Product quality defect analysis method, equipment and medium

Abstract

The invention discloses a product quality defect analysis method and device and a medium, and relates to the technical field of data processing. The method comprises the following steps: firstly, establishing a prediction model from a raw material performance index to a product quality index based on historical data by using a random forest algorithm; when product defects are detected, initial raw material data of the batch of products are obtained and subjected to multi-round random disturbance, a large number of virtual samples are generated, and the quality of the virtual samples is predicted through a model. Then, qualified samples are screened out from the virtual samples, and a sample subset with the highest model prediction reliability is further selected from the samples. And finally, performing statistics on the consistency of the disturbance direction of each raw material performance index in the subset, and judging the raw material index with the highest consistency as a key factor which is most likely to cause the current quality defect. Noise in industrial data can be effectively processed, product quality problems can be quickly and accurately attributed to specific raw material characteristics, and accurate guidance is provided for production control and process optimization.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to a method, device and medium for analyzing product quality defects. Background Technology

[0002] In complex modern industrial production, the quality of the final product is often determined by the performance indicators of dozens or even hundreds of raw materials. When quality defects occur (such as a failure to meet certain indicators), traditional manual experience analysis or simple statistical methods struggle to quickly and accurately pinpoint the root cause from a large amount of interrelated and potentially noisy raw material data. This leads to long troubleshooting cycles, high costs, and may even result in incorrect corrective measures due to misjudgment, causing even greater waste of resources. Summary of the Invention

[0003] This application provides a method, device, and medium for analyzing product quality defects. The main purpose is to solve the problem that the existing technology, which relies on manual experience to analyze which raw material performance indicators cause product quality defects, is inefficient.

[0004] In a first aspect, embodiments of this application provide a method for analyzing product quality defects. The method includes: when any quality indicator of a target product is detected to exceed the acceptable quality range, obtaining an initial raw material performance indicator vector corresponding to the product batch to which the target product belongs, and subjecting the initial raw material performance indicator vector to multiple rounds of random perturbation to obtain a raw material performance indicator vector after multiple perturbations. Based on the random forest model, the vector of raw material performance indicators after multiple perturbations is predicted to obtain multiple predicted values ​​and the standard deviation of each predicted value. The predicted values ​​are used to describe product quality indicators. Based on the predicted value and the quality qualified range, a first sample set of qualified raw material performance indexes is selected from the multiple perturbed raw material performance index vectors, and a preset number of perturbed raw material performance index vectors with the highest reliability ranking are selected from the first sample set based on the standard deviation, to obtain a second sample set. The perturbation direction of each raw material performance index in the perturbed raw material performance index vector of each item in the second sample set is statistically analyzed to obtain the consistency of the perturbation direction of each raw material performance index. The raw material performance index with the highest consistency in the direction of disturbance was identified as the key factor most likely to cause this quality defect.

[0005] In one implementation of this application, the step of subjecting the initial raw material performance index vector to multiple rounds of random perturbation to obtain a multi-perturbation raw material performance index vector includes: A random value is added to the initial value of each raw material performance index in the initial raw material performance index vector to obtain the perturbed raw material performance index vector. For each of the raw material performance indicators, a corresponding random value is randomly selected and added to its initial value for a second round of random perturbation. This process is repeated for N rounds of random perturbation to obtain N perturbed raw material performance indicator vectors. The random value corresponding to each of the raw material performance indicators follows a normal distribution with a mean of 0.

[0006] In one implementation of this application, the step of selecting a first sample set of qualified raw materials from the multiple perturbated raw material performance index vectors based on the predicted value and the qualified quality range includes: Filter out the perturbed raw material performance index vector corresponding to the predicted value that belongs to the quality qualified range; The first sample set is constructed based on the selected perturbation-induced raw material performance index vector.

[0007] In one implementation of this application, the step of selecting a predetermined number of perturbed raw material performance index vectors with high reliability ranking from the first sample set based on the standard deviation to obtain the second sample set includes: The reliability of the perturbed raw material performance index vector in the first sample set is sorted in ascending order based on the standard deviation. The smaller the standard deviation, the higher the reliability of the perturbed raw material performance index vector. Selecting a predetermined number of the perturbed raw material performance index vectors from the ascending order yields the second sample set.

[0008] In one implementation of this application, the consistency of the perturbation direction of each raw material performance index in the perturbed raw material performance index vector of each item in the second sample set is obtained by statistically analyzing the perturbation direction of each raw material performance index, including: For each of the perturbed raw material performance index vectors in the second sample set, the perturbation value corresponding to each raw material performance index is obtained by traversing the vector. The perturbation value is the sum of the corresponding random value and the initial value. Compare the magnitudes of all perturbation values ​​corresponding to the same raw material performance index in the different perturbation raw material performance index vectors with their corresponding initial values. If the disturbance value is greater than the initial value, a first marker indicating a larger disturbance direction is output; if the disturbance value is less than the initial value, a second marker indicating a smaller disturbance direction is output; if the disturbance value is equal to the initial value, an ignore marker is output. Count the first number of the first marker, the second number of the second marker, and the third number of the ignored markers, and add the first number, the second number, and the third number to calculate the total number; Divide the target difference by the total quantity to obtain the target quotient. Use the absolute value of the target quotient as the perturbation direction consistency score for the same raw material performance index. Repeat this process to obtain the perturbation direction consistency score for each raw material performance index.

[0009] In one implementation of this application, the raw material performance index with the highest consistency in disturbance direction is identified as the key factor most likely to cause this quality defect, including: The raw material performance index corresponding to the largest perturbation direction consistency score is taken as the key factor. The corresponding raw material performance indicators are adjusted according to the positive or negative nature of the target quotient corresponding to the maximum perturbation direction consistency score, so as to use raw materials that meet the adjusted performance indicators to produce the target product.

[0010] In one implementation of this application, after taking the raw material performance index corresponding to the largest perturbation direction consistency score as the key factor, the method includes: If the positive or negative property is positive, the raw material performance index is increased; otherwise, the raw material performance index is decreased.

[0011] In one implementation of this application, before predicting the multiple perturbed raw material performance index vectors based on a random forest model, the method includes: Obtain historical production data for the target product, including historical raw material performance indicators and historical product quality indicators; The historical production data is preprocessed, and a random forest regression model is trained separately for each product quality indicator based on the preprocessed historical production data to obtain the random forest model.

[0012] Secondly, embodiments of this application also provide a product quality defect analysis device, the device comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to: when detecting that any quality indicator of a target product exceeds the acceptable quality range, obtain an initial raw material performance indicator vector corresponding to the product batch to which the target product belongs, and subject the initial raw material performance indicator vector to multiple rounds of random perturbation to obtain multiple perturbed raw material performance indicator vectors; predict the multiple perturbed raw material performance indicator vectors based on a random forest model to obtain multiple The predicted value and the standard deviation corresponding to each predicted value are used to describe the product quality indicators. Based on the predicted value and the quality acceptable range, a first sample set of qualified raw material performance indicators is selected from the multiple perturbed raw material performance indicator vectors. Based on the standard deviation, a predetermined number of perturbed raw material performance indicator vectors with the highest reliability ranking are selected from the first sample set to obtain a second sample set. The perturbation direction of each raw material performance indicator in the perturbed raw material performance indicator vectors in the second sample set is statistically analyzed to obtain the consistency of the perturbation direction of each raw material performance indicator. The raw material performance indicator with the highest consistency of perturbation direction is determined as the key factor most likely to cause this quality defect.

[0013] Thirdly, embodiments of this application also provide a non-volatile computer storage medium for analyzing product quality defects, storing computer-executable instructions. These instructions are configured to: when any quality indicator of a target product is detected to exceed a acceptable quality range, obtain an initial raw material performance indicator vector corresponding to the product batch to which the target product belongs, and subject the initial raw material performance indicator vector to multiple rounds of random perturbation to obtain multiple perturbed raw material performance indicator vectors; predict the multiple perturbed raw material performance indicator vectors based on a random forest model to obtain multiple predicted values ​​and a standard deviation corresponding to each predicted value, where the predicted values ​​describe product quality indicators; based on the predicted values ​​and the acceptable quality range, select a first sample set of acceptable quality from the multiple perturbed raw material performance indicator vectors, and based on the standard deviation, select a preset number of perturbed raw material performance indicator vectors with high reliability ranking from the first sample set to obtain a second sample set; statistically analyze the perturbation direction of each raw material performance indicator in each perturbed raw material performance indicator vector in the second sample set to obtain the consistency of the perturbation direction of each raw material performance indicator; and determine the raw material performance indicator with the highest consistency in the perturbation direction as the key factor most likely to cause this quality defect.

[0014] This application provides a method, device, and medium for analyzing product quality defects, comprising: when any quality indicator of a target product is detected to exceed the acceptable quality range, obtaining an initial raw material performance indicator vector corresponding to the product batch to which the target product belongs, and subjecting the initial raw material performance indicator vector to multiple rounds of random perturbation to obtain multiple perturbed raw material performance indicator vectors; predicting the multiple perturbed raw material performance indicator vectors based on a random forest model to obtain multiple predicted values ​​and the standard deviation corresponding to each predicted value, wherein the predicted values ​​are used to describe product quality indicators; selecting a first sample set of acceptable quality from the multiple perturbed raw material performance indicator vectors based on the predicted values ​​and the acceptable quality range, and selecting a preset number of perturbed raw material performance indicator vectors with high reliability ranking from the first sample set based on the standard deviation to obtain a second sample set; statistically analyzing the perturbation direction of each raw material performance indicator in each perturbed raw material performance indicator vector in the second sample set to obtain the consistency of the perturbation direction of each raw material performance indicator; and determining the raw material performance indicator with the highest consistency in the perturbation direction as the key factor most likely to cause the current quality defect. It has the following beneficial effects: It can automatically and reliably identify the most critical raw material performance factors that cause quality defects in target products from high-dimensional, high-noise industrial production data, thereby achieving accurate and rapid attribution of product quality defects, guiding production adjustments, and improving the efficiency of analyzing which raw material performance indicators cause product quality defects. Attached Figure Description

[0015] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 A flowchart illustrating a product quality defect analysis method provided in this application embodiment; Figure 2 This is a schematic diagram of the internal structure of a product quality defect analysis device provided in an embodiment of this application. Detailed Implementation

[0016] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0017] To facilitate understanding of the following embodiments of this application, the embodiments of this application provide some conceptual explanations and definitions, including: Product Quality Indicators (PQI): refers to a series of quantitative parameters used to evaluate the quality of the final product, denoted as P1, P2, ..., Pn. For example, in the chemical industry, these can be purity, viscosity, and molecular weight; in the metallurgical industry, they can be tensile strength, hardness, elongation, etc.

[0018] Raw Material Performance Indicators (RMPIs) are a series of quantitative parameters used to describe the characteristics of raw materials used in production, denoted as R1, R2, ..., Rm. Examples include the concentration of a chemical solvent, the grade of a metal ore, and the melt flow index of a polymer.

[0019] Initial Raw Material Performance Index (RMPI): Specifically refers to the performance index data corresponding to the raw materials actually used in producing the batch of products with quality defects. The Initial Raw Material Performance Index vector is obtained by vector transformation of the initial raw material performance index.

[0020] Random Forest Model: An ensemble learning algorithm that makes predictions by constructing multiple decision trees, i.e., the random forest regression model (base learners), and then voting or averaging them. In this patent, the model takes raw material performance indicators (R1-Rm) as input and predicts product quality indicators (P1-Pn) as output.

[0021] Prediction Reliability: In this invention, it specifically refers to the standard deviation of the predictions of all base learners for a certain quality metric in the random forest model. A small standard deviation indicates that the prediction results of all trees are relatively consistent, and the prediction reliability is high; a large standard deviation indicates that there is a large divergence between trees, and the prediction reliability is low.

[0022] The Qualified Interval, also known as the acceptable quality range, is a pre-defined acceptable numerical range [P_low, P_high] for the product quality indicator P. Values ​​exceeding this range are considered quality defects.

[0023] Perturbation: This involves making small, random adjustments to the initial raw material performance parameters to simulate variations in performance within the natural range of fluctuations. Perturbation can be additive (R_perturbed = R_initial + Δ, where Δ is a random number) or multiplicative (R_perturbed = R_initial * (1 + Δ)). R_initial is the initial value, and R_perturbed is the perturbation value.

[0024] Virtual Sample: A prediction of product quality generated by perturbing the initial raw material performance indicators and inputting them into a trained random forest model. This sample is not generated in actual production, but rather represents computer-simulated data points illustrating "how the product would be if the raw materials were as they are." The perturbed raw material performance indicators are obtained through vector transformation using the virtual sample.

[0025] Deviation Direction Consistency: For a specific raw material performance index Ri, the degree of consistency among t selected high-reliability virtual samples in terms of whether the perturbation value is generally larger or smaller than the initial value. Higher consistency indicates that adjustments to the raw material index Ri have a more significant and stable effect on correcting product quality defects.

[0026] This application provides a method, device, and medium for analyzing product quality defects, in order to solve the following technical problem: how to achieve more accurate detection of abnormal operating behavior in software systems.

[0027] The technical solutions proposed in the embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0028] Figure 1 This is a flowchart illustrating a product quality defect analysis method provided in an embodiment of this application. Figure 1 As shown in the embodiment of this application, a method for analyzing product quality defects specifically includes the following steps: Step 101: If any quality indicator of the target product is detected to be outside the acceptable quality range, obtain the initial raw material performance index vector corresponding to the product batch to which the target product belongs, and subject the initial raw material performance index vector to multiple rounds of random perturbation to obtain the raw material performance index vector after multiple perturbations.

[0029] In some embodiments, step 101 is implemented by: for a newly produced batch of products, testing all product quality indicators. If a product quality indicator (denoted by P) is found to be outside its acceptable range, an attribution algorithm is triggered. The initial value of each raw material performance indicator in the initial raw material performance indicator vector corresponding to this batch of products is recorded as R_initial = (R1_i, R2_i, ..., Rm_i), where (R1_i, R2_i, ..., Rm_i) represents the initial values ​​of m types of raw material performance indicators.

[0030] Step 102: Based on the random forest model, predict the multiple perturbation raw material performance index vectors to obtain multiple predicted values ​​and the standard deviation corresponding to each predicted value. The predicted values ​​are used to describe product quality indicators.

[0031] Step 103: Based on the predicted value and the quality qualified range, select a first sample set of qualified raw material performance indexes from the multiple perturbation raw material performance index vectors, and based on the standard deviation, select a preset number of perturbation raw material performance index vectors with the highest reliability ranking from the first sample set to obtain a second sample set.

[0032] Step 104: Statistically analyze the perturbation direction of each raw material performance index in the perturbed raw material performance index vector of each item in the second sample set to obtain the consistency of the perturbation direction of each raw material performance index.

[0033] Step 105: The raw material performance index with the highest consistency in the direction of disturbance is determined as the key factor most likely to cause this quality defect.

[0034] This application provides a method for analyzing product quality defects, comprising: when any quality indicator of a target product is detected to exceed the acceptable quality range, obtaining an initial raw material performance indicator vector corresponding to the product batch to which the target product belongs, and subjecting the initial raw material performance indicator vector to multiple rounds of random perturbation to obtain multiple perturbed raw material performance indicator vectors; predicting the multiple perturbed raw material performance indicator vectors based on a random forest model to obtain multiple predicted values ​​and the standard deviation corresponding to each predicted value, wherein the predicted values ​​are used to describe product quality indicators; selecting a first sample set of acceptable quality from the multiple perturbed raw material performance indicator vectors based on the predicted values ​​and the acceptable quality range, and selecting a preset number of perturbed raw material performance indicator vectors with high reliability ranking from the first sample set based on the standard deviation to obtain a second sample set; statistically analyzing the perturbation direction of each raw material performance indicator in each perturbed raw material performance indicator vector in the second sample set to obtain the consistency of the perturbation direction of each raw material performance indicator; and determining the raw material performance indicator with the highest consistency in the perturbation direction as the key factor most likely to cause the current quality defect. It has the following beneficial effects: It can automatically and reliably identify the most critical raw material performance factors that cause quality defects in target products from high-dimensional, high-noise industrial production data, thereby achieving accurate and rapid attribution of product quality defects, guiding production adjustments, and improving the efficiency of analyzing which raw material performance indicators cause product quality defects.

[0035] As a refinement of the embodiments of this application, when performing step 101 to perform multiple rounds of random perturbation on the initial raw material performance index vector to obtain multiple perturbed raw material performance index vectors, the following implementation methods can also be adopted, but are not limited to: adding a random value to the initial value of each raw material performance index in the initial raw material performance index vector to obtain the perturbed raw material performance index vector; randomly selecting a corresponding random value for each raw material performance index and adding it to its initial value to perform a second round of random perturbation, and so on for N rounds of random perturbation to obtain N perturbed raw material performance index vectors, wherein the random value corresponding to each raw material performance index follows a normal distribution with a mean of 0.

[0036] In some embodiments, the initial value R_initial of each raw material performance index in the initial raw material performance index vector is randomly perturbed N times (N is typically 1000-10000 times) to generate N perturbed raw material performance index vectors R_perturbed_k (k=1,2,...,N). The perturbing method can be: adding a random number ε that follows a normal distribution with a mean of 0 to the initial value, i.e., R_perturbed_k = R_initial +ε, ε ~ N(0, σ), where the standard deviation σ can be determined based on the fluctuation range of the raw material index in the production history data.

[0037] In some embodiments, step 102 is further implemented by: inputting N perturbed raw material performance index vectors into a trained random forest model for index P to obtain N predicted values ​​P_pred_k and the corresponding N standard deviations of prediction reliability σ_pred_k (i.e., the standard deviation of the base learner prediction results corresponding to each predicted value).

[0038] As a refinement of the above embodiments, when performing step 103, which involves selecting a first sample set of qualified raw material performance indicators from the multiple perturbation-induced raw material performance indicator vectors based on the predicted value and the qualified quality range, the following implementation methods can also be adopted, for example: selecting the perturbation-induced raw material performance indicator vectors corresponding to the predicted value belonging to the qualified quality range; and constructing the first sample set based on the selected perturbation-induced raw material performance indicator vectors.

[0039] In some embodiments, the screening of the first sample set includes: selecting all samples from the N virtual samples whose predicted value P_pred_k is within the qualified interval [P_low, P_high], and assuming that the number of samples is n (n ≤ N).

[0040] As a refinement of the above embodiments, when performing step 103, which involves selecting a preset number of perturbed raw material performance index vectors with the highest reliability ranking from the first sample set based on the standard deviation to obtain the second sample set, the following implementation methods can also be adopted, but are not limited to: sorting the reliability of the perturbed raw material performance index vectors in the first sample set in ascending order based on the standard deviation, wherein the smaller the standard deviation value, the higher the reliability of the corresponding perturbed raw material performance index vector; selecting a preset number of perturbed raw material performance index vectors with the highest reliability ranking in the ascending order to obtain the second sample set.

[0041] In some embodiments, the screening of the second sample set includes: sorting the above n qualified virtual samples in ascending order according to their prediction reliability σ_pred_k (the smaller the standard deviation, the higher the reliability). Selecting the top t samples with the highest reliability (t can be a fixed value, such as 100; or selected proportionally, such as the top 10%) constitutes the high-reliability qualified sample set, i.e., the second sample set S. Utilizing the advantages of the random forest algorithm in handling high-dimensional features and tolerating noise, the accuracy of the prediction model, i.e., the random forest model, is ensured. Through the "reliability screening" step, noise points with high uncertainty in the model itself are further eliminated, making the attribution analysis based on the most reliable data, greatly improving the accuracy and reliability of the attribution results. By actively perturbing and screening "qualified" virtual samples, the causal scenario of "how to adjust raw materials to obtain qualified products" is simulated, thereby revealing the potential causal relationship more closely.

[0042] As a refinement of the above embodiment, when performing step 104 to statistically analyze the perturbation direction of each raw material performance index in the perturbed raw material performance index vectors of the second sample set and obtain the consistency of the perturbation direction of each raw material performance index, the following implementation methods can also be adopted, but are not limited to: For example, traversing each perturbed raw material performance index vector in the second sample set to obtain the perturbation value corresponding to each raw material performance index, wherein the perturbation value is the sum of the corresponding random value and the initial value; comparing the magnitudes of all perturbation values ​​corresponding to the same raw material performance index in different perturbed raw material performance index vectors with their corresponding initial values; if the If the disturbance value is greater than the initial value, a first marker indicating a larger disturbance direction is output; if the disturbance value is less than the initial value, a second marker indicating a smaller disturbance direction is output; if the disturbance value is equal to the initial value, an ignore marker is output. The first number of the first marker, the second number of the second marker, and the third number of the ignore marker are counted, and the first, second, and third numbers are added together to calculate the total number. The target difference is divided by the total number to obtain the target quotient. The absolute value of the target quotient is used as the disturbance direction consistency score for the same raw material performance index. This process is repeated to obtain the disturbance direction consistency score for each raw material performance index.

[0043] In some embodiments, the calculation process of the perturbation direction consistency score includes, but is not limited to, the following implementation method, for example: for each sample in the second sample set S, traverse each raw material performance index Ri (i=1,2,...,m) and determine the deviation direction of its perturbation value relative to the initial value R_initial: If R_perturbed > R_initial, then mark it as +1 (too large). If R_perturbed < R_initial, mark it as -1 (too small). If they are equal (very low probability), it can be ignored or randomly assigned a value.

[0044] For each raw material performance index Ri, count the deviation direction marks of all samples in the second sample set S.

[0045] Calculate the deviation direction consistency score of each raw material performance index Ri, that is, the perturbation direction consistency score Consistency_i. This score can be defined as: Consistency_i = | (Count_{+1} - Count_{-1}) / t |. The value range of this score is [0, 1]. The larger the value, the more consistent the adjustment direction of this raw material performance index in the S set. t represents the total quantity, Count_{+1} represents the first quantity, and Count_{-1} represents the second quantity.

[0046] Select the raw material performance index R* with the highest consistency score Consistency_i and determine it as the most likely cause of the unqualified product quality index P this time. Its consistent deviation direction is the direction of the corrective measure (for example, if the consistency direction is "too large", it means that this raw material index is "too small" in this production and should be adjusted upwards). The final attribution result not only points out which raw material has a problem, but also indicates the specific adjustment direction (increase or decrease) through the "deviation direction", providing direct and clear guidance for production operations and having strong engineering practical value. The entire process can be automatically completed by a computer, which can complete the analysis work that originally required experts several days or even weeks in a few minutes, greatly improving the troubleshooting efficiency and reducing the production downtime. This algorithm framework does not depend on a specific industry or process and can be widely applied to manufacturing fields such as chemical engineering, pharmaceuticals, metallurgy, and food processing that involve complex relationships between raw materials and product quality.

[0047] As a refinement of the above embodiment, when performing step 105 to determine the raw material performance index with the highest perturbation direction consistency as the key factor most likely to cause this quality defect, the following implementation methods can also be used but are not limited to: using the raw material performance index corresponding to the largest perturbation direction consistency score as the key factor; adjusting the size of the corresponding raw material performance index according to the positive or negative nature of the target quotient corresponding to the largest perturbation direction consistency score to produce the target product using the raw material that meets the adjusted performance index.

[0048] As a refinement of the above embodiments, after taking the raw material performance index corresponding to the largest perturbation direction consistency score as the key factor, the method may also adopt, but is not limited to, the following implementation methods: for example, if the positive and negative properties are positive, then the raw material performance index is increased; otherwise, the raw material performance index is decreased.

[0049] As a refinement of the above embodiments, before predicting the multiple perturbed raw material performance index vectors based on the random forest model, the method may also adopt, but is not limited to, the following implementation methods, for example: obtaining historical production data of the target product, the historical production data including historical raw material performance indexes and historical product quality indexes; preprocessing the historical production data, and training a random forest regression model separately for each product quality index based on the preprocessed historical production data to obtain the random forest model.

[0050] In some embodiments, the training process of the random forest model includes: collecting historical production data, the dataset of which should contain raw material performance indicators (R1, R2, ..., Rm) and corresponding product quality indicators (P1, P2, ..., Pn) for each batch of products. Data preprocessing includes data cleaning (handling missing values ​​and outliers) and standardization (Z-score standardization or Min-Max normalization). For each product quality indicator Px (x=1,2,...,n) to be monitored, a separate random forest regression model is trained. This model uses R1 to Rm as features and Px as the prediction target. Model hyperparameters (such as the number of trees, maximum depth, etc.) are optimized through cross-validation. A pass / fail interval [Px_low, Px_high] is defined for each product quality indicator Px.

[0051] In summary, the embodiments of this application can achieve the following effects, including: 1. Automatically and reliably identify the most critical raw material performance factors causing quality defects in target products from high-dimensional, high-noise industrial production data, thereby achieving accurate and rapid attribution of product quality defects, guiding production adjustments, and improving the efficiency of analyzing which raw material performance indicators cause product quality defects.

[0052] 2. A random forest model is trained offline and subjected to online perturbation simulation analysis. The standard deviation of the random forest model's predicted values ​​is used as a quantitative indicator of prediction reliability and to screen high-confidence virtual samples. First, "qualified" samples are screened, then "reliable" samples are screened, progressing step by step to ensure that the sample set for subsequent analysis possesses both the key attributes of "correct target" and "credible prediction." On the selected high-quality, high-reliability sample set, the consistency of the raw material perturbation direction is statistically analyzed, transforming the attribution problem into a clear statistical comparison problem. This method is simple and effective.

[0053] The above are embodiments of the method proposed in this application. Based on the same inventive concept, embodiments of this application also provide a product quality defect analysis device, the structure of which is as follows: Figure 2 As shown.

[0054] Figure 2 This is a schematic diagram of the internal structure of a product quality defect analysis device provided in an embodiment of this application. Figure 2 As shown, the device includes: At least one processor 201; And a memory 202 that is communicatively connected to at least one processor; The memory 202 stores instructions executable by at least one processor. These instructions are executed by at least one processor 201 to enable the processor 201 to: upon detecting that any quality indicator of the target product exceeds the acceptable quality range, acquire the initial raw material performance indicator vector corresponding to the product batch to which the target product belongs, and subject the initial raw material performance indicator vector to multiple rounds of random perturbation to obtain multiple perturbed raw material performance indicator vectors; predict the multiple perturbed raw material performance indicator vectors based on a random forest model to obtain multiple predicted values ​​and the standard deviation corresponding to each predicted value, where the predicted values ​​describe the product quality indicators; based on the predicted values ​​and the acceptable quality range, select a first set of acceptable quality samples from the multiple perturbed raw material performance indicator vectors, and based on the standard deviation, select a predetermined number of perturbed raw material performance indicator vectors with the highest reliability ranking from the first set to obtain a second set; statistically analyze the perturbation direction of each raw material performance indicator in the perturbed raw material performance indicator vectors in the second set to obtain the consistency of the perturbation direction of each raw material performance indicator; and determine the raw material performance indicator with the highest consistency in the perturbation direction as the key factor most likely to cause this quality defect.

[0055] Some embodiments of this application provide corresponding to Figure 1A non-volatile computer storage medium stores computer-executable instructions, which are configured to: when any quality indicator of a target product is detected to exceed the acceptable quality range, obtain an initial raw material performance indicator vector corresponding to the product batch to which the target product belongs, and subject the initial raw material performance indicator vector to multiple rounds of random perturbation to obtain multiple perturbed raw material performance indicator vectors; predict the multiple perturbed raw material performance indicator vectors based on a random forest model to obtain multiple predicted values ​​and the standard deviation corresponding to each predicted value, wherein the predicted values ​​are used to describe product quality indicators; based on the predicted values ​​and the acceptable quality range, select a first sample set of acceptable quality from the multiple perturbed raw material performance indicator vectors, and based on the standard deviation, select a preset number of perturbed raw material performance indicator vectors with high reliability ranking from the first sample set to obtain a second sample set; statistically analyze the perturbation direction of each raw material performance indicator in each perturbed raw material performance indicator vector in the second sample set to obtain the consistency of the perturbation direction of each raw material performance indicator; and determine the raw material performance indicator with the highest consistency in the perturbation direction as the key factor most likely to cause the current quality defect.

[0056] The various embodiments in this application are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the embodiments for IoT devices and media are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0057] The systems, media, and methods provided in this application are one-to-one correspondences. Therefore, the systems and media also have similar beneficial technical effects as their corresponding methods. Since the beneficial technical effects of the methods have been described in detail above, the beneficial technical effects of the systems and media will not be repeated here.

[0058] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0059] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0060] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0061] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0062] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0063] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0064] Computer-readable media include both permanent and non-permanent, removable and non-removable media that can store information by any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0065] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0066] The above description is merely an embodiment of this application and is not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A method for analyzing product quality defects, characterized in that, The method includes: If any quality indicator of the target product is detected to be outside the acceptable quality range, the initial raw material performance index vector corresponding to the product batch to which the target product belongs is obtained, and the initial raw material performance index vector is subjected to multiple rounds of random perturbation to obtain a raw material performance index vector after multiple perturbations. Based on the random forest model, the vector of raw material performance indicators after multiple perturbations is predicted to obtain multiple predicted values ​​and the standard deviation of each predicted value. The predicted values ​​are used to describe product quality indicators. Based on the predicted value and the quality qualified range, a first sample set of qualified raw material performance indexes is selected from the multiple perturbed raw material performance index vectors, and a preset number of perturbed raw material performance index vectors with the highest reliability ranking are selected from the first sample set based on the standard deviation, to obtain a second sample set. The perturbation direction of each raw material performance index in the perturbed raw material performance index vector of each item in the second sample set is statistically analyzed to obtain the consistency of the perturbation direction of each raw material performance index. The raw material performance index with the highest consistency in the direction of disturbance was identified as the key factor most likely to cause this quality defect.

2. The method for analyzing product quality defects according to claim 1, characterized in that, The process of subjecting the initial raw material performance index vector to multiple rounds of random perturbation to obtain multiple perturbed raw material performance index vectors includes: A random value is added to the initial value of each raw material performance index in the initial raw material performance index vector to obtain the perturbed raw material performance index vector. For each of the raw material performance indicators, a corresponding random value is randomly selected and added to its initial value for a second round of random perturbation. This process is repeated for N rounds of random perturbation to obtain N perturbed raw material performance indicator vectors. The random value corresponding to each of the raw material performance indicators follows a normal distribution with a mean of 0.

3. The method for analyzing product quality defects according to claim 2, characterized in that, The first sample set of qualified raw material performance indicators selected from the multiple perturbed raw material performance index vectors based on the predicted value and the qualified quality range includes: Filter out the perturbed raw material performance index vector corresponding to the predicted value that belongs to the quality qualified range; The first sample set is constructed based on the selected perturbation-induced raw material performance index vector.

4. The method for analyzing product quality defects according to claim 3, characterized in that, The step of selecting a predetermined number of perturbed raw material performance index vectors with the highest reliability ranking from the first sample set based on the standard deviation to obtain the second sample set includes: The reliability of the perturbed raw material performance index vector in the first sample set is sorted in ascending order based on the standard deviation. The smaller the standard deviation, the higher the reliability of the perturbed raw material performance index vector. Selecting a predetermined number of the perturbed raw material performance index vectors from the ascending order yields the second sample set.

5. The method for analyzing product quality defects according to claim 4, characterized in that, The perturbation direction of each raw material performance index in the perturbed raw material performance index vector of each item in the second sample set is statistically analyzed to obtain the consistency of the perturbation direction of each raw material performance index, including: For each of the perturbed raw material performance index vectors in the second sample set, the perturbation value corresponding to each raw material performance index is obtained by traversing the vector. The perturbation value is the sum of the corresponding random value and the initial value. Compare the magnitudes of all perturbation values ​​corresponding to the same raw material performance index in the different perturbation raw material performance index vectors with their corresponding initial values. If the disturbance value is greater than the initial value, a first marker indicating a larger disturbance direction is output; if the disturbance value is less than the initial value, a second marker indicating a smaller disturbance direction is output; if the disturbance value is equal to the initial value, an ignore marker is output. Count the first number of the first marker, the second number of the second marker, and the third number of the ignored markers, and add the first number, the second number, and the third number to calculate the total number; Divide the target difference by the total quantity to obtain the target quotient. Use the absolute value of the target quotient as the perturbation direction consistency score for the same raw material performance index. Repeat this process to obtain the perturbation direction consistency score for each raw material performance index.

6. The method for analyzing product quality defects according to claim 5, characterized in that, The raw material performance indicators with the highest consistency in disturbance direction were identified as the key factors most likely to cause this quality defect, including: The raw material performance index corresponding to the largest perturbation direction consistency score is taken as the key factor. The corresponding raw material performance indicators are adjusted according to the positive or negative nature of the target quotient corresponding to the maximum perturbation direction consistency score, so as to use raw materials that meet the adjusted performance indicators to produce the target product.

7. The method for analyzing product quality defects according to claim 6, characterized in that, After taking the raw material performance index corresponding to the largest perturbation direction consistency score as the key factor, the method includes: If the positive or negative property is positive, the raw material performance index is increased; otherwise, the raw material performance index is decreased.

8. A method for analyzing product quality defects according to any one of claims 1-7, characterized in that, Before predicting the vector of perturbed raw material performance indicators based on the random forest model, the method includes: Obtain historical production data for the target product, including historical raw material performance indicators and historical product quality indicators; The historical production data is preprocessed, and a random forest regression model is trained separately for each product quality indicator based on the preprocessed historical production data to obtain the random forest model.

9. A product quality defect analysis device, characterized in that, The device includes: At least one processor; And, a memory communicatively connected to the at least one processor; The memory stores instructions that can be executed by the at least one processor, which, when executed by the at least one processor, enables the at least one processor to perform a product quality defect analysis method as described in any one of claims 1-8.

10. A computer storage medium storing computer-executable instructions, characterized in that, When the computer-executable instructions are executed, they implement a product quality defect analysis method as described in any one of claims 1-8.

Images