Search

CN-121980524-A - Cosmetic full-link quality traceability control system based on artificial intelligence

CN121980524ACN 121980524 ACN121980524 ACN 121980524ACN-121980524-A

Abstract

The invention relates to the technical field of data processing, in particular to an artificial intelligence-based cosmetic full-link quality traceability control system. The invention locks the problem source in the production process domain by utilizing the orthogonal combination of the batch concentration degree and the space dispersion degree, eliminates the interference of local environmental factors in the circulation link, realizes the high-precision two-classification of two kinds of strong-correlation faults by a comparison competition mechanism of the pressure waveform matching degree and the temperature waveform matching degree, and forms a data driving treatment paradigm from a phenomenon layer to a mechanism layer and then to a regulation layer by taking after-sale feedback as a macroscopic probe and the equipment waveform as microscopic fingerprints. The invention effectively solves the problems of low recognition accuracy of abnormal reasons and lag adjustment of quality control parameters caused by relying on a single data source, lag indexes and artificial experience judgment.

Inventors

  • DENG JIACHUAN

Assignees

  • 妆溯(广东)数字科技有限公司

Dates

Publication Date
20260505
Application Date
20260403

Claims (10)

  1. 1. Cosmetic full-link quality traceability control system based on artificial intelligence, which is characterized by comprising: The acquisition module is used for acquiring after-sale feedback data of various abnormal products, filling head pressure waveforms and sealing temperature waveforms in the production process in real time; The judging module is used for judging that the production process is abnormal according to the comparison result of the process abnormality index and the preset index threshold value to obtain a judging result, wherein the process abnormality index is determined based on the batch concentration degree and the space dispersion degree, and the batch concentration degree and the space dispersion degree are determined based on the after-sales feedback data; The identification module is used for identifying that filling abnormality occurs according to the similarity between the target feature vector and the filling standard feature vector in the preset mapping database based on the judging result, wherein the target feature vector is determined based on the after-sales feedback data; The determining module is used for determining mechanical abrasion abnormality of the filling head or aging abnormality of the sealing heater according to time sequence characteristics of pressure waveform matching degree and temperature waveform matching degree based on the filling head pressure waveform, the sealing temperature waveform and a preset filling fault waveform library; the optimizing module is used for optimizing a pressure compensation coefficient of the filling head according to the impact pulse amplitude of the pressure waveform of the filling head based on the mechanical abrasion abnormality of the filling head, optimizing a sealing temperature compensation value according to the heating time constant of the sealing temperature waveform based on the aging abnormality of the sealing heater; The adjusting module is used for adjusting the preset index threshold according to the time sequence distribution characteristic of the judging result in a preset observation period based on the optimized pressure compensation coefficient or the optimized sealing temperature compensation value.
  2. 2. The artificial intelligence based cosmetic full link quality traceability control system according to claim 1, wherein when the process abnormality index is greater than the preset index threshold, it is determined that an abnormality occurs in the production process, and a determination result is obtained.
  3. 3. The artificial intelligence-based cosmetic full-link quality traceability control system according to claim 2, wherein the process anomaly index is obtained based on a product of a calculated lot concentration and a spatial dispersion, the lot concentration is obtained based on a calculated kurting coefficient of a production lot number of each of the anomaly products, and the spatial dispersion is obtained based on a standard deviation of distances between geographical coordinates and preset center point coordinates of each of the anomaly products.
  4. 4. The artificial intelligence based cosmetic full link quality traceability control system of claim 3, wherein said identification module comprises: the vector construction unit is used for constructing the target feature vector according to the co-occurrence frequency of the package defect keywords, the filling quantity fluctuation coefficient and the sealing defect rate; And the matching identification unit is used for calculating cosine similarity of the target feature vector and the filling standard feature vector to obtain filling matching degree, and judging that filling abnormality occurs when the filling matching degree is larger than a preset matching degree threshold.
  5. 5. The cosmetic full-link quality traceability control system based on artificial intelligence according to claim 4, wherein when the pressure waveform matching degree is greater than the temperature waveform matching degree, counting duration time that the pressure waveform matching degree is greater than the temperature waveform matching degree to obtain a first duration time, calculating a mean value of the pressure waveform matching degree in the first duration time to obtain a pressure matching mean value, and when the first duration time is greater than a preset risk duration threshold value and the pressure matching mean value is greater than a preset pressure threshold value, judging that the abnormal type of the filling abnormality is abnormal mechanical wear of a filling head; And when the temperature waveform matching degree is larger than the pressure waveform matching degree, counting the duration time when the temperature waveform matching degree is larger than the pressure waveform matching degree, obtaining a second duration time, calculating the average value of the temperature waveform matching degree in the second duration time, obtaining a temperature matching average value, and judging that the abnormal type of the filling abnormality is the aging abnormality of the sealing heater when the second duration time is larger than the preset risk duration threshold value and the temperature matching average value is larger than the preset temperature threshold value.
  6. 6. The artificial intelligence based cosmetic full link quality traceability control system of claim 5, wherein said pressure waveform match is determined based on cosine similarity of said filling head pressure waveform and standard abnormal pressure waveform of said preset filling fault waveform library, and said temperature waveform match is determined based on cosine similarity of said seal temperature waveform and standard abnormal temperature waveform of said preset filling fault waveform library.
  7. 7. The artificial intelligence based cosmetic full link quality traceability control system of claim 6, wherein said pressure compensation factor is optimized based on a comparison of a pulse amplitude mean value calculated based on an average value of impulse pulse amplitudes over a preset filling cycle and a preset pulse amplitude reference value, said impulse pulse amplitude being determined based on a periodic impulse component peak value of said filling head pressure waveform.
  8. 8. The artificial intelligence based cosmetic full link quality traceability control system of claim 7, wherein the seal temperature compensation value is optimized based on a comparison result of a mean value of a heating time constant and a preset heating time constant reference value, wherein the mean value of the heating time constant is calculated based on a mean value of the heating time constants in a preset seal period, and the heating time constant is determined based on a time required for the seal temperature waveform to rise from a starting temperature to a preset target temperature.
  9. 9. The artificial intelligence based cosmetic full link quality traceability control system of claim 8, wherein said adjustment module comprises: A production anomaly density calculation unit for calculating a production anomaly event density from time stamps of all the determination results in the preset observation period; and the adjusting unit is connected with the production abnormal event density calculating unit and is used for adjusting the preset index threshold according to the comparison result of the production abnormal event density and the preset production abnormal event density threshold.
  10. 10. The artificial intelligence based cosmetic full link quality traceability control system of claim 9, wherein said after-market feedback data includes said production lot number, said geographic coordinates, said package defect keyword co-occurrence frequency, said filling quantity fluctuation coefficient, and said closure defect rate for each of said abnormal products.

Description

Cosmetic full-link quality traceability control system based on artificial intelligence Technical Field The invention relates to the technical field of data processing, in particular to an artificial intelligence-based cosmetic full-link quality traceability control system. Background Currently, the cosmetic industry in China has entered trillion consumer markets, and with the comprehensive implementation of management regulations and matched regulations thereof, the quality safety tracing of the whole life cycle becomes the rigidity requirement of enterprise compliance management and brand trust construction. Meanwhile, the deep fusion of the industrial Internet and the artificial intelligence technology promotes the manufacturing industry to evolve towards self-sensing, self-deciding and self-optimizing intelligent factories. Under the background, a set of full-link quality traceability control system capable of communicating raw materials, production, circulation and consumption feedback is constructed, so that the method is not only a necessary choice under the trend of strict supervision, but also a core engine for realizing active quality risk prevention and control, continuous optimization of production process and digital transformation upgrading of cosmetic enterprises. The existing cosmetic quality tracing system mainly focuses on bar code scanning, batch recording and simple sensing monitoring. In the key procedures of filling, sealing and the like, enterprises usually deploy pressure sensors and temperature sensors to alarm single parameter threshold values, or adopt machine vision to carry out off-line spot inspection on the appearance of finished products. Some advanced production lines introduce Manufacturing Execution Systems (MES) to achieve production data collection, but the data is only used for post-query, lacking cross-process, cross-system correlation analysis. In an abnormal processing link, a multi-reliance quality inspector interprets waveforms or re-checks videos according to experience, the fault diagnosis period is long, the accuracy is low, the equipment parameter compensation and the alarm threshold setting are still mainly calibrated manually at regular intervals, and the dynamic change of the production line under high-speed continuous operation is difficult to adapt. Therefore, the method has the following problems that after-sales feedback and production process data are mutually split, abnormal feedback at a consumer end cannot be mapped to specific working procedures and equipment units quickly, process fluctuation is difficult to trace accurately, early degradation characteristics cannot be captured for progressive faults such as mechanical wear of a filling head and aging of a sealing heater, the traditional single-threshold alarm cannot be detected after batch waste products appear, process waveforms such as pressure, temperature and the like are only judged in an out-of-limit mode, fine granularity identification capability for an abnormal root cause is lacking, equipment parameter compensation and alarm threshold depend on manual periodic setting, and closed loop intelligent control cannot be formed after working condition drift. Disclosure of Invention Therefore, the invention provides an artificial intelligence-based cosmetic full-link quality traceability control system, which is used for solving the problems of low abnormality cause identification accuracy and delayed quality control parameter adjustment caused by relying on a single data source, a hysteresis index and artificial experience judgment in the prior art through multidimensional parameter analysis, after-sales feedback spatial clustering and equipment waveform characteristic matching and a dynamic adjustment mechanism. In order to achieve the above object, the present invention provides an artificial intelligence based cosmetic full link quality traceability control system, comprising: The acquisition module is used for acquiring after-sale feedback data of various abnormal products, filling head pressure waveforms and sealing temperature waveforms in the production process in real time; The judging module is used for judging that the production process is abnormal according to the comparison result of the process abnormality index and the preset index threshold value to obtain a judging result, wherein the process abnormality index is determined based on the batch concentration degree and the space dispersion degree, and the batch concentration degree and the space dispersion degree are determined based on the after-sales feedback data; the identification module is used for identifying abnormal filling conditions according to the similarity between the target feature vector and the filling standard feature vector in the preset mapping database based on the judging result, wherein the target feature vector is determined based on the after-sales feedback data; the determining module is used for determ