一种可能的解决方法是升级到Flink的版本大于1.11.2，并在Vectorized运算符配置文件中使用BatchPhysicalTypeCheckStrategy参数。下面是代码示例：

val conf = new Configuration()
conf.setString("table.exec.resource.default-parallelism", "1") 
val env = StreamExecutionEnvironment.createLocalEnvironmentWithWebUI(conf)

val streamSettings = EnvironmentSettings.newInstance()
      .useBlinkPlanner()
      .inStreamingMode()
      .build()

val tEnv = StreamTableEnvironment.create(env, streamSettings)

val srcDDL =
      """
        |CREATE TABLE src (
        |    id  BIGINT,
        |    name STRING,
        |    age INT,
        |    score DOUBLE
        |) WITH (
        |   'connector' = 'datagen'
        |)
      """.stripMargin
tEnv.executeSql(srcDDL)

val sinkDDL =
      """
        |CREATE TABLE sink (
        |    id  BIGINT,
        |    name STRING,
        |    age INT,
        |    score DOUBLE,
        |    invalid BOOLEAN
        |) WITH (
        |   'connector' = 'print'
        |)
      """.stripMargin
tEnv.executeSql(sinkDDL)

val func = new ExampleFunction

tEnv.createTemporaryFunction("func", func)

val query =
      """
        |SELECT
        |    id,
        |    name,
        |    age,
        |    score,
        |    func(name) as invalid
        |FROM src
      """.stripMargin

tEnv.executeSql(query).print()

使用以上代码示例，UDF应该可以正常工作。