitoring data of each terminal.

[0066] The effect evaluation and adjustment module includes an index calculation subunit, an evaluation judgment subunit, and a scheme adjustment subunit. After each batch is processed, the index calculation subunit calculates the comprehensive evaluation index. The evaluation judgment subunit judges whether the scheduling effect meets the standard based on the comprehensive evaluation value. When the evaluation does not meet the standard, the scheme adjustment subunit recalculates the vibration characteristic spectrum and phase difference of the batch to be processed and updates the low resonance batch sequence and scheduling scheme.

[0067] The module takes as input the actual running data after batch processing is completed, and outputs as the comprehensive evaluation results and the adjusted scheduling scheme.

[0068] The mapping module, effect evaluation and adjustment module, instruction issuance and execution monitoring module, scheduling scheme generation and constraint verification module, real-time phase modulation module, and phase conjugate matching scheduling module are all located in the edge computing layer.

[0069] The data management and iteration module, located in the cloud service layer, includes a data storage subunit, a report generation subunit, a model incremental training subunit, and an equipment lifespan prediction subunit. The data storage subunit employs a dual backup architecture (local and cloud), encrypting all historical data with AES-256 and enabling multi-dimensional data traceability. The report generation subunit generates daily, monthly, quarterly, and annual statistical reports. The model incremental training subunit incrementally updates the core model quarterly using newly added data. The equipment lifespan prediction subunit, based on the Weibull distribution model, predicts the remaining lifespan of equipment and generates preventative maintenance reminders.

[0070] The module inputs are the full-process operation data and model output results, and the outputs are statistical reports, model update versions, and equipment maintenance reminders.

[0071] Real-time running data between the edge and cloud is synchronized once per second, and historical running data is synchronized in batches once per hour. Model parameters are synchronized to the edge nodes immediately after being updated. Furthermore, when the network is interrupted, the edge nodes can independently complete all core functions, and the data can be resumed to the cloud after the network is restored.

[0072] Finally, it should be noted that the above embodiments are merely examples for clearly illustrating the present invention and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. A digital management system for resource scheduling and recovery of linoleic acid waste residue, characterized in that, The management system includes: The basic data acquisition module is used to acquire waste residue generation data, basic data of recycled resources, production plan data, and basic parameters of multi-order natural vibration modes of the processing equipment throughout the entire linoleic acid production process. The feature mapping module is used to extract the multidimensional component feature vector of each batch of linoleic acid waste residue, establish the physical relationship between the component features and the vibration excitation of each order of the equipment, and map the multidimensional component feature vector to the multi-order natural vibration feature spectrum of the entire process of the batch of waste residue treatment. The phase conjugate matching scheduling module is used to calculate the phase conjugate degree of each order of the multi-order natural vibration characteristic spectrum of the batch to be processed, and constructs a low resonance batch combination sequence with the goal of maximizing the vibration energy cancellation rate across the entire frequency band. The real-time phase modulation module is used to calculate the phase offset of each vibration mode based on the high-frequency real-time vibration data of the processing equipment, and to perform phase compensation of each vibration mode by dynamically adjusting the particle size of the feeding batch and the feeding time interval. The scheduling scheme generation module is used to generate an initial resource scheduling scheme by combining basic data on recycled resources, forecast data on waste residue recycling demand, and low-resonance batch combination sequences. The constraint verification and execution module is used to perform multi-dimensional constraint verification on the initial resource scheduling plan, including equipment load, transportation capacity, storage capacity, and environmental emissions. It outputs an optimized resource scheduling plan and converts it into execution instructions for distribution. The real-time monitoring module is used to synchronously collect the operating status of each execution terminal, process multi-level vibration data, real-time load data and temperature data of the equipment; The effect evaluation module is used to generate a multi-mode comprehensive cancellation rate based on the real-time vibration data of the processing equipment, and to generate a comprehensive evaluation value of resource scheduling effect by combining the scheduling execution data. Then, based on the multi-mode comprehensive cancellation rate, waste generation data and equipment operating data, the low resonance batch combination sequence and phase modulation parameters are adjusted synchronously.

2. The digital management system for resource scheduling of linoleic acid waste residue recycling according to claim 1, characterized in that, The basic data acquisition module specifically collects the following data: The production workshop of linoleic acid produces five types of waste residues: soap residue, waste clay, distillation residue, deodorization distillate, and hydrolysis residue. These waste residues include the amount, time of generation, storage location, and batch number of each type. The workshop also includes the reaction temperature, pressure, feed rate, catalyst addition, reaction time, raw material origin, raw material acid value, and moisture content of each corresponding process. The load capacity, mileage, maintenance status, average speed, loading and unloading time and leakage prevention level of transport vehicles; the storage capacity, location, temperature and humidity control capabilities, remaining capacity, leakage prevention level and ventilation capacity of storage facilities; the rated processing capacity, continuous load, operating time, maintenance cycle, historical vibration and fault data of processing equipment; the minimum feeding interval of automatic feeding system; and the job position, skill level, working hours and operational proficiency of operators. Enterprise medium- and long-term production plans and temporary production adjustment data, including daily product output, process scheduling, raw material batch switching, equipment maintenance plans, as well as emergency orders, raw material quality abnormalities, production failures, environmental inspections, and extreme weather warnings; The first to fifth natural vibration frequencies, modal damping ratios, mode shapes, safe acceleration thresholds, and energy percentages of the processing equipment were obtained through hammer impact modal testing.

3. The digital management system for resource scheduling of linoleic acid waste residue recycling according to claim 1, characterized in that, The execution steps of the feature mapping module are as follows: The rapid detection data of each batch of waste residue is integrated with the process parameters of the corresponding production process to generate an eight-dimensional component feature vector containing crude oil content, free fatty acid ratio, solid impurity content, saponification content, moisture content, ash content, pigment content, and phospholipid content, and a unique 32-bit hexadecimal component code is generated for each batch. Establish the physical correspondence between component characteristics and equipment vibration excitation. Among them, crude oil and free fatty acids mainly affect the first-order vibration amplitude, solid impurities and ash mainly affect the second-order vibration amplitude, moisture and saponification mainly affect the third-order vibration amplitude, and pigments and phospholipids mainly affect the fourth and fifth-order vibration amplitudes. Based on historical data from valid batches, a component fingerprint-multi-order vibration feature database is constructed, and a multi-order vibration prediction model is trained and generated. The eight-dimensional component feature vector of the batch to be processed is input into the model to generate the multi-order natural vibration feature spectrum of the batch on the corresponding processing equipment from feeding to processing end, including the frequency, amplitude, initial phase, duration and energy ratio of each order of vibration.

4. The digital management system for resource scheduling of linoleic acid waste residue recycling according to claim 3, characterized in that, The training process of the multi-order vibration prediction model is as follows: The data in the component fingerprint-multi-order vibration feature database were divided into training set, validation set and test set in a ratio of 8:1:

1. Empirical mode decomposition is performed on the original vibration data to separate the first to fifth order independent natural mode functions and filter out noise components; Using an eight-dimensional component feature vector as input and the frequency, amplitude, and initial phase of each order of intrinsic mode function as output, gradient boosting sub-models of orders one to five are trained respectively. The outputs of each sub-model are fused to generate a complete multi-order natural vibration characteristic spectrum. The error is corrected by test set data so that the amplitude prediction error of each vibration order does not exceed 8% and the phase prediction error does not exceed 15°.

5. The digital management system for resource scheduling of linoleic acid waste residue recycling according to claim 1, characterized in that, The execution steps of the phase conjugate matching scheduling module are as follows: The time axis of the multi-order natural vibration characteristic spectrum of all batches to be processed is standardized to be the longest processing time among all batches. Calculate the phase conjugate degree of each vibration mode between any two batches to be processed, where the phase conjugate degree = 1 - |phase difference - 180 degrees| / 180 degrees, and the value ranges from 0 to 1; Based on the energy proportion weight of each order of vibration, the overall cancellation rate of the full-frequency vibration of the two batches is calculated, where the overall cancellation rate of the full-frequency vibration is Σ (phase conjugation degree from the first to the fifth order × energy proportion of each order). Batch combinations with a comprehensive vibration cancellation rate of not less than 85% across the entire frequency band were selected and arranged in order of cancellation rate from high to low and total processing time from short to long to generate a low resonance batch combination sequence.

6. The digital management system for resource scheduling of linoleic acid waste residue recycling according to claim 5, characterized in that, The phase conjugate matching scheduling module also includes: If the cancellation rate of two batches does not reach 85%, the third and fourth batches are introduced in sequence for multi-stage phase matching until the cancellation rate reaches the target. The energy proportion weight of each vibration is dynamically adjusted based on the historical fault data of the processing equipment. For high-order vibrations that cause more than 3 faults, their weight is increased by 50%. When a single batch belongs to multiple qualified combinations, the combination with the highest offset rate and lowest energy consumption is selected first; if the energy consumption is the same, the combination with the shortest total processing time is selected. The total throughput of each low-resonance batch combination is controlled between 90% and 110% of the equipment's rated continuous load.

7. The digital management system for resource scheduling of linoleic acid waste residue recycling according to claim 1, characterized in that, The execution steps of the real-time phase modulation module are as follows: Empirical mode decomposition is performed on the real-time vibration data of the processing equipment to extract the first to fifth order actual vibration signals of the current batch combination and filter out environmental noise. The actual vibration signal is compared with the predicted vibration signal to calculate the phase shift and amplitude deviation of each order of vibration. The next complete batch to be processed is split into multiple micro-batches of equal weight, wherein the weight of a single micro-batch does not exceed 5% of the complete batch and is not less than the minimum feeding amount of the automatic feeding system. The feeding time interval for each micro-batch is calculated based on the phase shift and amplitude deviation of each vibration.

8. The digital management system for resource scheduling of linoleic acid waste residue recycling according to claim 7, characterized in that, The real-time phase modulation module further includes: The nonlinear coupling effect between vibration modes of each order is calculated, a coupling coefficient matrix is ​​established, and the time interval of micro-batch feeding is corrected, with the correction amplitude not exceeding 20% ​​of the original modulation amount; The feeding interval between adjacent micro-batches shall not be less than the minimum interval of the automatic feeding system, and shall not be greater than 1 / 20 of the processing time of a complete batch. After determining the feeding interval, pre-evaluate the offset rate after modulation. If it does not reach 85% or the energy consumption increases by more than 5%, readjust the number of micro-batch splits and the feeding interval. When any vibration acceleration exceeds 120% of the safety threshold of that order, feeding is immediately suspended and an audible and visual alarm is triggered. Once the vibration returns to a safe range, the modulation parameters are recalculated and feeding is resumed.

9. The digital management system for resource scheduling of linoleic acid waste residue recycling according to claim 1, characterized in that, The execution steps of the effect evaluation module are as follows: The cancellation rate of vibration modes from first to fifth order, the comprehensive cancellation rate of vibration across the entire frequency band, the effective value of equipment vibration acceleration, and the effective value of vibration velocity are calculated based on real-time vibration data. Among them, the cancellation rate of each order mode = 1 - actual vibration amplitude / predicted vibration amplitude. Calculate the average load rate, load fluctuation rate, and load peak-to-valley difference based on real-time load data of the equipment; The on-time completion rate of recycling tasks, average utilization rate of transport vehicles, average turnover rate of storage facilities, comprehensive energy consumption per unit of waste residue recycling, and cost per unit of waste residue treatment are calculated based on the scheduling and execution data. Determine the weight coefficients for each evaluation indicator, among which the weight of the comprehensive vibration cancellation rate across the entire frequency band is not less than 30%, the weight of the effective value of equipment vibration acceleration is not less than 20%, and the sum of all weight coefficients is 1; After normalizing the actual values ​​of each indicator, multiply them by their corresponding weights and sum them to generate a comprehensive evaluation value for resource scheduling effectiveness.

10. The digital management system for resource scheduling of linoleic acid waste residue recycling according to claim 1, characterized in that, The management system also includes a data management module, and the specific execution steps are as follows: Encrypt and store all historical waste generation data, compositional data, vibration characteristic spectrum data, batch combination data, phase modulation parameters, scheduling schemes, execution status, equipment operating conditions, and evaluation data. By using waste residue batch number, component code, transport vehicle number or processing equipment number, the entire process information of waste residue from generation, component detection, vibration prediction, phase matching, micro-batch modulation, transportation, storage to processing can be traced, as well as the vibration and operating condition data of the corresponding equipment at all times. Generate daily, monthly, and annual statistical reports, including total recovery volume, resource utilization rate, cost analysis, energy consumption analysis, vibration mitigation effect, equipment health status, and statistics on abnormal events; Each quarter, incremental training is performed on the component-vibration mapping model, phase matching model, and phase modulation model based on newly added historical data, ensuring that the prediction error of the updated model is not higher than that of the original model. Based on historical vibration data and vibration mitigation effects, a model for predicting the remaining life of equipment is established to generate preventative maintenance reminders.

Images