Despite decades of diligent work on their development, line-by-line (LBL) spectroscopic information systems, widely employed by scientists and engineers, may still contain a number of incorrect rovibronic lines. A novel heuristic protocol, relying on cycle bases of spectroscopic networks (SN) and a system of linear constraints, is proposed to unravel incorrect transitions present in the database. The algorithm is named autoECART, standing for automatic Energy Conservation Analysis of Rovibronic Transitions. The autoECART method is tested on synthetic SNs constructed from spectroscopic databases of nine water isotopologues. The systematic numerical tests demonstrate the outstanding capability of the autoECART procedure to identify almost all of the outliers generated randomly in these synthetic SNs. As a ‘real-life’ example, the GEISA-2019-H${}_2$${}^{18}$O database is scrutinized, revealing 17 rough outlier lines. Developers of spectroscopic information systems are encouraged to utilize autoECART for cleansing the huge number of transitions deposited in LBL databases.

尽管对其开发进行了数十年的辛勤工作，科学家和工程师广泛采用的逐行 (LBL) 光谱信息系统仍可能包含许多不正确的 rovibronic 谱线。提出了一种新的启发式协议，它依赖于光谱网络 (SN) 的循环基础和线性约束系统，以解开数据库中存在的不正确转换。该算法被命名为 autoECART，代表 Rovibronic Transitions 的自动能量守恒分析。autoECART 方法在由九种水同位素的光谱数据库构建的合成 SN 上进行测试。系统的数值测试证明了 autoECART 程序在识别这些合成 SN 中几乎所有随机生成的异常值方面的出色能力。作为一个“现实生活”的例子，GEISA-2019-H${}_2$${}^{18}$O 数据库被仔细检查，揭示了 17 条粗略的异常值线。鼓励光谱信息系统的开发人员利用 autoECART 来清理存储在 LBL 数据库中的大量跃迁。