A supply chain quality control method integrating cloud platform and industrial Internet

Abstract

The present invention provides a supply chain quality control method that integrates a cloud platform and the industrial Internet, belonging to the field of Internet technology. The method comprises: deploying a collection network to the corresponding supply link according to the supply model of each supply link in the supply chain and using industrial Internet technology, and transmitting the collected data of each supply link to the cloud platform; constructing a new standard set according to the current supply demand of each supply link and the mapping relationship between the collection network and the preset risk items of the corresponding supply link; analyzing the collected data of the corresponding supply link according to the new standard set, constructing a quality chain of the corresponding supply link, and performing a comparative analysis with the standard chain to determine quality problems; determining the management and control optimization strategy of the corresponding supply link based on the quality problems of each supply link and optimizing the deployment; and at the same time, feeding back the quality problems to the management end of each supply link based on the cloud platform for reminder. This ensures comprehensive and real-time intelligent management and control of all supply links in the supply chain.

Description

Technical Field

[0001] The present invention relates to the field of Internet technology, and in particular to a quality control method for a supply chain integrating a cloud platform and the industrial Internet. Background Art

[0002] With the development of global economic integration, supply chains have become increasingly complex, covering multiple links such as raw material procurement, manufacturing, product transportation, warehousing management, and sales. Quality issues at any link in this process can have a serious impact on the normal operation and ultimate quality of the entire supply chain.

[0003] Traditional supply chain quality control methods often rely on manual spot checks, paper records, and decentralized quality management systems. These methods suffer from issues such as delayed information transmission, poor data accuracy, and difficulty achieving cross-sector collaborative management. With the development of the Industrial Internet and cloud platform technologies, while some companies have begun to experiment with applying these new technologies to supply chain management, current applications largely remain at the level of simple data storage and transmission, failing to fully leverage the powerful computing and storage capabilities of cloud platforms and the efficient data collection and device interconnection capabilities of the Industrial Internet. Consequently, they are unable to achieve comprehensive and real-time intelligent control of supply chain quality.

[0004] Therefore, the present invention proposes a quality control method for a supply chain that integrates a cloud platform and the industrial Internet. Summary of the Invention

[0005] The present invention provides a supply chain quality control method that integrates a cloud platform and the industrial Internet. It is used to set up a collection network and construct a new standard set under supply demand, so as to facilitate the identification of existing quality problems and optimize management and control, thereby ensuring comprehensive and real-time intelligent management and control of all supply links in the supply chain.

[0006] The present invention provides a supply chain quality control method integrating a cloud platform and the industrial Internet, comprising:

[0007] Step 1: Based on the supply model of each supply link in the supply chain, deploy a data collection network to the corresponding supply link using industrial Internet technology, and transmit the collected data of each supply link to the cloud platform;

[0008] Step 2: Construct a new standard set based on the current supply demand of each supply link and the mapping relationship between the acquisition network and the preset risk items under the corresponding supply link;

[0009] Step 3: Analyze the collected data of the corresponding supply link based on the new standard set on the cloud platform, build a quality chain for the corresponding supply link, and compare and analyze it with the standard chain to identify quality issues, wherein the standard chain is constructed based on the minimum requirements of each new standard in the new standard set;

[0010] Step 4: Based on the quality issues of each supply link, determine the management and control optimization strategy of the corresponding supply link and optimize its deployment. At the same time, the quality issues are fed back to the management end of each supply link based on the cloud platform for reminder.

[0011] Preferably, a new standard set is constructed according to the current supply demand of each supply link and the mapping relationship between the acquisition network and the preset risk items under the corresponding supply link, including:

[0012] Extracting the business environment of each supply link in the supply chain under different historical supply demands from a historical database, analyzing the business environment to obtain a risk value for each preset risk item, and performing cluster analysis on the business environment to construct several risk matrices;

[0013] Standardizing and normalizing each risk matrix to obtain a characteristic vector of each processed matrix, wherein the characteristic vector includes a characteristic coefficient of each preset risk item;

[0014] Sort the characteristic coefficients in each eigenvector from large to small to obtain a size vector;

[0015] Based on the first element in the size vector, three elements are selected in sequence and mapped to a preset coordinate system for three-point connection drawing to determine the drawing shape;

[0016] assigning a first label to a first element of three consecutively screened elements depending on the drawn shape and in conjunction with a center point of the drawn shape;

[0017] Extract the labels under the same preset risk item from all first labels as second labels, and analyze the shape stability and numerical stability of the corresponding preset risk item based on all second labels;

[0018] Determining a first bias based on shape stability and a second bias based on numerical stability according to the drawn shapes and center points corresponding to each preset risk item under different risk matrices;

[0019] When the setting standard of the preset risk item under the corresponding supply demand does not match the first bias and the second bias, the corresponding setting standard is optimized based on the first bias and the second bias to obtain a new standard;

[0020] Match the current supply demand of each supply link with all historical demands of the same supply link. If there is a completely matching historical demand, a new standard set is obtained based on the mapping relationship between the acquisition network and the preset risk items of the corresponding supply link.

[0021] Otherwise, the incompletely matched historical demand with the largest matching coefficient is regarded as the reference demand;

[0022] Relying on the demand difference between the reference demand and the current supply demand, the demand difference is input into the demand analysis model to obtain the bias coefficient of each preset risk item, and the new standard based on the reference demand is optimized twice to obtain a new standard set.

[0023] Preferably, optimizing the corresponding set standard once based on the first bias and the second bias to obtain a new standard includes:

[0024] respectively obtaining a first mismatch factor and a second mismatch factor of the first bias and the second bias with a set standard under corresponding supply requirements;

[0025] Determining an optimization direction for a corresponding set standard according to the first mismatch factor and the second mismatch factor, wherein the optimization direction includes a standard improvement direction, a standard reduction direction, and a standard unchanged direction;

[0026] Adjust the corresponding data setting standards according to the optimization direction.

[0027] Preferably, performing a secondary optimization on the new standard based on the reference requirement includes:

[0028] Considering the new standard of the reference requirement as the first standard;

[0029] Obtain a bias coefficient for each preset risk item, and perform secondary optimization on the corresponding first standard in combination with the item weight of the preset risk item based on the reference demand.

[0030] Preferably, the quality chain of the corresponding supply chain is constructed, including:

[0031] Comparing and analyzing each new standard in the new standard set with corresponding sub-data in the collected data of the corresponding supply link to determine an initial anomaly factor, and constructing an initial vector based on the setting position of the corresponding new standard in the corresponding supply link, and inputting the initial vector into a vector analysis model to obtain an output representation based on the corresponding new standard, wherein the output representation includes: anomaly distribution and the degree of anomaly at each anomaly distribution position;

[0032] Obtain the blank chain corresponding to the supply link from the chain database;

[0033] The output representation of each new standard under the corresponding supply link is input into the blank chain in turn to obtain a quality chain.

[0034] Preferably, determining the initial abnormality factor includes:

[0035]

[0036] in, represents the initial abnormal factor; Indicates the quantization result of the corresponding sub-data; Indicates the quantitative threshold corresponding to the new standard; represents the inverse tangent function; Indicates the set weight corresponding to the new standard; represents a Bernoulli random variable; represents a symbolic function; Represents the contrast sensitivity coefficient.

[0037] Preferably, a comparative analysis with a standard chain is performed to identify quality issues, including:

[0038] Acquire an abnormal combination of abnormal distribution positions based on the quality chain;

[0039] Extracting quality analysis indicators related to the lowest standard combination corresponding to the abnormal distribution position from the standard chain, and extracting an analysis model from the indicator-model database to analyze the abnormal combination, and determining a clear sub-problem and a fuzzy sub-problem corresponding to the abnormal distribution position, wherein the fuzzy sub-problem is between clear and unclear and has a fuzzy probability;

[0040] Obtaining a first indicator for comparison with the clear sub-problem and a second indicator for comparison with the fuzzy sub-problem, performing same-indicator extraction on the first indicator and the second indicator, and determining whether there are the same indicators;

[0041] According to the first indicator under the same indicator, the corresponding clear sub-problem is traced to determine the first proportion problem. At the same time, according to the second indicator under the same indicator, the corresponding fuzzy sub-problem is traced to determine the second proportion problem.

[0042] Performing a correlation analysis on the first proportion question and the second proportion question to determine a dependency relationship between the corresponding second indicator and the corresponding first indicator;

[0043] The fuzzy sub-problems under the corresponding second indicator are clarified according to the dependency relationship and combined with the fuzzy probability, and the quality problem of the abnormal distribution position is obtained by combining the sub-problems under the remaining independent indicators.

[0044] Preferably, based on the quality issues of each supply link, a management and control optimization strategy for the corresponding supply link is determined, including:

[0045] Aggregate the quality issues in the supply chain, and perform normal distribution on the impact range of the quality issues involved in each aggregation result to obtain a concentrated impact range;

[0046] Extract the quality issues under any maximum impact range in each aggregation result to construct the auxiliary function to be optimized. At the same time, extract the quality issues that best match the corresponding concentrated impact range from each aggregation result to construct the main function to be optimized.

[0047] Solving the main function to be optimized and the auxiliary function to be optimized to obtain a quality optimization binary of each aggregation result;

[0048] Use the normal distribution range variance group of each aggregation result as a reference indicator, and match the analysis strategy of the corresponding aggregation result from the aggregation quality type-indicator-strategy comparison table;

[0049] The binary group is input into the analysis strategy to obtain an optimal function, and the optimal function is converted to obtain a management and control optimization strategy.

[0050] Compared with the prior art, the present invention has the following advantages:

[0051] By setting up a collection network and constructing a new set of standards under supply requirements, it is easy to identify existing quality problems and optimize management and control, ensuring comprehensive and real-time intelligent management and control of all supply links in the supply chain.

[0052] Other features and advantages of the present invention will be described in the following description, and in part will become apparent from the description, or will be understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description and the accompanying drawings.

[0053] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0055] Figure 1 This is a flow chart of a quality control method for a supply chain integrating a cloud platform and the industrial Internet in an embodiment of the present invention;

[0056] Figure 2 This is a structural diagram of a preset coordinate system in an embodiment of the present invention. DETAILED DESCRIPTION

[0057] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0058] The present invention provides a supply chain quality control method integrating cloud platform and industrial Internet, such as Figure 1 Shown, including:

[0059] Step 1: Based on the supply model of each supply link in the supply chain, deploy a data collection network to the corresponding supply link using industrial Internet technology, and transmit the collected data of each supply link to the cloud platform;

[0060] Step 2: Construct a new standard set based on the current supply demand of each supply link and the mapping relationship between the acquisition network and the preset risk items under the corresponding supply link;

[0061] Step 3: Analyze the collected data of the corresponding supply link based on the new standard set on the cloud platform, build a quality chain for the corresponding supply link, and compare and analyze it with the standard chain to identify quality issues, wherein the standard chain is constructed based on the minimum requirements of each new standard in the new standard set;

[0062] Step 4: Based on the quality issues of each supply link, determine the management and control optimization strategy of the corresponding supply link and optimize its deployment. At the same time, the quality issues are fed back to the management end of each supply link based on the cloud platform for reminder.

[0063] In this embodiment, the supply chain supply link refers to the business stage with independent functions in the supply chain, such as raw material procurement, production and processing, warehousing and logistics, sales services, etc. Taking the electronic equipment supply chain as an example, the supply links include: "chip procurement", "motherboard production", "finished product warehousing", "e-commerce sales", etc.

[0064] The supply model refers to the business operation mode of each link, including procurement model (such as JIT just-in-time procurement), production model (such as assembly line production), logistics model (such as third-party logistics TPL), etc.

[0065] Industrial Internet technology refers to the interconnection of physical devices and networks through sensors, RFID, embedded systems and other devices to collect real-time data. For example, temperature sensors on production lines and RFID shelf tags in warehouses are all Industrial Internet devices.

[0066] The collection network is composed of various sensors and communication modules, which are used to collect data from each link in real time. Different links involve different network components and are interdependent with the corresponding supply links themselves. For example, the raw material transportation link uses GPS sensors and humidity sensors to build a transportation status collection network.

[0067] Specifically: Select sensors according to the characteristics of the link: for example, deploy vibration sensors in the production link to monitor the operating status of equipment, and deploy temperature and humidity sensors in the warehousing link to monitor the storage environment.

[0068] Build a communication network: Use technologies such as 5G and LoRa to achieve wireless transmission of sensor data, or use industrial Ethernet to achieve wired connection.

[0069] Data access to cloud platforms: The collected data is pre-processed (such as denoising and normalization) through the edge computing gateway and then transmitted to the database of cloud platforms such as Alibaba Cloud and AWS.

[0070] In this embodiment, the current supply demand refers to the specific requirements of the current business scenario, such as order volume, delivery cycle, quality standards, etc. For example, during an e-commerce promotion period, "daily delivery order volume ≥ 100,000 orders" is the current supply demand in the logistics link.

[0071] Preset risk items refer to potential quality risk points set based on historical experience or industry standards, such as "excessive impurities in raw materials" and "transportation delays." For example, the preset risk items in the procurement of electronic components include "component voltage resistance not meeting standards" and "packaging damage rate."

[0072] The mapping relationship refers to the corresponding association between collected network data and risk items, such as "temperature sensor data" mapping to "risk of excessive temperature during transportation", for example, humidity sensor data in the warehousing link directly maps to "risk of goods getting damp".

[0073] The new standard set refers to a set of quality standards that have been dynamically optimized based on current needs, such as raw material acceptance standards, production process parameter thresholds, etc. For example, the new standard set includes "screen touch sensitivity threshold ≥95%" and "battery charging efficiency standard ≥98%".

[0074] In this embodiment, the quality chain is a chain that visualizes the quality data at each time point with the process of the corresponding link as the axis, and is used to show abnormal distribution. The standard chain is a theoretical quality baseline constructed based on the minimum requirements of the new standard set, which is used to compare the actual quality chain.

[0075] The initial anomaly factor is a quantitative indicator calculated by a formula, which is used to characterize the degree of deviation of data from the standard.

[0076] In this embodiment, abnormal distribution refers to the location and degree of abnormal points in the quality chain, such as "the yield rate of the third process in the production link drops sharply."

[0077] In this embodiment, the management and control optimization strategy is an improvement measure formulated for quality issues, such as equipment maintenance plan, process parameter adjustment, personnel training, etc. For example, "adjust the wire bonding machine pressure from 70g to 80g" and "increase the lens cleaning frequency to 2 times a day."

[0078] The management-side reminder pushes early warning information to the persons in charge of each link through the cloud platform, such as APP notifications and email reminders. For example, the warehouse management side receives the reminder that "the temperature in the cold storage exceeds the standard", triggering emergency cooling measures.

[0079] The beneficial effects of the above technical solution are: by setting up a collection network and constructing a new set of standards under supply demand, it is easy to identify existing quality problems and optimize management and control, thereby ensuring comprehensive and real-time intelligent management and control of all supply links in the supply chain.

[0080] The present invention provides a supply chain quality control method that integrates a cloud platform and the industrial Internet. A new standard set is constructed based on the current supply demand of each supply link and the mapping relationship between the acquisition network and the preset risk items in the corresponding supply link, including:

[0081] Extracting the business environment of each supply link in the supply chain under different historical supply demands from a historical database, analyzing the business environment to obtain a risk value for each preset risk item, and performing cluster analysis on the business environment to construct several risk matrices;

[0082] Standardizing and normalizing each risk matrix to obtain a characteristic vector of each processed matrix, wherein the characteristic vector includes a characteristic coefficient of each preset risk item;

[0083] Sort the characteristic coefficients in each eigenvector from large to small to obtain a size vector;

[0084] Based on the first element in the size vector, three elements are selected in sequence and mapped to a preset coordinate system for three-point connection drawing to determine the drawing shape;

[0085] assigning a first label to a first element of three consecutively screened elements depending on the drawn shape and in conjunction with a center point of the drawn shape;

[0086] Extract the labels under the same preset risk item from all first labels as second labels, and analyze the shape stability and numerical stability of the corresponding preset risk item based on all second labels;

[0087] Determining a first bias based on shape stability and a second bias based on numerical stability according to the drawn shapes and center points corresponding to each preset risk item under different risk matrices;

[0088] When the setting standard of the preset risk item under the corresponding supply demand does not match the first bias and the second bias, the corresponding setting standard is optimized based on the first bias and the second bias to obtain a new standard;

[0089] Match the current supply demand of each supply link with all historical demands of the same supply link. If there is a completely matching historical demand, a new standard set is obtained based on the mapping relationship between the acquisition network and the preset risk items of the corresponding supply link.

[0090] Otherwise, the incompletely matched historical demand with the largest matching coefficient is regarded as the reference demand;

[0091] Relying on the demand difference between the reference demand and the current supply demand, the demand difference is input into the demand analysis model to obtain the bias coefficient of each preset risk item, and the new standard based on the reference demand is optimized twice to obtain a new standard set.

[0092] In this embodiment, the historical database stores a database of past operation data of the supply chain, covering business information of each supply link in different periods, such as raw material procurement prices, production output, transportation time, etc. For example, the historical database of a certain automobile manufacturing company contains data such as the monthly parts procurement quantity, supplier delivery on-time rate, and number of production line equipment failures for the past five years.

[0093] In this embodiment, the historical supply demand refers to the business demand faced by supply chain operations in a certain period in the past, such as a specific order volume, delivery deadline, product quality requirements, etc. For example, in June last year, an e-commerce platform proposed a historical supply demand for the logistics link to process 1 million orders per day.

[0094] In this embodiment, the business environment refers to the internal and external conditions that affect the operation of the supply chain, including market demand, supplier status, production equipment status, etc. The preset risk items are pre-set based on experience or industry standards, and may affect the potential risks of the supply chain quality or normal operation.

[0095] In this embodiment, the environmental risk analysis model is obtained by training the neural network model based on different business environments and experts setting risk values ​​for possible preset risk items in the environment. Therefore, the risk value of each preset risk item can be directly obtained.

[0096] In this embodiment, the preset risk items are set in advance. It is only necessary to fill the corresponding risk points into the corresponding preset risk item filling positions to obtain a matrix of corresponding clustering results. The columns of the matrix correspond to the same preset risk items, and the behaviors correspond to all preset risk items in the environmental scenario.

[0097] In this embodiment, standardization and normalization processes transform the data to have a uniform scale or distribution. Standardization generally converts the data into a standard normal distribution with a mean of 0 and a standard deviation of 1; normalization maps the data to the interval [0, 1].

[0098] In this embodiment, the eigenvector is a vector that can reflect the core characteristics of the risk matrix after processing, and contains the characteristic coefficient of each preset risk item. For example, in the eigenvector of a certain risk matrix, the characteristic coefficient of the "raw material price fluctuation" risk item is 0.7, indicating that it accounts for a large proportion of the matrix characteristics.

[0099] In this embodiment, the size vector is a vector obtained by arranging the characteristic coefficients in the characteristic vector in descending order. For example, after sorting a certain characteristic vector, the size vector [0.8, 0.6, 0.3, 0.2] is obtained, which corresponds to the characteristic coefficients of different preset risk items respectively. The first three elements (0.8, 0.6, 0.3) are selected and a triangle is drawn in the preset coordinate system. Assuming that the size vector is [0.5, 0.5, 0.5, 0.2], a straight line is drawn in the element and coordinate system. It should be noted that the horizontal coordinate of the preset coordinate system is the number of elements, and the vertical coordinate is the characteristic coefficient, such as Figure 2 shown.

[0100] In this embodiment, the first label is an identifier assigned to the first element of the three consecutively screened elements based on the drawn shape and the center point, and the second label is a set of labels extracted from all the first labels and belonging to the same preset risk item.

[0101] In this embodiment, shape stability is to evaluate the similarity of the shapes drawn by the preset risk items under different risk matrices, reflecting their changing patterns. If the shapes drawn by the "transportation delay" risk item under multiple risk matrices are similar, it means that its shape stability is high. Specifically, the shape stability is evaluated by calculating the shape similarity (such as the Hausdorff distance), and the numerical stability is evaluated by calculating the variance index of the characteristic coefficient.

[0102] Numerical stability is a measure of the degree of numerical fluctuation of the characteristic coefficient of the preset risk item under different risk matrices. For example, the characteristic coefficient of the "equipment failure" risk item fluctuates less in different risk matrices, indicating that its numerical stability is high.

[0103] The first bias is the direction of adjustment of the preset risk item standard based on shape stability, such as whether the standard is increased, decreased, or remains unchanged. The second bias is the direction of adjustment of the preset risk item standard based on numerical stability. For example, the standard for "product defective rate" is set at 5%. After analyzing its shape and numerical stability, it is shown that the risk has an upward trend. The first bias and the second bias are both directions of standard increase. According to the mismatch factor calculation, the standard is adjusted to 3% to obtain a new standard.

[0104] In this embodiment, the set standard is the quality or operation standard currently set for the preset risk item.

[0105] In this embodiment, the new standard is a new quality or operation standard obtained by optimizing the set standard according to the first bias and the second bias. For example, adjusting the "transportation time" setting standard from 24 hours to 18 hours is the new standard. The matching coefficient is a numerical value that measures the similarity between the current supply demand and the historical demand, which is calculated through the similarity function.

[0106] Reference demand is the incompletely matched historical demand with the largest matching coefficient with the current supply demand. Demand differences include the differences between the current supply demand and the reference demand in terms of order quantity, delivery time, product specifications, etc.

[0107] In this embodiment, a demand parsing model based on a neural network is used as input, with different demand differences as input samples. The input samples are the bias coefficients of each preset risk item corresponding to the demand differences. The bias coefficients reflect the degree to which the demand differences affect the preset risk item standards. The new standard set is a collection of all new standards for preset risk items obtained after combining primary and secondary optimization for each supply link. The demand parsing model calculates a bias coefficient of 0.2 for the risk item "special specifications for raw materials supply." Combined with the new standard for the reference demand, this risk item standard is further adjusted to form a new standard set suitable for current demand.

[0108] The beneficial effects of the above technical solution are: by analyzing historical data and current demand, the quality standards of each link in the supply chain can be quickly adjusted to make the standards more in line with actual business conditions. Based on the risk matrix and stability analysis, the optimization standards can effectively identify potential risks and take measures in advance to reduce the probability and impact of risks. The formulation of a new set of standards will help each supply link to operate more efficiently, reduce resource waste, and provide data support for supply chain management.

[0109] The present invention provides a quality control method for a supply chain integrating a cloud platform and the industrial Internet, which relies on the first bias and the second bias to optimize the corresponding set standards to obtain new standards, including:

[0110] respectively obtaining a first mismatch factor and a second mismatch factor of the first bias and the second bias with a set standard under corresponding supply requirements;

[0111] Determining an optimization direction for a corresponding set standard according to the first mismatch factor and the second mismatch factor, wherein the optimization direction includes a standard improvement direction, a standard reduction direction, and a standard unchanged direction;

[0112] Adjust the corresponding data setting standards according to the optimization direction.

[0113] In this embodiment, the first mismatch factor is a value that quantifies the degree of mismatch between the first bias and the set standard under the corresponding supply demand.

[0114] For example, if the first bias of the "transportation delay" risk item is in the direction of standard improvement, and the transportation delay rate is high under the current set standard, the first mismatch factor is calculated to be 0.6 (a larger value indicates a higher degree of mismatch).

[0115] The second mismatch factor is a quantitative indicator that measures the degree to which the second bias does not match the set standard.

[0116] For example, for the risk item "damaged product packaging," the second bias indicates that the standard needs to be improved, but the existing set standard fails to effectively control the damage rate, and the calculated second mismatch factor is 0.5.

[0117] The optimization direction is the adjustment direction of the set standard determined based on the mismatch factor, including the standard improvement direction (improving standard requirements), the standard reduction direction (reducing standard requirements), and the standard unchanged direction (maintaining the existing standard).

[0118] For example, if the first and second mismatch factors of a risk item are both high, the optimization direction is to improve the standard; if the mismatch factor is close to 0, the optimization direction is to keep the standard unchanged.

[0119] In this embodiment, the first mismatch factor F1 is calculated according to the following formula: F1 = (T1 - S1) / S1, where T1 is the risk value estimated based on the shape trend, and S1 is the risk value corresponding to the set standard.

[0120] The second mismatch factor F2 is calculated according to the following formula: F2 = (T2-S2) / S2, where T2 is the risk value estimated based on numerical stability, and S2 is the risk value corresponding to the set standard.

[0121] When F1≥0.1 and F2≥0.1, the optimization direction is determined to be the standard improvement direction. When 0<F1<0.1 and 0<F2<0.1, the optimization direction is determined to be the standard unchanged direction. Otherwise, the optimization direction is determined to be the standard reduction direction.

[0122] The beneficial effect of the above technical solution is: by quantifying the mismatch factor and accurately determining the optimization direction, the set standards are more in line with the actual risk situation of the supply chain, effectively solving the problem of lagging or unreasonable standards.

[0123] The present invention provides a quality control method for a supply chain integrating a cloud platform and the industrial Internet, which performs secondary optimization on a new standard based on the reference requirements, including:

[0124] Considering the new standard of the reference requirement as the first standard;

[0125] Obtain a bias coefficient for each preset risk item, and perform secondary optimization on the corresponding first standard in combination with the item weight of the preset risk item based on the reference demand.

[0126] In this embodiment, the item weight is a weight value assigned according to the importance of the preset risk item to the overall quality or operation of the supply chain, and is used to measure the relative importance of each risk item in the standard optimization.

[0127] In this embodiment, the secondary optimization is obtained by using Sc×(1+w0×b0), where Sc is the threshold corresponding to the first criterion, w0 is the item weight, and b0 is the bias coefficient.

[0128] The beneficial effect of the above technical solution is: by combining the bias coefficient and the item weight to perform secondary optimization on the first standard, the quality standard can more accurately match the current complex and changeable supply demand, avoiding resource waste or quality problems.

[0129] The present invention provides a supply chain quality control method that integrates a cloud platform and the industrial Internet, and constructs a quality chain for the corresponding supply link, including:

[0130] Comparing and analyzing each new standard in the new standard set with corresponding sub-data in the collected data of the corresponding supply link to determine an initial anomaly factor, and constructing an initial vector based on the setting position of the corresponding new standard in the corresponding supply link, and inputting the initial vector into a vector analysis model to obtain an output representation based on the corresponding new standard, wherein the output representation includes: anomaly distribution and the degree of anomaly at each anomaly distribution position;

[0131] Obtain the blank chain corresponding to the supply link from the chain database;

[0132] The output representation of each new standard under the corresponding supply link is input into the blank chain in turn to obtain a quality chain.

[0133] Preferably, determining the initial abnormality factor includes:

[0134]

[0135] in, represents the initial abnormal factor; Indicates the quantization result of the corresponding sub-data; Indicates the quantitative threshold corresponding to the new standard; represents the inverse tangent function; Indicates the set weight corresponding to the new standard; represents a Bernoulli random variable; represents a symbolic function; Represents the contrast sensitivity coefficient.

[0136] In this embodiment, The threshold value is , The value of is 1.

[0137] In this embodiment, The value of is ±1, .

[0138] In this example, the inverse tangent function initially rises rapidly (sensitive to small deviations) and then flattens out (saturates to large deviations). This aligns with the practical logic that small deviations in the supply chain require prompt correction, while large deviations are often systemic issues (requiring separate investigation). For example, if the production temperature standard is 25°C, and R increases from 25°C to 26°C (a small deviation), the tan−1 output rises rapidly, providing a sensitive warning of the abnormal factor Q. However, if R increases from 26°C to 35°C (a large deviation), the tan−1 output approaches saturation, indicating the need for downtime and maintenance rather than simply parameter adjustment. This aligns with the tiered response requirements of on-site management.

[0139] The output Q not only has a deviation quantization value (such as 0.3 represents moderate anomaly), but also Reflect the direction of deviation (positive / negative, corresponding to "exceeding the standard" or "not meeting the standard"), through Implicit random fluctuation information. These dimensions enrich the description of anomalies at quality chain nodes (e.g., "Process 3: Q = 0.2, positive deviation, including random fluctuations"). Subsequent analysis can quickly determine whether it is a persistent process issue (negative deviation and no random items) or an occasional equipment fluctuation (positive deviation and random items), improving quality traceability efficiency.

[0140] In this embodiment, the corresponding sub-data is data in the collected data that is directly associated with a new standard.

[0141] The initial anomaly factor quantification is an indicator that reflects the degree of deviation of the sub-data from the new standard.

[0142] In this embodiment, the setting location is the corresponding node of the new standard in the supply link process (such as the "third process" in the production link).

[0143] The initial vector is a vector consisting of the initial anomaly factor and the setting position, which is used for model input. The initial anomaly factor setting position is encoded (such as [0.33, 5, ...], where "5" represents the 5th process).

[0144] The vector analysis model is obtained by training a neural network model with different vectors as output and the position distribution and degree of abnormality of abnormal elements in the vectors as output, and can be obtained directly.

[0145] In this embodiment, the abnormal distribution is the spatial / process distribution of abnormal data in the supply link. For example, in the production link, "the abnormality rate of the 3rd and 5th processes is high", showing a distribution of "the processes are dispersed but the key nodes are concentrated".

[0146] The abnormality level is a quantitative description of the severity of the abnormality (between 0 and 1, with 1 being a severe abnormality).

[0147] The chain database is a database that stores blank chain templates and historical quality chain data of each supply link. A blank chain is a chain structure with no actual data and only a link process framework (such as an empty template in which the production links are arranged as "process 1-process 2-..."). For example, the automobile manufacturing chain database contains blank chains for links such as machining, painting, and assembly.

[0148] In this embodiment, the quality chain is a chain that fully reflects the quality status of the supply chain after the blank chain is filled with the new standard output representation (abnormality distribution, abnormality degree).

[0149] The beneficial effects of the above technical solution are: by combining the initial abnormality factor with the vector model to quantify the degree of abnormality and locate the distribution, the ambiguity of traditional empirical judgment is resolved. The quality chain integrates scattered standards and data into a process-based view, allowing managers to quickly grasp the quality status of each link, realize accurate diagnosis and management of supply chain quality, improve the efficiency of abnormality handling, and reduce quality risks.

[0150] The present invention provides a quality control method for a supply chain that integrates a cloud platform and the industrial Internet. The method compares and analyzes the supply chain with a standard supply chain to identify quality issues, including:

[0151] Acquire an abnormal combination of abnormal distribution positions based on the quality chain;

[0152] Extracting quality analysis indicators related to the lowest standard combination corresponding to the abnormal distribution position from the standard chain, and extracting an analysis model from the indicator-model database to analyze the abnormal combination, and determining a clear sub-problem and a fuzzy sub-problem corresponding to the abnormal distribution position, wherein the fuzzy sub-problem is between clear and unclear and has a fuzzy probability;

[0153] Obtaining a first indicator for comparison with the clear sub-problem and a second indicator for comparison with the fuzzy sub-problem, performing same-indicator extraction on the first indicator and the second indicator, and determining whether there are the same indicators;

[0154] According to the first indicator under the same indicator, the corresponding clear sub-problem is traced to determine the first proportion problem. At the same time, according to the second indicator under the same indicator, the corresponding fuzzy sub-problem is traced to determine the second proportion problem.

[0155] Performing a correlation analysis on the first proportion question and the second proportion question to determine a dependency relationship between the corresponding second indicator and the corresponding first indicator;

[0156] The fuzzy sub-problems under the corresponding second indicator are clarified according to the dependency relationship and combined with the fuzzy probability, and the quality problem of the abnormal distribution position is obtained by combining the sub-problems under the remaining independent indicators.

[0157] In this embodiment, an abnormal combination is a set of multiple abnormal factors at a specific abnormal distribution location (such as a certain process or node). For example, an abnormal combination in the tire assembly process may include "insufficient bolt torque" (abnormal factor A), "uneven tire pressure" (abnormal factor B), etc.

[0158] The standard chain is a structured data chain that stores the minimum quality standards for each link, corresponding to each node in the quality chain. For example, the standard chain node of the tire assembly process includes minimum standards such as "bolt torque ≥ 100N·m" and "tire pressure 2.2±0.1bar".

[0159] Quality analysis indicators are quantitative parameters used to evaluate anomalies, such as yield, failure rate, temperature deviation, etc.

[0160] Clear sub-problems are abnormal problems whose causes can be directly identified through indicators, without ambiguity. For example, if the measured bolt torque value is 80N·m (lower than the standard 100N·m), it is directly judged as "insufficient torque".

[0161] Fuzzy subproblems are abnormal problems with multiple possible causes that require probabilistic judgment. Uncertainty is expressed using fuzzy probability. For example, if tire pressure fluctuates between 2.0 and 2.4 bar (the standard is 2.2 ± 0.1 bar), possible causes include "inflation equipment error" (probability 0.6) or "tire air tightness problem" (probability 0.4).

[0162] The first indicator and the second indicator correspond to the analysis indicators of the clear sub-problem and the fuzzy sub-problem respectively. For example, the first indicator "bolt torque" corresponds to the clear sub-problem; the second indicator "air pressure fluctuation range" corresponds to the fuzzy sub-problem.

[0163] Dependency relationship: The causal probabilistic relationship between a fuzzy subproblem and a crisp subproblem. For example, "inflator error" (fuzzy subproblem) is 70% dependent on "air pressure sensor calibration bias" (clean subproblem).

[0164] The first proportion problem is to clarify the contribution of each cause in the sub-problem. For example, in the problem of insufficient bolt torque, "tool calibration error" accounts for 60% and "worker operation error" accounts for 40%.

[0165] The second proportion problem is the probability distribution of each possible cause in the fuzzy sub-problem. For example, in the tire pressure fluctuation problem, the probability of "inflating equipment error" is 0.6, and the probability of "tire air tightness problem" is 0.4.

[0166] In this embodiment, the indicator-model database includes analysis models under different indicator combinations. For example, the regression model is pre-set and stored in the database. For example, the "solder paste thickness-printing pressure" regression model (Y=0.05X+0.1) in the database is called to analyze the relationship between thickness and pressure.

[0167] In this embodiment, an indicator comparison table of the clear sub-problem (C) and the fuzzy sub-problem (F) is constructed, and a common indicator (Icommon=C∩F) is identified based on the comparison indicator set.

[0168] Clarify sub-question C: "Solder paste thickness deviation" (metrics: thickness, printing pressure).

[0169] Fuzzy sub-problem F: "High solder joint void rate" (indicators: thickness, reflow temperature).

[0170] Operation: Compare and get the common indicator "solder paste thickness".

[0171] Result: "Solder paste thickness" is determined to be the same indicator, and its impact on the two issues needs to be further analyzed.

[0172] In this embodiment, a causal graph model (such as a structural equation model) is constructed to quantify the dependencies between indicators.

[0173] Calculate the conditional probability P(F|C) to determine the dependence strength of the fuzzy subproblem on the clear subproblem.

[0174] 100 sets of data were collected and it was found that the occurrence rate of F was 80% when C occurred and 20% when C did not occur. P(F|C)=0.8 was calculated, which means that 80% of solder joint voids depend on solder paste thickness deviation.

[0175] Result: Dependency relationship F0.8C is established.

[0176] In this embodiment, based on the dependency relationship, the fuzzy sub-problems are decomposed into:

[0177] Dependent part: Fdependent=P(F|C)×F.

[0178] Independent part: Findependent=F−Fdependent.

[0179] For dependent parts, directly relate the causes of the clear sub-problems; for independent parts, analyze them separately.

[0180] The beneficial effects of the above technical solution are: through indicator association and probabilistic reasoning, fuzzy problems are transformed into clear and traceable causes, and dependency association analysis directly points to key indicators, reducing trial and error costs. The established indicator dependency relationship can be used to predict potential problems.

[0181] The present invention provides a supply chain quality control method that integrates a cloud platform and the industrial Internet. The method determines the control optimization strategy for the corresponding supply link based on the quality issues of each supply link, including:

[0182] Aggregate the quality issues in the supply chain, and perform normal distribution on the impact range of the quality issues involved in each aggregation result to obtain a concentrated impact range;

[0183] Extract the quality issues under any maximum impact range in each aggregation result to construct the auxiliary function to be optimized. At the same time, extract the quality issues that best match the corresponding concentrated impact range from each aggregation result to construct the main function to be optimized.

[0184] Solving the main function to be optimized and the auxiliary function to be optimized to obtain a quality optimization binary of each aggregation result;

[0185] Use the normal distribution range variance group of each aggregation result as a reference indicator, and match the analysis strategy of the corresponding aggregation result from the aggregation quality type-indicator-strategy comparison table;

[0186] The binary group is input into the analysis strategy to obtain an optimal function, and the optimal function is converted to obtain a management and control optimization strategy.

[0187] In this embodiment, the K-Means clustering algorithm is used to aggregate quality issues by "process + type", and the impact range of each type of problem (such as the affected product batches and number of customers) is counted. Python is used to fit the normal distribution, and μ and σ are output. For example, five categories are aggregated, such as "screen problems" (200 items) and "mainboard problems" (300 items). The number of orders affected by "screen problems" is fitted, and μ=95, σ=15 is obtained. The normal distribution R2=0.92 (high fit). The result is a clear distinction between the concentrated impact boundaries of the problems (68% of the problems affect 80-110 orders).

[0188] In this embodiment, the auxiliary function is to filter out the instances with the largest impact in the aggregation problem (such as screen bubbles that affect 300 orders in a batch), define the goal as "contain the spread within 12 hours", and the variables are "temporary sampling frequency" and "manual rework investment" to construct a linear function.

[0189] The main function is to select typical examples with an impact range close to μ (such as regular bubbles that affect 95 orders), define the target as "defective rate from 5% to 1% within 30 days", and use the variables "equipment upgrade budget" and "process adjustment range" to construct a nonlinear function (including equipment cost and yield-related items).

[0190] In this embodiment, a genetic algorithm (GA) or a gradient descent method is used to solve the two functions and find the optimal solution.

[0191] Calculate the range variance group of the aggregated problem (e.g., the screen problem variance σ2=225), and query the comparison table for matching strategies (large variance → select the "quick response + fluctuation control" strategy, such as the PDCA cycle).

[0192] The binary data is input into the policy model, and the decision tree algorithm is used to generate the optimal execution path (for example, "perform expedited spot checks on days 1-3, initiate equipment assessment on day 4, and determine the upgrade plan on day 10").

[0193] Input a binary pair into the PDCA strategy model and output:

[0194] P (Plan): Develop sampling inspection procedures and equipment upgrade plans within 3 days.

[0195] D (Execution): Start random inspections on the 4th day and pilot new equipment on the 15th day.

[0196] C (Inspection): Evaluate the inspection results and equipment yield on the 20th day.

[0197] A (Improvement): Optimize the plan on the 25th day and fully promote it on the 30th day.

[0198] Result: The management and control strategy can be quantified down to the day and is highly implementable.

[0199] In this embodiment, the quality optimization binary, for example, after solving two functions, obtains a combination of "short-term response plan + long-term cure plan" (such as "increasing the full inspection link (auxiliary) + upgrading the bonding equipment (main)").

[0200] In this embodiment, the range variance group is used to describe the degree of dispersion of the influence of different aggregation issues in the normal distribution (large variance → large fluctuation of influence).

[0201] The aggregate quality type-indicator-strategy table is a predefined rule base that associates problem types (such as assembly / welding), indicators (variance groups) and analysis strategies (such as Six Sigma, PDCA).

[0202] The beneficial effect of the above technical solution is: through the closed loop of aggregation-quantification-function solving-strategy matching, quality problems are upgraded from scattered responses to system optimization, providing a quantifiable and executable technical path for quality control.

[0203] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.

Claims

1. A quality control method for a supply chain integrating a cloud platform and the industrial Internet, characterized in that: include: Step 1: Based on the supply model of each supply link in the supply chain, deploy a data collection network to the corresponding supply link using industrial Internet technology, and transmit the collected data of each supply link to the cloud platform; Step 2: Construct a new standard set based on the current supply demand of each supply link and the mapping relationship between the acquisition network and the preset risk items under the corresponding supply link; Step 3: Analyze the collected data of the corresponding supply link based on the new standard set on the cloud platform, build a quality chain for the corresponding supply link, and compare and analyze it with the standard chain to identify quality issues, wherein the standard chain is constructed based on the minimum requirements of each new standard in the new standard set; Step 4: Based on the quality issues in each supply link, determine the corresponding supply link management and control optimization strategy and optimize its deployment. At the same time, the quality issues are fed back to the management end of each supply link through the cloud platform for reminder. The construction of the quality chain of the corresponding supply link includes: Comparing and analyzing each new standard in the new standard set with corresponding sub-data in the collected data of the corresponding supply link to determine an initial anomaly factor, and constructing an initial vector based on the setting position of the corresponding new standard in the corresponding supply link, and inputting the initial vector into a vector analysis model to obtain an output representation based on the corresponding new standard, wherein the output representation includes: anomaly distribution and the degree of anomaly at each anomaly distribution position; Obtain the blank chain corresponding to the supply link from the chain database; Input the output representation of each new standard under the corresponding supply link into the blank chain in turn to obtain a quality chain; Determine the initial abnormal factors, including: ;in, represents the initial abnormal factor; R represents the quantization result of the corresponding sub-data; r represents the quantization threshold corresponding to the new standard; represents the inverse tangent function; Indicates the set weight corresponding to the new standard; represents a Bernoulli random variable; represents a symbolic function; Represents the contrast sensitivity coefficient.

2. The quality control method for the supply chain integrating the cloud platform and the industrial Internet according to claim 1 is characterized in that: According to the current supply demand of each supply link and the mapping relationship between the acquisition network and the preset risk items under the corresponding supply link, a new standard set is constructed, including: Extracting the business environment of each supply link in the supply chain under different historical supply demands from a historical database, analyzing the business environment to obtain a risk value for each preset risk item, and performing cluster analysis on the business environment to construct several risk matrices; Standardizing and normalizing each risk matrix to obtain a characteristic vector of each processed matrix, wherein the characteristic vector includes a characteristic coefficient of each preset risk item; Sort the characteristic coefficients in each eigenvector from large to small to obtain a size vector; Based on the first element in the size vector, three elements are sequentially selected and mapped to a preset coordinate system for three-point connection drawing to determine the drawing shape; assigning a first label to a first element of three consecutively screened elements depending on the drawn shape and in conjunction with a center point of the drawn shape; Extract the labels under the same preset risk item from all first labels as second labels, and analyze the shape stability and numerical stability of the corresponding preset risk item based on all second labels; Determining a first bias based on shape stability and a second bias based on numerical stability according to the drawn shapes and center points corresponding to each preset risk item under different risk matrices; When the setting standard of the preset risk item under the corresponding supply demand does not match the first bias and the second bias, the corresponding setting standard is optimized based on the first bias and the second bias to obtain a new standard; Match the current supply demand of each supply link with all historical demands of the same supply link. If there is a completely matching historical demand, a new standard set is obtained based on the mapping relationship between the acquisition network and the preset risk items of the corresponding supply link. Otherwise, the incompletely matched historical demand with the largest matching coefficient is regarded as the reference demand; Relying on the demand difference between the reference demand and the current supply demand, the demand difference is input into the demand analysis model to obtain the bias coefficient of each preset risk item, and the new standard based on the reference demand is optimized twice to obtain a new standard set.

3. The quality control method for the supply chain integrating the cloud platform and the industrial Internet according to claim 2 is characterized in that: Relying on the first bias and the second bias to optimize the corresponding set standard to obtain a new standard, including: respectively obtaining a first mismatch factor and a second mismatch factor of the first bias and the second bias with a set standard under corresponding supply requirements; Determining an optimization direction for a corresponding set standard according to the first mismatch factor and the second mismatch factor, wherein the optimization direction includes a standard improvement direction, a standard reduction direction, and a standard unchanged direction; Adjust the corresponding number setting standards according to the optimization direction.

4. The quality control method for the supply chain integrating the cloud platform and the industrial Internet according to claim 2 is characterized in that: Secondary optimization of the new standard based on the reference requirements, including: Considering the new standard of the reference requirement as the first standard; Obtain a bias coefficient for each preset risk item, and perform secondary optimization on the corresponding first standard in combination with the item weight of the preset risk item based on the reference demand.

5. The quality control method for the supply chain integrating the cloud platform and the industrial Internet according to claim 1 is characterized in that: Comparative analysis with standard chains to identify quality issues, including: Acquire an abnormal combination of abnormal distribution positions based on the quality chain; Extracting quality analysis indicators related to the lowest standard combination corresponding to the abnormal distribution position from the standard chain, and extracting an analysis model from the indicator-model database to analyze the abnormal combination, and determining a clear sub-problem and a fuzzy sub-problem corresponding to the abnormal distribution position, wherein the fuzzy sub-problem is between clear and unclear and has a fuzzy probability; Obtaining a first indicator for comparison with the clear sub-problem and a second indicator for comparison with the fuzzy sub-problem, performing same-indicator extraction on the first indicator and the second indicator, and determining whether there are the same indicators; According to the first indicator under the same indicator, the corresponding clear sub-problem is traced to determine the first proportion problem. At the same time, according to the second indicator under the same indicator, the corresponding fuzzy sub-problem is traced to determine the second proportion problem. Performing a correlation analysis on the first proportion question and the second proportion question to determine a dependency relationship between the corresponding second indicator and the corresponding first indicator; The fuzzy sub-problems under the corresponding second indicator are clarified according to the dependency relationship and combined with the fuzzy probability, and the quality problem of the abnormal distribution position is obtained by combining the sub-problems under the remaining independent indicators.

6. The quality control method for the supply chain integrating the cloud platform and the industrial Internet according to claim 1 is characterized in that: Determine the management and control optimization strategy for each supply link based on the quality issues of each supply link, including: Aggregate the quality issues in the supply chain, and perform normal distribution on the impact range of the quality issues involved in each aggregation result to obtain a concentrated impact range; Extract the quality issues under any maximum impact range in each aggregation result to construct the auxiliary function to be optimized. At the same time, extract the quality issues that best match the corresponding concentrated impact range from each aggregation result to construct the main function to be optimized. Solving the main function to be optimized and the auxiliary function to be optimized to obtain a quality optimization binary of each aggregation result; Use the normal distribution range variance group of each aggregation result as a reference indicator, and match the analysis strategy of the corresponding aggregation result from the aggregation quality type-indicator-strategy comparison table; The binary group is input into the analysis strategy to obtain an optimal function, and the optimal function is converted to obtain a management and control optimization strategy.

Images