基于场动力理论的施工工人行为安全智能管理研究

方琦

摘要：我国建筑行业的事故量目前仍然处于高位,其中绝大部分事故由工人的不安全行为造成。为了减少工人不安全行为引起的事故,提高建筑业安全管理水平,本文以Behavior-Based Safety（BBS）方法为基础,结合人工智能技术与心理学场动力理论,研究工人不安全行为形成机理,并对工人不安全行为进行观测、分析和干预管理。具体研究内容如下:（1）施工工人不安全行为形成机理。以场动力理论为基础,结合心理空间与心理动力在施工工人不安全行为研究中的内涵与意义,明确不安全行为的影响要素包括心理场因素以及以自控力为特征的人格特质,指出不安全行为的产生是心理动力与自控力水平的综合作用结果,并据此建立施工领域工人行为安全分析模型。（2）施工工人不安全行为智能化检测方法。鉴于不同场景下施工行为规范具有较大差别,提出基于场景的工人不安全行为智能检测方法。首先分别从场景关键对象的识别、定位以及空间范围划分等,提出施工场景自动识别方法;接着,在深度学习算法的基础上,结合建筑施工领域应用特点,设计多个施工不安全行为基础算法模块;再根据施工不安全行为判定规则的语义特点,利用规则拆解建立“不安全行为自动化识别任务”与“基于深度学习的施工不安全行为基础算法模块”之间的程序调用与组合机制,最终实现不同场景下的工人不安全行为智能识别。（3）施工工人行为安全分类管理策略。考虑到施工工人的行为安全模式具有较强的个性化特征与显著的个体差异,本文通过行为安全分析模型对工人行为安全档案中的记录进行挖掘处理,分析工人自控力特征。设立安全绩效考核、分类引导疏导等分类管理办法,根据工人自控力特征智能地推送对应管理策略,辅助工人不安全行为矫正。（4）施工工人行为安全分类管理实例研究。在某地地铁的三个工地开展实例研究,验证实验组与对照组施工不安全行为率改善效果是否具有显著差异,并通过对比采取不同管理策略前后工人自控力水平的变化,分析分类管理对工人行为矫正的效果。结果表明,采用具有针对性的分类行为安全管理策略比统一策略更能有效提升工人行为安全水平。本文将场动力理论与人工智能技术应用于施工工人行为安全管理,梳理不安全行为产生的根本原因和致因路径,实时获取工人不安全行为和心理场指标记录,分析工人行为模式,采用分类管理策略,有助于深入探究工人行为安全规律,矫正工人不安全行为。

关键词：场动力理论；不安全行为检测；行为安全管理；深度学习；自控力特质

ABSTRACT

The construction industry is a high hazard industry. For years, construction has led all industries in the total number of worker deaths, most of which are caused by unsafe behaviors of workers. Behavior-Based Safety (BBS) has always been an important research topic to reduce accidents and facilitate safety management. On the basis of the traditional BBS program, this study has introduced artificial intelligence technology and psychological field theory to support the root cause analysis and automatic observation and management for workers' unsafe behavior.

(1) An unsafe behavior formation mechanism for construction workers. According to the relevant concepts of field theory, a behavioral safety mechanism analysis model of construction workers is proposed. This dissertation has illustrated the connotation and influences of psychological space and motivating forces on workers' behaviors. It is also pointed out that the vector sum of the motivating forces of the psychological field and the self-control of personality traits ultimately determines whether unsafe behaviors occur or not. It provides a basis for the following analysis and classified management of self-control personality traits.

(2) Intelligent  detection  method  for  construction  workers'  unsafe  behaviors.Considering the great differences of construction behavior specifications in different scenarios, this dissertation proposes a scenario based intelligent detection method for construction workers' unsafe behaviors. Firstly, the construction scenarios are defined and detected according to the locations and space division of central objects. Then, the calling and combination mechanisms of the vision-based subprogram modules are established to solve the automatic recognition task for construction workers' unsafe behaviors.

(3) A classified management system for construction workers' unsafe behaviors.According to the behavior intervention theory, a safety performance evaluation system and a "4I" based classified management system are designed for workers with different self-control personality traits. It aims to motivate the workers to work in a safe manner, help them with sufficient support, and finally reshape and improve workers' behavioral safety habits. In the design of the personalized behavioral safety management system, full consideration is given to the personal characteristics of their self-control traits, and specific management measures are formulated accordingly to achieve personalized correction of unsafe behaviors of workers.

(4) Case study. An intelligent identification and classification management system for behavior safety which is used to automatically collect and obtain workers' behavior safety data was developed. Then, a comparative experiment was carried out on three construction sites. It had been verified whether there is a significant difference between the experimental group and the control group in reducing the rate of unsafe behaviors. The performance of the classified management system was analyzed by comparing the changes of workers' self-control level before and after adopting classified management strategies.    In this study, psychological field theory and artificial intelligence technology are applied to behavior-based safety management of construction workers. Root causes and paths of unsafe behaviors are discussed. Unsafe behavior records and psychological field parameters are obtained to mine behavior patterns of workers. Classified management strategies are adopted to help reduce unsafe behaviors of workers.

Keywords: Psychological Field Theory; Unsafe Behavior Detection; Behavior-Based Safety Management; Deep Learning; Self-control Trait