Big data based tall oil production batch quality traceability management system
By constructing a tall oil production batch quality traceability management system using big data technology, the problem of uncertain material batch responsibility boundaries under flow field fluctuations has been solved. This system enables real-time quality traceability and responsibility attribution during the tall oil production process, improving the accuracy and real-time nature of quality supervision.
Patent Information
- Application Number
- CN202610776175.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-06-01
- Publication Date
- 2026-06-30
Smart Images

Figure CN122311972A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to a big data-based batch quality traceability management system for tall oil production, belonging to the field of tall oil production supervision and management technology. Background Technology
[0002] Currently, in the forest chemical production system, tall oil distillation systems use continuous feed and discharge buffer tanks. Existing quality traceability methods establish a correspondence between inbound ledger records and finished product batch numbers. The basic logic is to establish a mapping path between raw material batches and product quality, thereby meeting the quality supervision needs in the production process. Tall oil contains resin acid components. When the ambient temperature or material temperature fluctuates and reaches the crystallization critical point, the resin acid components precipitate and crystallize inside the tank. These crystals form deposits on the inner wall or bottom of the tank, reducing the effective volume of the tank participating in fluid circulation and creating flow stagnation areas. Under these non-ideal fluid flow conditions, the residence time distribution of different raw material batches in the tank exhibits non-linear changes, making the static matching logic based on ledgers unable to accurately represent the actual dynamic mixing ratio of materials.
[0003] To address the uncertainty in attribution caused by flow field fluctuations, existing solutions attempt to calculate material particle trajectories using fluid dynamics simulation algorithms. However, these algorithms involve solving complex partial differential equations, consuming significant processor resources and resulting in data output lagging behind production nodes, failing to meet real-time auditing requirements. Furthermore, conventional linear models cannot accommodate state transitions caused by crystallization dead zones. While physical hardware improvements can offer some optimization to enhance material flow, relying solely on hardware enhancements is insufficient for achieving accurate quality anchoring when dealing with complex batch traceability logic. For example, the Chinese utility model patent with authorization announcement number CN209973308U... The patent discloses a tall oil storage and feeding tank used in multi-tower vacuum depressurization continuous distillation production. Through a steam heating device and inner cover design, the temperature of the material in the storage tank and pipeline is maintained to ensure smooth transportation in cold weather. This solution is essentially a passive protection of the physical properties of the material. It does not address the data weight flow law of the continuous mixing process. Under dynamic conditions such as production load changes or local cold spots causing resin acid crystallization, it cannot detect the real-time loss of circulation volume. It lacks a logic operator to physically isolate and freeze the historical weights in the flow stagnation area, causing the traceability data to deviate from the actual batch attribution of the physical entity. As a result, it is impossible to clarify the quality responsibility of the raw material supplier in the finished product quality defect judgment process.
[0004] Therefore, the technical problem to be solved by this invention is how to reconstruct the responsibility boundary between material batches and establish a stable traceability model, in view of the sudden changes in the physical properties of tall oil during the mixing process and the nonlinear characteristics generated by the continuous flow field. Summary of the Invention
[0005] To address the problems mentioned in the background art, the technical solution of the present invention is as follows: A big data-based batch quality traceability management system for tall oil production, comprising: The data acquisition module is used to acquire time-series data of flow rate, material level change data, and material temperature data of the storage tank system. The batch weight calculation module is used to convert raw material batches injected into the storage tank at different times into corresponding batch mixing weight matrices based on flow time series data and material level change data. The batch weight calculation module is also used to determine the logical address segment of the corresponding flow stagnation zone in the storage tank when the material temperature is lower than the preset temperature threshold, based on the logical association between material temperature data and preset temperature threshold. It also performs address masking on the data vector of the corresponding logical address segment in the batch mixing weight matrix, divides the batch mixing weight matrix into a temporary suspended subset and a dynamic iteration subset, and only overwrites the data of the dynamic iteration subset through the weight update step size. The traceability map generation module is used to determine the proportion weight of each historical batch at the discharge node based on the batch mixing weight matrix, and to construct a batch-testing association model by combining the quality inspection data of the discharge node, thereby generating a quality traceability map that represents the ownership of responsibility for the quality of raw materials.
[0006] Preferably, when performing address masking, the batch weighting calculation module marks the logical address segment as read-only using mapping instructions and skips the read-only logical address segment during the calculation cycle; when the material temperature rises above the preset temperature threshold, the batch weighting calculation module removes the masking of the logical address segment and incorporates the temporarily suspended subset into the dynamic iterative subset to participate in the calculation.
[0007] Preferably, the batch weight calculation module initializes the batch mixing weight matrix through a preset fluid diffusion probability model, and uses the real-time flow pulse signal as the trigger pulse for weight update step size, mapping the fluid residence pattern in the physical storage tank into the data migration pattern in the logical space.
[0008] Preferably, the traceability map generation module also includes a deviation correction submodule, which is used to perform time-position alignment correction on the proportion weight at the discharge node based on the pipeline volume and flow rate data of the distillation unit.
[0009] Preferably, the traceability map generation module is also used to locate abnormal raw material batches whose proportion weight exceeds a preset threshold by searching the batch-testing association model when the quality inspection data exceeds the preset tolerance range, and output a quality abnormality alarm command indicating the abnormal raw material batch.
[0010] Preferably, the data acquisition module is also used to acquire pressure change data inside the storage tank, and the batch weight calculation module uses the pressure change data to calculate the real-time volume correction coefficient of the storage tank, and calibrates the spatial proportion of the batch mixed weight matrix based on the real-time volume correction coefficient.
[0011] Preferably, when determining the logical address segment, the batch weighting calculation module calculates the dead zone volume ratio at a preset temperature threshold by calling a preset resin acid crystallization rate model, and maps the dead zone volume ratio to the blocking length of the logical address segment.
[0012] Preferably, the system also includes a source ledger synchronization module, which is used to synchronize the warehousing records of upstream raw material batches. The traceability map generation module maps the warehousing records with historical batches in the quality traceability map to trace the responsibility from the end product to the original supply source.
[0013] Preferably, the traceability map generation module is also used to output a quality audit analysis table, which records the proportion weight of each raw material batch, the material temperature change trend, and the corresponding address masking switching sequence within each discharge cycle.
[0014] Compared with the prior art, the beneficial effects of the present invention are: 1. In the batch quality traceability management of tall oil production, the batch weight calculation module is used to translate the mixing process of continuous fluid in the storage tank into a batch mixing weight matrix in the data space. The weight update step size is used to calculate the proportion weight of each historical batch at the current moment in real time, which solves the technical bottleneck of the fuzzy batch boundary under the continuous fluid mixing condition. It realizes the deterministic mapping between the mixing state in the physical space and the traceability data in the logical space, and provides the underlying data support for the quality supervision system with a legally valid evidence chain.
[0015] 2. By introducing a thermodynamic state-aware data topology partitioning mechanism, when the real-time temperature drops to the critical crystallization temperature threshold, the processor intervenes in the storage medium and performs address segment addressing isolation, dividing the weight vector into a temporary suspended subset and a dynamic iteration subset. This technique, which transforms the nonlinear mutation of the physical dead zone into data operator branch control, avoids the high overhead and easily divergent fluid dynamics finite element calculation, ensuring the accuracy of management and supervision data while meeting the hard requirements for real-time data processing in enterprise-level production environments.
[0016] 3. By constructing a bipartite graph mapping structure through the traceability graph generation module, the test data of the discharge node is deeply coupled with the batch proportion weight of the raw materials at the front end. With the help of nonlinear attenuation compensation calculation, the system can not only complete the post-event responsibility attribution audit, but also generate high-precision raw material quality early warning indicators. This transformation from single-dimensional book matching to multi-mechanism collaborative pre-control eliminates the regulatory blind spots caused by sensor white noise and fluid nonlinear fluctuations, so that the final generated responsibility traceability ratio matrix has high engineering robustness. Attached Figure Description
[0017] Figure 1 This is a flowchart of the weight calculation and graph generation process for the tall oil quality traceability system of the present invention; Figure 2 This is an address masking and state transition diagram for a subset of the weight matrix of this invention.
[0018] The objectives, features, and advantages of this invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation
[0019] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.
[0020] A big data-based batch quality traceability management system for tall oil production, characterized in that it includes: The data acquisition module is used to acquire time-series data of flow rate, material level change data, and material temperature data of the storage tank system. The batch weight calculation module is used to convert raw material batches injected into the storage tank at different times into corresponding batch mixing weight matrices based on flow time series data and material level change data. The batch weight calculation module is also used to determine the logical address segment of the corresponding flow stagnation zone in the storage tank when the material temperature is lower than the preset temperature threshold, based on the logical association between material temperature data and preset temperature threshold. It also performs address masking on the data vector of the corresponding logical address segment in the batch mixing weight matrix, divides the batch mixing weight matrix into a temporary suspended subset and a dynamic iteration subset, and only overwrites the data of the dynamic iteration subset through the weight update step size. The traceability map generation module is used to determine the proportion weight of each historical batch at the discharge node based on the batch mixing weight matrix, and to construct a batch-testing association model by combining the quality inspection data of the discharge node, thereby generating a quality traceability map that represents the ownership of responsibility for the quality of raw materials.
[0021] Preferably, when performing address masking, the batch weighting calculation module marks the logical address segment as read-only using mapping instructions and skips the read-only logical address segment during the calculation cycle; when the material temperature rises above the preset temperature threshold, the batch weighting calculation module removes the masking of the logical address segment and incorporates the temporarily suspended subset into the dynamic iterative subset to participate in the calculation.
[0022] Preferably, the batch weight calculation module initializes the batch mixing weight matrix through a preset fluid diffusion probability model, and uses the real-time flow pulse signal as the trigger pulse for weight update step size, mapping the fluid residence pattern in the physical storage tank into the data migration pattern in the logical space.
[0023] Preferably, when generating the quality traceability map, the traceability map generation module constrains the proportion weight of each raw material batch through a normalization operator, so that at the discharge node at the current time t, the proportion weight of all associated historical raw material batches i is... The following quantification rules must be met: Where n is the total number of historical raw material batches that constitute the current output volume and are not masked. For the i-th historical raw material batch at time i The weighting coefficients.
[0024] Preferably, the traceability map generation module also includes a deviation correction submodule, which is used to perform time-position alignment correction on the proportion weight at the discharge node based on the pipeline volume and flow rate data of the distillation unit.
[0025] Preferably, the traceability map generation module is also used to locate abnormal raw material batches whose proportion weight exceeds a preset threshold by searching the batch-testing association model when the quality inspection data exceeds the preset tolerance range, and output a quality abnormality alarm command indicating the abnormal raw material batch.
[0026] Preferably, the data acquisition module is also used to acquire pressure change data inside the storage tank, and the batch weight calculation module uses the pressure change data to calculate the real-time volume correction coefficient of the storage tank, and calibrates the spatial proportion of the batch mixed weight matrix based on the real-time volume correction coefficient.
[0027] Preferably, when determining the logical address segment, the batch weighting calculation module calculates the dead zone volume ratio at a preset temperature threshold by calling a preset resin acid crystallization rate model, and maps the dead zone volume ratio to the blocking length of the logical address segment.
[0028] Preferably, the system also includes a source ledger synchronization module, which is used to synchronize the warehousing records of upstream raw material batches. The traceability map generation module maps the warehousing records with historical batches in the quality traceability map to trace the responsibility from the end product to the original supply source.
[0029] Preferably, the traceability map generation module is also used to output a quality audit analysis table, which records the proportion weight of each raw material batch, the material temperature change trend, and the corresponding address masking switching sequence within each discharge cycle.
[0030] Example 1: In a continuously operating tall oil continuous distillation unit with an annual production capacity of 100,000 tons, the unit adopts a continuous feed and discharge buffer tank system. The crude tall oil raw material is purchased in multiple batches from different forest areas. Under winter production conditions, the temperature of the material entering the tank fluctuates beyond the standard deviation due to the decrease in ambient temperature. Tall oil contains resin acids. When the material temperature drops to the resin acid precipitation threshold range, the resin acid components precipitate and crystallize at the bottom and inner wall of the tank, forming a flow stagnation zone. This causes a nonlinear abrupt change in the volume of the tank that effectively participates in fluid circulation. Under the condition that conventional batch matching logic based on static inventory records cannot characterize the actual residence time and mixing ratio of different batches of raw materials in the storage tank, when the rosin acid content of the end product deviates from the preset tolerance range, the management system lacks a data mapping chain to trace the abnormal upstream raw material batches. The data acquisition module continuously acquires the flow time-series data, material level change data, and material temperature data of the storage tank system. At the same time, it acquires the pressure change data in the storage tank through the pressure transmitter at the bottom. The batch weight calculation module has a built-in state equation algorithm, which combines the pressure change data with the material level change data. The data is used to calculate the change in hydrostatic pressure of the liquid and substitute it into the Poisson's ratio formula for elastic deformation of the tank and the fluid compressibility formula to derive the real-time volume correction coefficient characterizing the micro-expansion of the tank due to stress and the compaction of the fluid itself. In the specific derivation process, the processor first multiplies the change in hydrostatic pressure by the isothermal volume compressibility coefficient of tall oil to calculate the volume reduction caused by the fluid's own compaction under pressure. Then, the change in hydrostatic pressure is substituted into the Poisson's ratio formula for elastic deformation based on the cylindrical thin film stress theory to solve for the radial micro-strain of the tank wall. Finally, the expansion of the tank's geometric boundary is obtained by integrating the micro-strain. The volume expansion is calculated, and the net volume variable after deducting the volume reduction is divided by the nominal total volume of the storage tank to obtain the correction coefficient. The batch mixing weight matrix has a clear data reconstruction topology and reproducibility. The specific flow and construction logic of the batch weight calculation module, based on the flow time series data and material level change data, converts the raw material batches injected into the storage tank at different times into the corresponding batch mixing weight matrix is as follows: The batch mixing weight matrix is stored in memory using a two-dimensional data array topology, where the row vector dimension of the matrix corresponds to the weight update step size. Discretization of time steps in units The column vector dimension of the matrix corresponds to the unique numerical identifier of each historical batch of raw materials injected into the storage tank system at different time periods. The intersection elements in the matrix are used to represent the dynamic proportion weight of a specific raw material batch at a specific time. First, the processor monitors the real-time feed flow rate in the flow time series data. The step change is used to automatically identify and segment the physical boundaries of new raw material batches. When the flow time series data indicates that the feed flow rate changes from zero to an input state greater than the preset full load threshold, the batch weight calculation module dynamically adds a new column vector dimension to the batch mixed weight matrix to map the new input raw material batch. Secondly, the processor dynamically extracts the real-time geometric height of the liquid level in the storage tank through the material level change data, and performs product calibration by combining the nominal physical geometric parameters pre-stored in the storage tank and the derived real-time volume correction coefficient, thereby calculating the current moment. The total effective volume of material actually participating in the full mixed-flow cycle within the storage tank; subsequently, the processor converts the absolute feed volume obtained from the flow time series data within this time step, i.e., the real-time feed flow rate. With weight update step size The product of the two components is normalized and ratioed to the actual total effective volume of the fully mixed flow, and this ratio is used to dynamically generate new batches of raw materials. The initial weight element values are read from the matrix and then filled into the matrix at the corresponding current time step. With new batch identification At the intersection of rows and columns, for the column vector of previous historical raw material batches already existing in the storage tank, the discharge flow rate is used as the basis. The resulting overall volume reduction is proportionally attenuated and covered by historical weight elements in the row vectors, thereby completely reconstructing the responsibility matrix representing the multi-batch mixing boundary of the physical flow field in the logical space; the batch weight calculation module receives the real-time feed flow rate. Real-time discharge flow rate The processor calculates the dynamic viscosity correction factor based on the real-time temperature T(t) and maintains a dynamic data array representing the weight of historical batches within the physical storage medium. When the real-time temperature T(t) is greater than or equal to the preset crystallization critical temperature threshold, the processor outputs the transient effective storage capacity. The time step iteration logic is applied to all batch elements in the dynamic data array, and when the material temperature is detected to be lower than the preset crystallization critical temperature threshold.
[0031] The batch weighting calculation module determines the logical address segment of the corresponding flow stagnation zone in the storage tank based on the logical relationship between the material temperature data and the preset temperature threshold. It then retrieves the preset resin acid crystallization rate model based on the Avramyi isothermal crystallization phase transition kinetics theory and collects material temperature data to calculate the relative crystallinity at the current temperature. The calculation formula satisfies Where Z is the crystallization rate constant characterizing the nucleus growth rate, n is the Avramie exponent determined by the nucleation mechanism, and t is the cumulative time the material remains below a preset temperature threshold. The total physical volume of the storage tank is obtained. According to relative crystallinity The physical dead zone volume is calculated by multiplying the product of the maximum precipitation volume fraction of the resin acid. A one-dimensional queue data structure based on a time-first-in, first-out (FIFO) rule is maintained in memory, mapping the tank volume to a continuous logical address space. The head of the queue corresponds to the latest incoming batch, and the tail of the queue corresponds to the historical batch with the longest residence time. Due to the laminar heat transfer characteristics of tall oil in the tank, the fluid turnover rate in the low-temperature stagnation zone near the tank wall and bottom is extremely low. This causes the resin acid components that precipitate and crystallize in this area to be highly concentrated at the very end of the fluid circulation in physical space. In terms of fluid dynamics, the average residence time of the locally crystallized material entrained in this cold spot is much longer than that of the mainstream material in the center of the tank. Thus, a three-dimensional spatial crystallization dead zone and logical time are established. The system establishes a physical isomorphic mapping relationship between the earliest batches entering the tank on the axis. During this mapping process, the system establishes physical boundary conditions based on the dynamic viscosity change and density increment after the resin acid phase transition. The highly viscous fluid that precipitates crystals settles directionally under the influence of gravity and adheres to the bottom of the tank, forming a high-resistance viscous isolation layer at the dead zone interface that blocks the convection exchange of new materials entering the central mainstream zone. This microscopic density stratification effect eliminates secondary backmixing interference from fluids at multiple time scales in three-dimensional space. This ensures that the fluid in the physically enclosed dead zone must correspond to the earliest batch entering the tank that did not participate in dynamic convection on the time axis in terms of macroscopic residence time. According to the physical dead zone volume... With physical total volume The proportion is calculated by truncating data units of corresponding length from the tail of the one-dimensional queue data structure as logical address segments. The batch weight calculation module marks the logical address segments as read-only through mapping instructions, masking the data vectors of the corresponding logical address segments in the batch mixed weight matrix. Based on the real-time proportion weight distribution at the threshold, the processor divides the dynamic data array into a temporary suspended subset of the first storage address segment and a dynamic iteration subset of the second storage address segment. The processor issues a calculation lock-up instruction to the batch elements in the temporary suspended subset, terminating their time step iteration logic within the subsequent time step Δt, so that their proportion weight remains statically constant in the database. At the same time, the processor calculates the residual probability space of the data corresponding to the dynamic iteration subset and sets the real-time feed flow rate. The newly generated batch of data is only mapped to the updated weights within the dynamically iterative subset, and the iterative calculation formula for the percentage weights satisfies... ,in, For the i-th historical raw material batch at time i The weighting coefficients, The weighting coefficients from the previous time step. For real-time discharge flow rate, The feed flow rate is real-time, and Δt is the weight update step size. For transient effective reserves, A Boolean input variable indicating whether the current feed belongs to batch i.
[0032] When the material temperature rises above the preset temperature threshold, the batch weighting calculation module removes the shielding of the logical address segment, merges the temporarily suspended subset into the dynamic iterative subset to resume calculation, and the traceability graph generation module determines the proportion weight of each historical batch at the discharge node based on the batch mixed weight matrix. The traceability graph generation module constrains the proportion weight of each raw material batch through a normalization operator, so that the proportion weight of all associated historical raw material batches i at the discharge node at the current time t is equal. Satisfy quantification rules Where n is the total number of historical raw material batches that constitute the current output volume and are not masked. For the i-th historical raw material batch at time i The weighting coefficients are determined based on the principle of mass conservation and advection transport of fluid in a closed pipeline. A position alignment correction procedure is then performed to obtain the physical pipeline space volume between the discharge node and the target sampling point. Combined with real-time discharge flow rate Calculate the advection delay time in the pipeline The calculation formula satisfies Delay time for pipeline flow Converting to discrete time step counts, the weight array at the discharge node is shifted along the time axis to ensure that the weights at time t are equal. With sampling and testing time Establish temporal synchronization associations for quality inspection data and correct the mapping relationship to To eliminate data phase differences caused by spatial transport distance, when the system detects that the quality inspection data of the discharge node exceeds the preset tolerance range, the traceability map generation module combines the quality inspection data of the discharge node to construct a batch-test correlation model, locate abnormal raw material batches whose proportion weight exceeds the preset proportion threshold, and the system outputs a quality abnormality alarm command indicating the abnormal raw material batch.
[0033] Example 2: In a continuously operating 500L pilot-scale tall oil buffer mixing system, a flow pulse signal containing 5% random white noise was set to simulate pumping fluctuations in an industrial setting. The system acquired inlet and outlet flow rate data using a Coriolis mass flow meter with a measurement accuracy of 0.2%, and collected material temperature data using a platinum resistance temperature sensor with a measurement accuracy of 0.1℃. Three inert tracers with gas chromatography-identifiable characteristics were injected into the system inlet at different time intervals to calibrate different batches of input materials. Regarding the setting of the sampling period Δt, the system established a judgment model relating the fluid mixing time constant to the flow pulse bandwidth based on an engineering trade-off between the data acquisition time domain resolution and the processor memory iteration load. When the flow rate change rate captured by the sensor is greater than or equal to... When the flow rate is 10 L / min, to avoid signal aliasing and distortion, the system sets Δt to tend towards the lower limit of the range and sets the value of Δt to 10 s. The global update of the dynamic data array is completed within a calculation time window of no more than 100 ms. In view of the difference between the calculation step size of the second-level flow pulse and the time scale of the macroscopic crystallization process, the system introduces a dual-time axis decoupled operation mechanism in the processor architecture. That is, the high-frequency iteration with a step size of 10 s is strictly limited to driving the advection migration operation of the weight matrix, while the call to the resin acid phase change kinetic model is allocated to an independent low-frequency integration thread. This low-frequency thread outputs the smoothed transient effective storage parameters with an hourly sampling window, thereby avoiding the logical conflict of using a microscopic high-frequency clock to physically divide the slow thermodynamic phase change.
[0034] The experiment established a control group based on the full volume integral rule and an experimental group using a data topology segmentation and weight vector freezing mechanism. The mechanism's response characteristics were verified by adjusting the input material temperature to generate a gradient change. When the material temperature was maintained at 65℃, the fluid in the tank exhibited a fully mixed flow state. The absolute deviation between the batch tracking weight output by the control group and the tracer concentration measured by gas chromatography was 1.2%, while the absolute deviation for the experimental group was 1.1%, both within the standard tolerance range. The preset critical crystallization temperature threshold of 55℃ was also determined. At that time, localized cold spots in the tank triggered initial crystallization. The processor calculated the dynamic viscosity correction factor and output the decreased transient effective storage capacity. Due to a lack of data mapping on changes in physical state, the absolute deviation of the control group nonlinearly increased to 8.5%. The experimental group, however, dynamically updated the transient effective reserves. To suppress the absolute deviation to 1.5%, when the material temperature drops to 45℃, a large amount of resin acid crystals are deposited at the bottom of the storage tank, forming a physical dead zone occupying one-third of the total volume. The batch weight calculation module determines the logical address segment corresponding to the flow stagnation zone based on the logical association between the material temperature data and the preset temperature threshold. The batch weight calculation module marks the logical address segment as read-only through the mapping instruction, and masks the data vector corresponding to the logical address segment in the batch mixed weight matrix. The processor divides the dynamic data array into a temporary suspended subset of the first storage address segment and a dynamic iteration subset of the second storage address segment. It issues a calculation lock-up instruction to the batch elements in the temporary suspended subset, stopping their time step iteration calculation in the subsequent time step Δt, so that the proportion weight of the early historical batches stuck in the physical dead zone remains static and constant in the database.
[0035] Simultaneously, the processor calculates the residual probability space of the data corresponding to the dynamic iterative subset, and the real-time feed flow rate. The newly generated batch of data is mapped only to the dynamically iterative subset, and the processor follows the formula. Perform weight updates, where, Let be the weighting coefficient of the i-th historical raw material batch at time t. The input variable indicates whether the current feed belongs to batch i. Under this physical dead zone mutation condition, the absolute deviation of the control group tracked the data exponentially to 34.2%, exceeding the tolerance range. The experimental group eliminated the interference of the flow stagnation zone on the mixing ratio calculation by blocking the data flow path in the memory structure and redistributing the iterative subset. The absolute deviation between the output ratio weight of the experimental group and the measured concentration of the tracer remained at 2.1%. When the material temperature rose above the preset temperature threshold, the batch weight calculation module removed the shielding of the logical address segment, merged the temporarily suspended subset into the dynamic iterative subset to resume the calculation, and the traceability map generation module continuously output the responsibility traceability ratio matrix according to the normalization operator constraining the ratio weight of each raw material batch, forming a continuous quality tracking data chain under non-steady-state conditions.
[0036] Example 3: During a quality audit of a 50,000-ton-per-year tall oil distillation production line, an anomaly was observed in the final rosin product where the unsaponifiable matter content exceeded the standard. The data acquisition module read the feed-in / out time sequence array of the preceding time window containing the weights of the raw material batches corresponding to the dynamic iterative subset in the storage tank system. The batch weight calculation module retrieved the batch mixing weight matrix corresponding to the dynamic iterative subset from the physical storage medium. The traceability map generation module extracted the percentage weight of each historical batch at the discharge node based on the batch mixing weight matrix. To quantify the contribution of each batch of raw materials to the quality fluctuations of the final product, the system acquires real-time detection indicators at the discharge node. Compared with the preset quality benchmark value The system calculates real-time detection indicators. Compared with the preset quality benchmark value The absolute deviation value is calculated and mapped to the batch-analysis correlation model. The batch-analysis correlation model calculates the weight of each historical batch. With initial mass parameters The sum of the products of these factors predicts theoretical output indicators. Theoretical output indicators The calculation formula satisfies Where n is the total number of historical raw material batches, Let be the initial quality parameters tested by the testing equipment for the i-th batch during the warehousing stage, when the theoretical output index With real-time detection indicators When the absolute residual is greater than the half-width σ of the dynamic confidence interval, the system determines that the variation in raw material properties has caused a quality deviation and triggers the source map generation logic.
[0037] The traceability mapping generation module performs a reverse search on the batch mixed weight matrix to locate the proportion weight. For specific batches exceeding a preset proportion threshold H, the traceability map generation module generates a quality traceability map representing the ownership of raw material quality responsibility based on the unique digital identifier of the specific batch. The system outputs a quality anomaly alarm command based on the quality traceability map and writes an audit data packet containing the anomaly batch identifier, quality offset, and associated supplier number into the credit rating database associated with the corresponding raw material supplier via a network interface. The half-width σ of the dynamic confidence interval is calculated according to the formula... Determined, where k is the confidence coefficient and m is the total number of sampling periods. These are the finished product test observation values over the past m sampling periods. The system generates a dynamic confidence interval based on the historical test standard deviation, which is the average value of the finished product test observations over the past m sampling periods. It suppresses the inherent noise of the sensor and the signal interference caused by the drift of the test instrument, and outputs the corresponding abnormal batch identifier to the credit evaluation database.
[0038] Example 4: In the production management system of an 80,000-ton-per-year tall oil distillation production line, the system addresses the mixing mechanism deviation caused by the uneven volume distribution of newly connected raw material storage tanks and the differences in sensor placement. The processor establishes a three-dimensional probability diffusion weight space based on the acquired physical geometric parameters of the target storage tanks. By injecting a calibration fluid containing tracer molecules into the feed end in stages and collecting the concentration time-series response curve of the tracer molecules at the discharge end using an online spectrometer, the processor calculates the phase lag τ and time distribution variance of the discharge end concentration response relative to the feed pulse. Based on the calculation results, the fluid diffusion probability model in the batch weighting calculation module is filled. By adjusting the diffusion coefficient α and the advection transport weight coefficient β in the fluid diffusion probability model, the residual between the predicted concentration response curve and the measured time-series response curve is reduced, so that the residual falls within the preset standard tolerance range.
[0039] When the effective circulation volume decreases due to material buildup on the inner wall of the storage tank, the processor, under normal operating conditions where the material temperature is above the preset crystallization critical temperature threshold, uses material level fluctuation data combined with the cumulative difference between inlet and outlet flow rates to perform volume balance verification. If the system deviation between the material level height measured within 5 consecutive sampling cycles and the theoretical liquid level height calculated based on flow integral is greater than twice the measurement resolution, the processor determines that a non-crystalline dead zone exists within the storage tank and corrects the transient effective storage capacity based on the system deviation. The baseline offset is injected as a negative feedback term into the iterative calculation logic of the weight proportion to offset the loss of weight iteration accuracy due to changes in physical volume. The system maintains the ownership determination error of the quality traceability map within the preset accuracy limit.
[0040] Example 5: In a tall oil storage and transfer station containing multiple tank arrays, the mixing intensity and residence time distribution of different tanks deviate due to differences in aspect ratio. The system obtains the characteristic aspect ratio of a specific tank during the initialization phase. Based on this, the processor retrieves the corresponding first-order Markov chain diffusion probability model from physical memory. This model characterizes the migration probability distribution of batch elements across memory logical address segments. The processor discretizes the one-dimensional queue data structure into M state nodes and constructs an M×M state transition matrix. The main diagonal elements of this matrix represent the probability of material remaining in the current address segment, while the secondary diagonal elements are set according to the local Reynolds number of the fluid as the probability of diffusion to adjacent address segments. This probability is dynamically adjusted using the diffusion coefficient described below. By injecting a calibration fluid containing tracer molecules into the feed end in stages and monitoring the concentration time-series response curve at the discharge end, the processor calculates the deviation of the peak time of the concentration time-series response curve from the theoretical advection time, thereby determining whether the diffusion coefficient α satisfies the calculation relationship. ,in, Δ is the reference diffusion coefficient value corresponding to the characteristic aspect ratio. The dimensionless correction factor determined by the deviation is used to adjust the diffusion coefficient α and its corresponding advection transport weighting coefficient β so that the residual between the predicted concentration response curve and the measured time series response curve falls within the preset standard tolerance range, thereby completing the benchmark calibration of the mixing characteristics of the storage tank.
[0041] When the system faces a data breakpoint challenge due to the instantaneous loss of the feed sensor signal, the processor constructs a second-order Lagrange interpolation model using the feed flow time-series data from the previous five sampling cycles and calculates the missing moment. Estimated feed flow rate Estimate feed flow rate The data is filled into the flow time series array to maintain the continuity of the weight update step size. The system adjusts the dynamic viscosity correction factor accordingly. A linearized segmented calibration method based on material composition characteristics is adopted. When the real-time temperature T(t) is within a preset temperature range, the dynamic viscosity correction factor... Satisfy quantification rules Where a is the fitting slope and b is the fitting intercept, both a and b are physical constants pre-fitted based on the component ratios of a specific raw material sample. The system utilizes a dynamic viscosity correction factor. Corrected transient effective reserves The attenuation slope eliminates the impact of material composition fluctuations on the accuracy of virtual batch allocation, ensuring that the output quality traceability chart maintains the preset judgment sensitivity during material switching. In actual operation, to address the blind spots in raw material composition caused by multiple batch purchases, the system is equipped with an online impedance viscometer and densitometer at the feed main pipe. At the initial stage of each new batch injection, the processor collects the basic physical property parameters of the batch in real time and inputs them into a multivariate regression support vector machine model pre-constructed from historical test records. This multivariate regression support vector machine model specifically integrates the initial apparent viscosity measured by the online impedance viscometer and the fluid density measured by the online densitometer. The model uses temperature and real-time fluid temperature as three-dimensional input feature vectors. Its training set is constructed from 500 historical sets of feature vectors of raw materials from different production areas calibrated by full-component analysis and their corresponding viscosity-temperature fitting constants. The model internally performs high-dimensional space mapping through radial basis kernel function and solves for the classification hyperplane that minimizes the residual between the feature vector and the target fitting constant. This results in the real-time output of the predicted values of constants a and b, which characterize the viscosity-temperature response sensitivity of the current unknown batch of material. Through implicit feature matching, the model dynamically calls the constants a and b that best match the current blind zone material, thereby completing the adaptive assignment of the dynamic viscosity correction factor without relying on offline full-component analysis.
[0042] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.
[0043] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.
Claims
1. A big data-based batch quality traceability management system for tall oil production, characterized in that, include: The data acquisition module is used to acquire time-series data of flow rate, material level change data, and material temperature data of the storage tank system. The batch weight calculation module is used to convert raw material batches injected into the storage tank at different times into corresponding batch mixing weight matrices based on flow time series data and material level change data. The batch weight calculation module is also used to determine the logical address segment of the corresponding flow stagnation zone in the storage tank when the material temperature is lower than the preset temperature threshold, based on the logical association between material temperature data and preset temperature threshold. It also performs address masking on the data vector of the corresponding logical address segment in the batch mixing weight matrix, divides the batch mixing weight matrix into a temporary suspended subset and a dynamic iteration subset, and only overwrites the data of the dynamic iteration subset through the weight update step size. The traceability map generation module is used to determine the proportion weight of each historical batch at the discharge node based on the batch mixing weight matrix, and to construct a batch-testing association model by combining the quality inspection data of the discharge node, thereby generating a quality traceability map that represents the ownership of responsibility for the quality of raw materials.
2. The tall oil production batch quality traceability management system based on big data as described in claim 1, characterized in that, When performing address masking, the batch weighting calculation module marks the logical address segment as read-only using mapping instructions and skips the read-only logical address segment during the calculation cycle. When the material temperature rises above the preset temperature threshold, the batch weighting calculation module removes the masking of the logical address segment and incorporates the temporarily suspended subset into the dynamic iterative subset to participate in the calculation.
3. The tall oil production batch quality traceability management system based on big data as described in claim 1, characterized in that, The batch weighting calculation module initializes the batch mixing weight matrix through a preset fluid diffusion probability model, and uses the real-time flow pulse signal as the trigger pulse for weight update step size, mapping the fluid residence pattern in the physical storage tank into the data migration pattern in the logical space.
4. The tall oil production batch quality traceability management system based on big data as described in claim 1, characterized in that, The traceability map generation module also includes a deviation correction submodule, which is used to perform time-position alignment correction on the proportion weight at the discharge node based on the pipeline volume and flow rate data of the distillation unit.
5. A big data-based batch quality traceability management system for tall oil production according to claim 1, characterized in that, The traceability map generation module is also used to locate abnormal raw material batches with a weight exceeding a preset percentage threshold by searching the batch-test correlation model when the quality inspection data exceeds the preset tolerance range, and output a quality abnormality alarm command indicating the abnormal raw material batch.
6. The tall oil production batch quality traceability management system based on big data according to claim 1, characterized in that, The data acquisition module is also used to acquire pressure change data inside the storage tank. The batch weight calculation module uses the pressure change data to calculate the real-time volume correction coefficient of the storage tank and calibrates the spatial proportion of the batch mixed weight matrix based on the real-time volume correction coefficient.
7. The tall oil production batch quality traceability management system based on big data according to claim 1, characterized in that, When determining the logical address segment, the batch weighting calculation module calculates the dead zone volume ratio at a preset temperature threshold by calling the preset resin acid crystallization rate model, and maps the dead zone volume ratio to the blocking length of the logical address segment.
8. A batch quality traceability management system for tall oil production based on big data as described in claim 1, characterized in that, The system also includes a source ledger synchronization module, which synchronizes the warehousing records of upstream raw material batches. The traceability map generation module maps the warehousing records with historical batches in the quality traceability map to trace the responsibility from the end product to the original supply source.
9. A big data-based batch quality traceability management system for tall oil production according to claim 1, characterized in that, The traceability map generation module is also used to output a quality audit analysis table. The quality audit analysis table records the proportion and weight of each raw material batch, the material temperature change trend, and the corresponding address masking switching sequence within each discharge cycle.
Citation Information
Patent Citations
Tall oil storage and supply tank used in multi-tower vacuum decompression continuous rectification production
CN209973308U
