基于多源异构数据的氧化铀陶瓷表面缺陷智能检测数据
收藏资源简介：
本数据为氧化铀陶瓷产品缺陷检测数据，共包含15个字段，涵盖产品标识、检测参数、计算系数、结果判定及执行指令五大类信息，用于完整记录检测流程及结果。检测序号为字符串类型，作为产品唯一检测标识，可精准区分不同检测样本；检测时间为日期型，记录产品具体检测时间。图像检测的缺陷面积占比S1、声波信号异常幅度S2均为浮点型，分别由图像、声波传感器采集；缺陷密度D为浮点型，单位为个/cm²，表征每平方厘米内产品缺陷数量；缺陷连续性指数C为浮点型，基于形态学分析得出；产品厚度T为浮点型，单位为mm，记录产品实际厚度。w1、w2为浮点型权重系数，分别对应图像、声波传感器，用于缺陷存在概率计算；α、β、γ均为浮点型模型调节系数，经大量数据训练优化得出。缺陷存在概率P为字符串类型，按公式P=α*{（w1*S1+w2*S2）/（w1+w2）}+β*log(D+1.3)+γ*C/T计算得出，格式为百分比；缺陷风险等级为字符串类型，依据P值划分为低度、中度、高度缺陷风险三类；分拣指令为字符串类型，对应风险等级触发对应动作，分别为自动放行至包装、人工复检、剔除高风险产品。
This dataset contains defect detection data for uranium oxide ceramic products, with a total of 15 fields covering five categories of information: product identification, detection parameters, calculation coefficients, result judgment, and execution instructions, designed to fully record the entire detection workflow and its outcomes. The detection serial number is a string-type field serving as the unique identifier for each detection, enabling accurate distinction between different test samples. The detection time is a date-type field that records the exact timestamp when the product was inspected. The defect area ratio S1 obtained via image detection and the abnormal acoustic signal amplitude S2 are both floating-point fields, collected by image and acoustic sensors respectively. The defect density D is a floating-point field with the unit of counts/cm², representing the number of product defects per square centimeter. The defect continuity index C is a floating-point field derived through morphological analysis. The product thickness T is a floating-point field with the unit of mm, recording the actual physical thickness of the product. w1 and w2 are floating-point weight coefficients corresponding to the image and acoustic sensors respectively, used to calculate the defect presence probability. α, β, and γ are all floating-point model adjustment coefficients optimized and obtained through training on large-scale datasets. The defect presence probability P is a string-type field calculated using the formula: P = α * ((w1*S1 + w2*S2)/(w1 + w2)) + β * log(D + 1.3) + γ * (C/T), with its format being a percentage value. The defect risk level is a string-type field categorized into three classes based on the P value: low, medium, and high defect risk. The sorting instruction is a string-type field that triggers corresponding actions according to the risk level, specifically automatically releasing products to packaging, performing manual re-inspection, and rejecting high-risk products respectively.
